Opinion for the Court filed by Circuit Judge WILLIAMS.
WILLIAMS, Circuit Judge:The National Gas Policy Act of 1978 (the “NGPA” or the “Act”), 15 U.S.C. §§ 3301 et seq. (1982), sets ceiling prices for certain sales of natural gas. Under § 110, producers may raise their prices above these ceilings “to the extent necessary to recover ... [s]tate severance taxes.” 15 U.S.C. § 3320(a) (1982). Such taxes are in turn defined as “any severance, production, or similar tax, fee, or other levy imposed on the production of natural gas.” Id. § 3320(c). The Federal Energy Regulatory Commission has classified Kansas’s ad valorem property tax as a severance tax under § 110. Colorado Interstate Gas Company and Northern Natural Gas Company,1 interstate pipelines that buy price-regulated natural gas, object to this classification.
The Commission has stated by way of explanation that the tax is “based on production factors.” Northern Natural Gas Co., 38 FERC ¶ 61,062 (1987), at 61,176. This is undoubtedly so: past production is used to estimate the net present value of future production from a gas-producing property, and thus the property’s taxable value. But that link is not necessarily enough. The capital value of any property is in essence the net present discounted value of its anticipated income stream; this is true even for a shirt, which generates a stream of (non-pecuniary) income in the form of satisfaction as it is worn. We conclude that the Commission’s treatment of the issue fell short of reasoned decision-making. It failed to offer any principle for determining what relation to production is enough for a tax to qualify under § 110. That failure is underscored by its inability to supply a persuasive explanation for simultaneously denying severance tax treatment for Texas’s seemingly indistinguishable tax.
I.
The Commission's predecessor, the Federal Power Commission (here also referred to as the Commission), first had occasion to give separate treatment to state severance taxes when, in setting producer rates under the Natural Gas Act of 1938, 15 U.S.C. §§ 717 et seq. (1982), it abandoned the evidently impossible task of setting them individually for each producer and started to set them for a single producing area on the basis of average production costs. In Area Rate Proceeding (Permian Basin), 34 FPC 159 (1965), aff'd sub nom. Permian Basin Area Rate Cases, 390 U.S. 747, 88 S.Ct. 1344, 20 L.Ed.2d 312 (1968), the first regional rate order, it allowed producers to add on to the ceilings “the full amount of taxes actually incurred either in Texas or New Mexico.” Id. at 206. Although the Commission did not explain the separate treatment, one may surmise that it regarded it as appropriate because the state boundary formed a bright line between the two different cost levels. Thus there was not the same practical need to use average figures for severance taxes as there was for other costs. The Commission again followed this approach when it adopted national ceilings in Just and Reasonable National Rates for Sales of Natural Gas, Opinion No. 699, 51 FPC 2212, reh. denied *78in relevant part, 52 FPC 1604 (1974), aff'd sub nom. Shell Oil Co. v. FPC, 520 F.2d 1061 (5th Cir.1975), cert. denied sub nom. California Co. v. FPC, 426 U.S. 941, 96 S.Ct. 2660, 49 L.Ed.2d 394 (1976). There it provided generically for adjustment of the national rates upward for “all ... production, severance, or similar taxes.” Just and Reasonable National Rates, supra, at 2301-02. The words are, of course, exactly those used in part of § 110’s definition of allowable severance taxes, though § 110 slightly alters the order.
In Opinion No. 699 the Commission did not explicitly state why producers should be able to recover severance or production taxes but not, say, property taxes. The reason that immediately comes to mind, of course, is that non-recovery of a cost that varies with production will work as a disincentive to production, while non-recovery of invariant costs will not. A production tax is a cost of producing, a property tax a cost of holding an asset. Indeed, to the extent that extraction reduces the base against which a property tax is applied, its imposition may tend to accelerate production. Compare Anthony Scott, Natural Resources: The Economics of Conservation 195-96 (Carleton Library ed. 1973) (effects of property tax) with id. at 185 (effects of production tax). A special feature of natural gas (or any “fugacious” resource) blurs the distinction: in a reservoir with multiple producers, the extraction by a taxpayer’s fellow owners may have far more effect on the annual decline in a property’s recoverable reserves than his own.
In October 1974, at the behest of the Kansas State Corporation Commission, the Commission declared that Kansas’s ad valorem tax on “all oil and gas wells, producing or capable of producing oil or gas in paying quantities,” Kan.Stat.Ann. § 79-329, was eligible for Opinion No. 699’s adjustmnet. Opinion No. 699-D, 52 FPC 915 (1974).
Without claiming to grasp the refinements of the Kansas tax, we can briefly describe its calculation. The market value of annual production is calculated by multiplying past production (averaged over a three- or five-year period where available) by the price actually paid the producer on January 1 of the assessment year. Kansas then capitalizes annual production (i.e., infers a present capital value from income expected to be received over the future) by multiplying it by a “present worth factor” or “PWF.”
As would any capitalization formula, the PWF incorporates an assumed discount rate and expectations as to future price changes. Because natural gas is an exhaustible resource, the PWF also builds in assumptions (or information) about rates of decline in production and the probable length of production; these naturally vary from property to property. The Kansas Department of Revenue has calculated PWFs for many specific Kansas gas fields, ranging from 6.30 for the Hugoton (above 3000 feet) to 2.50 for the Glick. See Kansas Oil & Gas Appraisal Guide (1986), Joint Appendix (J.A.) 83, 98. For “all other Kansas gas fields,” Table B sets forth PWFs for rates of production decline experienced and for either a five- or eight-year estimated future life. See Appraisal Guide, J.A. at 101. The PWFs in Table B range from 3.915 (for a zero decline rate and an eight-year life) to .673 (for a 50% decline rate and a five-year life). After annual production is multiplied by the PWF, the resulting figure is adjusted downwards for operating costs and upwards for the assumed value of equipment. Although production plays an important role in tax computation, the variation caused by other elements — included because of the state’s interest of ascertaining capital value — may make the ultimate tax on equal production vary nearly by a factor of ten.
The Commission found the Kansas tax “similar” to a production or severance tax within the meaning of its offset rule because “the bulk of the Kansas ad valorem tax is based upon production factors, and, as such, is in fact, a severance or production tax merely bearing the title ‘ad valorem tax.’ ” Just and Reasonable National Rates for Sales of Natural Gas, Opinion No. 699-D, 52 FPC 915 (1974).
*79Two years later the Commission found that Texas’s ad valorem tax on real and tangible personal property was not eligible for add-on treatment under Opinion No. 699. Mobil Oil Corp., 55 FPC 917, 918 (1976). It said that “[although annual and future gas production is quite naturally used to determine the value of gas properties for tax purposes, the tax itself does not depend on production. It depends on the value of the remaining recoverable reserves.” Id. We will return later to the meaning and validity of the Commission’s efforts to distinguish the two taxes.
In 1978 Congress substantