Opinion for the Court filed by Circuit Judge WALD.
WALD, Circuit Judge:
The Staggers Rail Act of 1980, Pub.L.No. 96-4M8, 95 Stat. 1895 (to be codified in scattered sections of 11, 45 & 49 U.S.C.) (“Staggers Act” or “Act”) provides the Interstate Commerce Commission (“ICC” or “Commission”) with jurisdiction to review the reasonableness of rates charged by “market dominant” 1 railroads. Under Title II of the Act, however, certain railroad rates and rate increases are deemed reasonable and are thus immune from Commission review. The issues in this case are (1) whether the formula adopted by the Com[330]*330mission for translating cost increases into presumptively reasonable rate increases comports with the statutory description for such a formula and (2) whether the Commission acted arbitrarily and capriciously in selecting among possible formulae. For the reasons set forth below, we find that the Commission acted within its statutory authority and that it offered an adequate rationale for adopting the specific inflationary cost index at issue in this case.
I. BACKGROUND
A. Statutory Context
Under the Staggers Act, market dominance of a railroad in a particular market is determined by the ratio of rates to variable costs. If this “revenue-variable cost percentage” exceeds a certain level, the Commission has jurisdiction to review the reasonableness of the rate charged. See 49 U.S.C.A. §§ 10701a(b)(1), 10709(d)(2) (West Supp.1981). This threshold test, set initially at 160% for the year ending September 30, 1981, will increase to 175% for the year ending September 30, 1984.2
Rates that met this test of market dominance at the time of the Staggers Act’s enactment could have been challenged within 180 days of the passage of the Act. See 49 U.S.C.A. § 10701a note (West Supp. 1981). Rates determined in such a proceeding became “base rates” for the applicable service. Similarly, rates that were not challenged or were not initially subject to challenge also became base rates. Pursuant to section 203 of the Act, 49 U.S.C.A. § 10707a (West Supp.1981), these base rates are to be adjusted annually in accordance with an inflation cost index developed or verified by the Commission. The resulting “adjusted base rate” is not necessarily a rate that is actually charged by any railroad. Instead, it serves as a measuring stick by which the Commission considers the reasonableness of rate increases.3 Those rate increases that simply follow the inflation adjustment formula are immune from attack. See 49 U.S.C.A. § 10707a(b)(2) (West Supp.1981). Those that exceed the inflation adjustment may be challenged, but the burden of proof in a rate proceeding and the Commission’s authority to suspend and investigate rate increases depend on the size of the rate increase as a percent of the adjusted base rate. For the four years following passage of the Act, additional rate increases of six percent or less of the adjusted base rate fall within the railroad’s “zone of rate flexibility.” For these rate increases, the Commission generally lacks the power to suspend or investigate rates4 and the burden of proof in a rate proceeding rests on the one challenging the rate’s reasonableness.5 See 49 U.S.C.A. § 10701a(b) (West Supp.1981). For larger rate increases, the Commission may suspend and investigate rates and the burden of proof in a rate proceeding rests on the railroad.6 Because the “zone of rate flexibility” applies to rate increases above the adjusted base rate, the magnitude of the underlying inflation adjustment is of [331]*331critical concern to both shippers and railroads.7
B. Adoption of the Modified AAR Index
The Commission’s search for an index that would properly measure inflationary cost increases preceded the enactment of the Staggers Act. On April 22, 1980, the Commission issued an Advance Notice of Proposed Rulemaking in Ex Parte No. 290 (Sub. No. 2), Railroad Cost Recovery Procedures soliciting suggestions for measuring average cost increases for railroads. The Commission explained that it wished to establish a mechanism by which railroads could recover increased costs without being subjected to the length and uncertainty of individualized rate proceedings. See 45 Fed.Reg. 29,103-04 (1980); Joint Appendix (“J.A.”) 42—43. In response, the Commission received extensive comments from railroads, shippers, and purchasers of commodities shipped by rail. The railroads proposed that the Commission adopt the Association of American Railroad’s (“AAR”) index of rail costs. See Verified Statement of Dr. Paul O. Roberts & Dr. Brian C. Kullman, J.A. 52; Verified Statement of Dr. Harvey A. Levine, J.A. 420-24; Verified Statement of Mary Maher, J.A. 431—40. The shippers did not propose a specific cost index. They did suggest, however, that if the Commission adopted the AAR index it should adjust that index so that it would reflect cost savings from productivity gains. See Verified Statement of Leroy E. Peabody, Jr., J.A. 389-400.
The AAR index is a classic Laspeyre’s price index,8 which measures the changing cost of buying a fixed “market basket” of items purchased by the railroad industry. AAR selected items for the index’s market basket on the basis of surveys of purchases by major railroads. The index includes 65 items that AAR found to be widely purchased in large quantities and whose price trends follow those of related materials. See Verified Statement of Mary Maher, J.A. 434. AAR updates the index’s “market basket” periodically to reflect shifts in the typical railroad market basket. In the past, this updating has occurred every six to ten years. See id., J.A. 441.9
At the heart of the shippers’ criticism of the AAR index was the difference between the prices railroads pay to purchase inputs (such as labor and materials) and the costs they incur in producing their output (rail service). The shippers argued that a proper measure of railroad “costs” must account for the possibility that productivity gains allow railroads to produce more rail service with the same “market basket” of railroad inputs. By accounting for these gains, they suggested that an input index (such as the AAR index) could and should be converted into an “output index” — namely an index of cost per unit of output.10
Before the Commission completed its proceedings on inflationary rate adjustments, [332]*332Congress passed the Staggers Act. Section 203 of the Act requires the Commission to develop or verify an inflationary cost index with “appropriate adjustments to reflect the changing composition of railroad costs, including the quality and mix of material and labor .. .. ” 49 U.S.C.A. § 10707a(a)(2)(B) (West Supp.1981). In accordance with this mandate, the Commission issued a new Notice of Proposed Rule-making in Ex Parte No. 290 (Sub. No. 2), 45 Fed.Reg. 81,217 (Dec. 10, 1980); J.A. 587-99. In this notice, the Commission stated that it found the “most acceptable measure of inflation for [its] purposes to be the AAR input price index, as modified by the Commission.” J.A. 589.11 The Commission stated its intention to review the statistical validity of the AAR index, check for data errors,12 and compare the index with other data sources. But the Commission rejected the suggestion that the index should be modified to reflect productivity gains. While it acknowledged the “theoretical validity” of the shippers’ argument that “productivity gains tend to offset the upward pressure of input prices on overall rail costs,” it concluded that several considerations counseled against adoption of such an index. J.A. 591. First, it found that there is no acceptable measure of productivity available to convert an input index into an output index. Second, it noted that competition among railroads ought to curb price increases caused by an inflated cost index. Third, the Commission expressed its concern that downward adjustments for productivity gains would act as a disincentive to productivity growth. J.A. 591-92.
Following further comments, the Commission issued a final rule adopting the modified AAR index as initially proposed. See 364 I.C.C. 841 (1981); J.A. 928. In the explanation accompanying this rule, the Commission reexamined its original rationale for not adjusting the AAR index to reflect productivity gains. First, the Commission conceded that methodological problems do not pose an absolute bar to developing measures of productivity; but it noted that these problems were nonetheless “grave.” See id. at 846; J.A. 931. As to the effect of a productivity deflator on incentives for productivity growth, the Commission stated its agreement with the shippers that an adjustment reflecting average productivity trends would not eliminate all incentives for individual railroads to maximize their own productivity gains. Id. at 847-48; J.A. 931-32. But the Commission said that this argument cuts two ways. While the incentive to achieve better than average productivity gains remains, those railroads that are unable to meet the industry’s average performance would be unable to recover (through non-reviewable rate increases) the full measure of their inflation-induced increased costs. Finally, the Commission expressed its skepticism that, even if rates were set by competitive forces, productivity gains would be passed along to consumers. It reasoned that under competitive conditions, firms must retain revenues from cost savings if they are earning insufficient revenues to cover costs. Because it had elsewhere found that the rail industry typically earns inadequate revenues to cover economic costs, it concluded that it would be inappropriate to pass along productivity-related savings. Id. In summation, the Commission stated that “as long as the railroad industry continues to earn inadequate revenues,” it would not adjust the AAR input index to reflect productivity. Id.
II. ANALYSIS
A. Consistency With the Staggers Act
The first question before us is whether the Commission properly interpreted its [333]*333statutory mandate. The Commission stated:
[T]he Rail Act does not specify whether an input cost index or an output cost index is to be used as the rail cost adjustment factor. Instead, the Rail Act leaves this matter up to the Commission, as a matter of policy, to be determined in a manner consistent with the purpose of the Rail Act as a whole.
364 I.C.C. at 846; J.A. 931. Turning to the considerations that should guide its policy determination, the Commission explained that “[in] broad terms, . . . the purpose of section 203 is to permit rail carriers to neutralize the effects of inflation through rate increases with minimum regulatory intervention.” Id. After discussing why an input index properly serves this objective under current conditions, the Commission carefully preserved its authority to require productivity adjustments in the future, stating that “[it] will not adjust the rail cost index to reflect productivity gains as long as the railroad industry continues to earn inadequate revenues.” Id. at 847 (emphasis added); J.A. 931.
In their petitions for review, the Western Coal Traffic League (“WCTL”)13 and the Chemical Manufacturer’s Association (“CMA”), joined by several intervenors,14 argue that the Act does not leave the Commission with discretion to choose between output and input cost indices.15 They contend that the plain meaning of section 203, its legislative history, and the overall structure of the Act require that the Commission develop an index that reflects productivity changes. A third view of section 203 is presented by the Association of American Railroads (“AAR”) as intervenor-respondent. AAR argues that the Commission was required to adopt an input index. AAR Brief 32. Indeed, at some points in their brief, the Commission’s attorneys appear to be sliding into this third position, see Government’s Brief 20, even though we read the Commission’s decision as clearly finding room under the Act to allow for productivity adjustments to the inflationary cost index at some future time.
1. The “Plain Meaning” Debate
All sides of this controversy lay claim to the plain meaning of section 203. As we have often observed, the meaning of some statutory provisions is sufficiently clear to resolve disputes over their interpretation. See, e.g., McCord v. Bailey, 636 F.2d 606, 614-15 (D.C.Cir.1980), cert. denied, 451 U.S. 983, 101 S.Ct. 2314, 68 L.Ed.2d 839 (1981). The language of section 203, however, does not fit into that category. In particular, it does not point us towards a specific statistical methodology for calculating inflationary cost increases.
Section 203 requires that:
Commencing with the fourth quarter of 1980, the Commission shall, as often as practicable but in no event less often than quarterly, publish a rail cost adjust[334]*334ment factor which shall be a fraction, the numerator of which is the latest published Index of Railroad Costs (which index shall be compiled or verified by the Commission, with appropriate adjustments to reflect the changing composition of railroad costs, including the quality and mix of material and labor), and the denominator of which is the same index for the fourth quarter of 1980, or for the fourth quarter of 1982 or for the fourth quarter of every fifth year thereafter, as appropriate.
49 U.S.C.A. § 10707a(a)(2)(B) (West Supp. 1981) (emphasis added). CMA suggests that this language clearly indicates Congress’ concern with railroad costs. Thus, “the index must be adjusted to reflect cost changes due to changes in the mix of expenses incurred by railroads, including the use of less labor or fuel due to more efficient technology and operating methods.” CMA Brief 14. In short, an adequate measure of changing costs must account for more efficient uses of input items through productivity gains. The Commission’s attorneys, in an argument we find somewhat at odds with the Commission’s articulated position, contend that the common sense meaning of section 203 requires adoption of an input index. They state that “the natural, common sense meaning of the statutory phrase ‘composition of railroad costs’ is the formation of a total cost by putting together the different prices paid by railroads for the goods and services they purchase.” Government’s Brief 19. After noting that “the prices of labor and materials mentioned in the statute are indisputably inputs,” they state that: “Because none of the words in the statute embody concepts related to output, the statute cannot be read to embody productivity, which is a concept inextricably linked to output.” Government’s Brief 19.
We find neither interpretation of the “plain meaning” of section 203 very persuasive.16 Turning first to the Commission counsel’s argument, it is hardly self-evident that the required adjustments for “quality and mix” are not output-related. For example, gains in labor productivity could be considered an improvement in the “quality” of labor. Similarly, the “mix” of labor and materials could refer to total quantities of inputs, and not simply the proportionate share of different types of inputs. If either of these interpretations is correct, a productivity adjustment would fall within the adjustments authorized by the Act. But a more basic problem with the Commission’s argument is that it sets up an artificially neat dichotomy between “input” and “output” indices.17 Although the two types of indices are theoretically different — with input indices measuring the cost of a “market basket” of inputs and outputs indices measuring cost per unit of output — an “output” index is still critically concerned with the cost of inputs. The main difference between the two approaches is that an output index looks at additional factors that affect a firm’s cost per unit of output. The fact that an output index would take additional steps to determine cost per unit of output— steps that, arguably, are not specifically mentioned in section 203 — does not mean that Congress meant to preclude such adjustments. Congress drafted section 203’s mandate for appropriate adjustments with inclusive, not exclusive, language. To read it as “plainly” rejecting an index reflecting [335]*335all factors bearing on cost per unit of output is unwarranted.
CMA’s “plain” reading is equally faulty. CMA asserts that a true cost index must include all factors bearing on cost per unit of output, even though the AAR index stands as a stark reminder that not all such indices make the fine adjustments it recommends. Furthermore, CMA overlooks the full statutory language of section 203, which only requires “appropriate” adjustments. Since one critical issue in this case is whether estimates of productivity gains on a quarterly basis are feasible, CMA’s contention that productivity is a cost factor, even if accepted, is insufficient to prove that a productivity adjustment is “appropriate” and, therefore, statutorily mandated.18
2. Legislative History of Section 203
a. Redrafting of Section 203
As further support for their respective interpretations of section 203, the parties turn to the history of its revision in committee. The original version of S. 1946, the bill that became the Staggers Act, described the inflationary cost index as an “Index of Railroad Material Prices and Wages.” See S. 1946, 96th Cong., 2d Sess. § 103 (1980). This description clearly pointed towards an input index, if not the AAR index itself.19 Following hearings on S. 1946, during which shippers advocated adoption of an output index, the bill was redrafted to refer to the inflationary cost index as an “Index of Railroad Costs [with] appropriate adjustments to reflect the quality and mix of material and labor.”20 In arguments paralleling their analysis of the “plain meaning” of section 203, petitioners, the Commission, and the railroads, seek to draw from this redrafting of section 203 (section 103 of S. 1946) unequivocal support for their conceptions of the proper index. Petitioners argue that the committee redrafted section 203 in response to the concerns of shippers, and therefore intended to reject input indices. The Commission and the railroads argue that the committee redrafted section 203 only to provide a more general description of an input index; they reason that since the committee failed to draft in any specifically output-related concepts, its actions should be read as a rejection of output indices. Frankly, we find it difficult to agree with either of these rigid theories of what the Senate Committee revisions meant. What little can be discerned from the redrafting of section 203 indicates to us only that the committee sought to replace specific language pointing towards an input index with a more general mandate to the Commission to select an index with “appropriate” adjustments. Whether adjustments serving to convert an input index into an output index are authorized or required must therefore depend largely on whether [336]*336they serve the underlying purposes of the statutorily mandated index.
Section 203 attracted considerable attention during hearings on S. 1946. Numerous witnesses appearing before the Senate Committee on Commerce, Science, and Transportation argued that language describing an input index was inappropriate if the purpose of indexing base rates was simply to neutralize the effect of cost increases on railroad profits. These witnesses explained that the bill’s formula was “not flexible enough to recognize that certain carriers may come up with innovation, that change in volume, that change in traffic mix should be considered in the automatic adjustment.” See 1 Railroad Transportation Policy Act of 1979: Hearings on S. 1946 Before the Committee on Commerce, Science and Transportation of the United States Senate, 96th Cong., 1st Sess. 395 (“Senate Hearings”) (statement of Edwin Wheeler).
These criticisms were repeated throughout the Senate hearings,21 with one witness proposing that section 203 be redrafted to specifically require that “[the] rail cost adjustment factor ... be adjusted to reflect changes in the productivity of labor, and changes in the volume and mix of traffic.” See id. at 399 (statement of Edwin Wheeler). Finally, on the last day of the hearings, Senator Cannon discussed changes in the language of section 203 with Edmund B. Frost, vice-president and general counsel of CM A:
Senator Cannon: With regard to section [203], you note that the formula set forth in S.1946 doesn’t specifically require consideration of such factors as changes in traffic volume and labor productivity, although it does call for commission compilation or verification of any index.
Several witnesses have suggested that in compiling the index, specific consideration be given to changes in volume, productivity, and traffic mix.
I assume language along those lines, if adopted, would satisfy your concerns on that point.
Mr. Frost. I think it might very well satisfy my concerns.
As redrafted by the committee, however, section 203 did not “specifically require consideration of such factors as changes in traffic volume and labor productivity.” Instead, it provided the Commission with a general mandate to make “appropriate adjustments to reflect the quality and mix of material and labor.” This language was modified once more in conference to require “appropriate adjustments to reflect the composition of costs, including the quality and mix of material and labor.” 22 Neither change in the language of section 203 was explained in the accompanying committee reports.23
Courts must exercise caution before drawing inferences regarding legislative intent from changes made in committee without explanation. See Trailmobile Co. v. Whirls, 331 U.S. 40, 61, 67 S.Ct. 982, 992, 91 L.Ed. 1328 (1947). Although a succession of draft bills may point toward a clear legislative purpose, see J. Hurst, Dealing With Statutes 42-43 (1982), amendments to a bill’s language are frequently latent with ambiguity: they may either evidence a substantive change in legislative design or simply a better means for expressing a provision in the original bill. See C. Sands, Sutherland Statutory Construction § 48.18 (1973).
[337]*337In this case, the legislative record supports an inference that the committee redrafted section 203 in response to the shippers’ testimony, but it collapses as a foundation for understanding how the committee chose to resolve the conceptual and methodological problems of cost indexing. It is possible that the committee sought to accommodate the shippers’ concerns, viewing language on the quality and mix of labor and materials as a general formulation of the kind of adjustments needed to make an input index accurately reflect costs per unit of output. Alternatively, the committee may have sought to reaffirm the approach of the original bill, requiring only those refining adjustments that fit easily within an input methodology. Neither the committee’s rejection of the bill’s original language, nor its rejection of the shippers’ proposed amendment provides a reliable basis for choosing between these two readings. We must therefore search for such guidance in an examination of the purposes and placement of section 203 in the overall statutory scheme.
b. Underlying Purpose of Section 203
The shippers ground their argument that section 203 mandates an output index on the purpose of the inflationary cost index. They contend that the underlying purpose of the inflationary cost index is to enable railroads to recover costs; no more and no less. Thus, they conclude, section 203 requires an output index, or at least an input index modified to approximate an output index. Otherwise, the index will authorize market dominant railroads to overrecover costs during periods of productivity growth at the expense of captive shippers.
We agree with the shippers that the principal function of the section 203 inflationary cost index is to provide a mechanism for recovering inflation-based costs, not to enhance revenues. But the legislative history of section 203 also demonstrates that Congress sought a cost-recovery mechanism that would avoid regulatory lag by operating with dispatch to assure that rates actually reflect current costs. Insofar as any particular methodology would interfere with achievement of this practical objective, its adoption cannot be characterized as mandated by the Act.
In introducing S.1946, Senator Cannon explained that the bill’s overall approach was “to provide the railroads with more pricing flexibility while protecting captive shippers” and to “streamlin[e] the administrative process” so as “to simplify rather than further complicate railroad regulation.” 125 Gong.Rec.S 15309 (daily ed. Oct. 29, 1979). To accomplish these purposes, S.1946 set up a simple test for determining whether a rate is subject to Commission review, and specified percentages by which railroads could increase rates without facing any threat of rate suspension. The bill, like the Act itself, set forth a two-part test for judging rate increases.24 Rate increases to recover costs would be free from attack while additional rate increases up to a specified percentage could not be suspended and could only be challenged by shippers who would bear the burden of proving unreasonableness. Discussing the first threshold.— the percentage increase authorized by the inflationary cost index — Senator Cannon explained:
The railroad’s financial difficulties may be attributed in part to the timelag involved in recovering cost increases. This section will assure that rail rates reflect a carrier’s current costs.
125 Cong.Rec.S 15315 (daily ed. Oct. 29, 1979) (emphasis added). In contrast, he stated that the bill’s “rate flexibility zone above cost increases” within which shippers bear the burden of proving unreasonableness would “assist the railroads in achieving adequate revenue levels and [would] provide additional flexibility for rate changes without regulatory interference.” Id.
Throughout hearings on S.1946, witnesses adhered to the view that the function of the inflationary cost index was to allow railroads to receive rate increases warranted by cost increases without the expense and de[338]*338lay of rate review proceedings. See, e.g., Senate Hearings 122—23 (Statement of A. Daniel O’Neal); 444 (Statement of Richard E. Miller); id. at 541 (Statement of Bruce A. Melaas). Indeed, the bill’s limitation of its “free zone” to cost-based increases stood in stark contrast to the administration’s proposed bill — which would have allowed for annual seven percentage point rate increases above cost increases. See Senate Hearings 136-37 (Statement of William H. Dempsey).
Although witnesses at the Senate hearings proposed changes in the bill’s language to assure that the index would accurately reflect incurred costs or to explicitly allow free rate increases above costs,25 no one suggested that the function of the inflationary cost index was anything other than to provide a rapid mechanism for translating cost increases into rate increases. This view of the underlying purpose of the inflation index was reaffirmed when S.1946 was reported out of committee. See S.Rep.No. 470, 96th Cong., 1st Sess. 19-20 (1979); 126 Cong.Rec.S 3278, S 3280 (daily ed. Apr. 1, 1980) (Statement of Senator Cannon).26
This consistent view of the function of the inflationary cost index, however, falls short of proving that section 203 envisions an output index. As the Commission noted, the theoretical validity of increasing rates in accordance with costs per unit of output is a different matter from the soundness of insisting that whatever “output” methodology is currently available be immediately incorporated into the cost index regardless of its reliability. If that methodology is unreliable, or involves undue delays, the resulting index could hamper rather than help the statutory goal of timely rate increases to cover current rail costs. Congress may well have had in mind the difficulties of translating statistical theory into practice when it required the Commission to make only “appropriate” adjustments to its cost index. Thus, while we agree with petitioners that the purpose of section 203 was to allow railroads to keep up with cost increases, we cannot conclude that this purpose mandates adoption of an output index regardless of the accuracy of such an index or the feasibility of designing it so as to allow railroads to obtain rate increases in a timely manner.
B. Application of Section 208
Petitioners argue that even if section 203 does not mandate adoption of an output index, the Commission acted arbitrarily and capriciously by not adjusting the AAR index to reflect productivity growth.27 They assert that the Commission, relying on inappropriate considerations, engaged in an unexplained departure from its settled practice of considering productivity data in ratemaking proceedings. We disagree. In developing its cost index the Commission faced a new task: how to estimate, on an [339]*339average and current basis, the impact of inflation on railroad costs. In formulating an answer to that question, the Commission squarely addressed the propriety of a productivity adjustment. Although we find some weaknesses in the Commission’s reasons for rejecting a productivity adjustment, we conclude that the Commission did provide a sufficiently adequate explanation for not including such an adjustment at this time. We would emphasize, however, that the Commission’s reasoning is barely adequate, and that we firmly anticipate that its experience with the modified AAR index will shed further light on the soundness of its chosen course, and may well affect its discretion to continue to omit any productivity adjustment in the future.
1. Scope of Review
When an agency rule is challenged as arbitrary and capricious, the scope of judicial review is narrow. Courts may not substitute their judgment for that of the agency, see Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971); Ethyl Corp. v. EPA, 541 F.2d 1, 34 (D.C.Cir.) (en banc), cert. denied, 426 U.S. 941, 96 S.Ct. 2663, 49 L.Ed.2d 394 (1976). Nor may they require agencies to follow procedures not required by statute. See Vermont Yankee Nuclear Power Corp. v. NRDC, 435 U.S. 519, 548, 98 S.Ct. 1197, 1214, 55 L.Ed.2d 460 (1978). But in reviewing rules promulgated under the notice and comment rulemaking requirements of the Administrative Procedure Act, 5 U.S.C. § 553 (“APA”), courts must assure that the agency has provided a reasoned explanation for its rule. In particular, a reasoned explanation for agency action must be based on a consideration of relevant factors, see Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. at 416, 91 S.Ct. at 823, providing reasons that do not contravene “ascertainable legislative intent.” See Environmental Defense Fund v. Costle, 657 F.2d 275, 283 (D.C.Cir.1981). Furthermore, an agency decision may not be reasoned if the agency ignores vital comments regarding relevant factors, rather than providing an adequate rebuttal. See Alabama Power Co. v. Costle, 636 F.2d 323, 384-85 (D.C.Cir.1979); PPG Industries, Inc. v. Costle, 630 F.2d 462, 467 (6th Cir. 1980). Within these bounds of reasoned decision-making, agencies possess a wide measure of discretion, allowing them the flexibility to adapt to changing circumstances and to reevaluate past policies. See American Public Gas Association v. FPC, 567 F.2d 1016, 1059 (D.C.Cir.1977).
In this case, the narrowness of our scope of review is reinforced by the narrowness of the Commission’s action. Although the Commission adopted a methodology for indexing rail costs that does not include a specific productivity adjustment, it stated its intention to periodically check the results of its current methodology against other data sources, including its own cost data. See 364 I.C.C. at 843; J.A. 929. Furthermore, the Commission expressed its willingness to consider altering its methodology to include a specific productivity adjustment at some later date. See id. at 847; J.A. 931. Therefore, the issue before us is simply whether the Commission has adequately supported its decision to omit such an adjustment at this time.
2. Adequacy of the Commission’s Decision
Despite the narrowness of the issue and the scope of review in this case, we find the Commission’s decision is only minimally adequate. The Commission has not responded to some important comments and has relied, in part, on inappropriate considerations. Nonetheless, it has sketched out a sufficient justification to omit a productivity adjustment from its initial cost indexing methodology.
The crux of the Commission’s rationale is that available productivity measures are sufficiently unreliable, and the consequences of overestimating productivity growth are sufficiently grave, that it is inappropriate to adjust the cost index on the assumption of continued productivity growth. The Commission’s assessment of the unreliability of productivity measures [340]*340finds support in the record. Although one witness baldly stated that a productivity adjustment is simply “an exercise in applied econometrics,” see Verified Statement of Dr. George H. Borts, J.A. 566, the record before the Commission indicated that estimates of productivity growth are hardly a simple exercise. Some estimates are not proper measures of the impact of productivity growth on overall costs because they look at a single input. For example, the Bureau of Labor Statistics’ output per manhour index fails to account for inputs other than labor. See Reply Verified Statement of Dr. Harvey A. Levine, J.A. 511. Efforts to estimate total factor productivity (output per unit of input), however, post significant methodological problems. The critical difficulty facing these studies is establishing a proper measure of railroad output. Witnesses explained that the conventional measure of railroad output — the “ton-mile” — produces distorted estimates of productivity trends. Under the ton-mile measure, one ton carried over ten miles and ten tons carried over one mile are treated as the same amount of railroad output. When the mix of rail traffic shifts to heavier items, like coal, “ton-miles” may increase independent of changes in railroad productivity. See id.; J.A. 508, 513. Such a shift may represent no net increase in railroad revenues; indeed, if railroads charge low rates for heavier commodities, increased “ton-miles” may even represent a decline in revenues. See id.; J.A. 508. In these circumstances, a downward adjustment in the cost index would be inappropriate.
The difficulties of accurately estimating productivity growth are also reflected in the diverse results of past studies. One study, by Dr. John W. Kendrick, found productivity growth to be as high as 5.2 percent per year. See id. (citing J. W. Kendrick, Post War Productivity Trends in the United States, 1948-1969, 78-79 (1973)), J.A. 512. Another study estimated that average productivity growth was low as 0.8 to 1.8 percent per year. See id. (citing Council on Economic Advisers & National Commission on Productivity, Task Force on Railroad Productivity, Improving Railroad Productivity 78 (1973)), J.A. 512. But even if the Commission could reconcile the expertise in estimating past productivity growth, it is by no means clear that it could develop a satisfactory mechanism for evaluating current productivity trends. This “exercise in applied econometrics” may yield a formula too low in predictive value to be reliable. Furthermore, it may be considerably more difficult to gather the relevant productivity data on a current basis than it is to price items in a market basket. Reliable current estimates, however, are of central importance to section 203’s function of allowing railroads to recover costs promptly. Were the Commission to simply follow the shippers’ suggestions and assume that future productivity trends will match past experience, it could over-deflate the cost index, thereby denying railroads the timely and unencumbered rate increases they deserve.28
[341]*341These methodological problems take on greater importance in the light of the Commission’s short statutory deadline for developing a cost index 29 and the generally inadequate state of rail revenues. In times of inadequate rail revenues, erroneous assumptions about efficiency gains pose a more serious threat to the solvency of railroads. As the Commission noted in its Notice of Proposed Rulemaking, crude measures of productivity growth, such as the Bureau of Labor Statistics’ output per manhour series, “tend to be distorted by short run cyclical factors.” 45 Fed.Reg. 81218; J.A. 601. Similarly, a simple assumption that productivity growth will follow average trends, ignores the impact of the business cycle on a railroad’s ability to reduce its costs through efficiency measures. See Verified Statement of Dr. Paul O. Roberts & Dr. Brian C. Kullman, J.A. 69. Rather than adopt one of these measures and risk the possibility of overdeflating the cost index at times when railroads are also suffering adverse effects of the business cycle, the Commission opted for an index that provides a greater assurance of cost recovery. So long as the Commission follows through on its stated intention to review the index against current cost data, taking appropriate action in the event of overrecovery, we cannot say that its decision to err on the side of assuring cost recovery is arbitrary.
3. The Commission’s Remaining Justification
More troubling are the Commission’s remaining policy justifications for omitting a productivity adjustment. These policy arguments suggest that the Commission may be taking the position that a productivity adjustment would never be appropriate, or at least would never be appropriate during periods of revenue inadequacy.30 If it were so, our doubts as to the Commission’s rationale would be multiplied. We reiterate that in the event that the Commission’s experience with the modified AAR index indicates that it provides significantly distorted estimates of railroad costs, the Commission would have to provide a more adequate explanation of its rationale for ignoring the impact of productivity on railroad costs.
a. Penalties for Inefficiency
The Commission’s decision suggests that the effect of a productivity adjustment on railroads with below average productivity performance justifies omitting productivity as a factor in its cost index. The Commission explained that a productivity adjustment would reduce the ability of railroads to recoup their costs even if their below average performance is beyond their control. As a hypothetical example, the Commission posed the case of a productivity breakthrough for bulk commodities. It noted that “such a breakthrough would improve industry average statistics, but would be of little benefit to railroads whose traffic was predominantly nonbulk in nature.” 364 I.C.C. at 847; J.A. 931.
Petitioners argue that the fallout of any statistical average on individual carriers is [342]*342not a proper concern in developing an index. While we see no need to decide whether the Commission could ever bias its index in favor of those railroads that experience above average costs, we think it noteworthy that the Commission has elsewhere rejected such a biased methodology. In particular, the Commission rejected a proposal by the railroads that would have allowed railroads to increase their rates by either the national index or a regional index. Its ground for rejecting that proposal was precisely the ground urged by petitioners in this case: that such an approach “would permit the actual national average level of rates to increase more rapidly than the national average of costs.” 364 I.C.C. at 852; J.A. 934. Should the Commission adhere to its “penalty” analysis, we would expect it to explain why productivity should not be averaged, while disparate regional prices are averaged into the same index.
b. Revenue Adequacy
A familiar principle of rate regulation is that regulated rates should approximate rates under competitive conditions. See A. Kahn, 1 The Economics of Regulation 65 (1970). Relying on this principle, shippers argued that under competitive conditions, productivity gains would be passed along to consumers. See, e.g., Verified Statement of Leroy E. Peabody, J.A. 820-21. The Commission rejected this justification for a productivity adjustment, reasoning that productivity gains are not passed along to consumers under competitive conditions when a firm’s revenues fail to cover its costs. See 364 I.C.C. at 847; J.A. 931. Since the Commission had previously found in revenue adequacy proceedings that railroads were receiving inadequate revenues, it concluded that productivity savings should not be passed along to shippers. See id.
We agree with petitioners that revenue adequacy is not a proper justification for omitting a productivity adjustment. The Staggers Act provides a twofold process for assuring revenue adequacy: an inflationary cost index and a rate flexibility zone for rate increases above costs. Under this two part structure, the purpose of the inflationary cost index is simply to allow railroads to raise their rates to reflect current costs. See pp. 16-20 supra. While the statute does not prohibit adoption of an index that incidentally boosts a railroad’s net revenues, increasing net revenues is not a proper reason for refusing to make otherwise appropriate adjustments. Nonetheless, as we have observed, the inadequate state of rail revenues does provide the Commission with grounds for caution in the light of the unreliability of productivity measurements. Thus, while we disagree with the Commission that inadequate revenues provide an independent reason for not adjusting the index, We do concede that the Commission properly concluded that an imprecise adjustment would pose greater risks in times of revenue inadequacy.
c. Incentives for Productivity Growth
We also fail to find a reasoned basis for the Commission’s assumption that an average productivity adjustment would undermine incentives for productivity growth. Although the Commission retreated from its original statement that a productivity adjustment would eliminate such incentives, it has continued to assert that a productivity adjustment would have an adverse effect on incentives for productivity growth.31 In adhering to this position, the Commission has failed to provide any supporting analysis, or answer cogent arguments made by the shippers’ witnesses. These witnesses [343]*343argued that an average adjustment for productivity growth should not affect the incentives facing any individual railroad.32 They suggested that if automatic rate increases were adjusted for average productivity growth, individual railroads would have an incentive to recoup gains from above average performance or to avoid the “penalty” effects of below average performance. While we recognize that the science of quantifying economic incentives is far from exact, we would expect the Commission to provide some indication of its answer to these significant comments. Otherwise, the Commission’s bald assertion that a productivity adjustment would adversely affect incentives for productivity growth is an inadequate basis for ignoring a potentially significant element of railroad costs.
III. CONCLUSION
In affirming the Commission’s order, we share some of the doubts expressed by ICC Chairman Taylor that the modified AAR index may not be precisely the type of index envisioned by the statute. See Ex Parte No. 290 (Sub. No. 2), Railroad Cost Recovery Procedures, Clarification of Decision, 365 I.C.C. 302, 303 (1981) (Chairman Taylor, concurring); J.A. 947. Our role on review, however, is limited to determining whether the Commission has reasonably interpreted its statutory mandate and adequately justified its choice of index. We conclude that within the present state of the art — at least as reflected in this record — the difficulties of accurately measuring productivity growth on a current basis, coupled with the Commission’s assessment of the importance of assuring railroads unencumbered rate increases to cover increased costs, support the Commission’s conclusion that a productivity adjustment is inappropriate under present circumstances. We expect, however, that in accordance with its stated intentions, the Commission will continuously review the accuracy of its chosen index and will revise the index in an appropriate fashion, including consideration of measurable productivity gains, as the circumstances warrant.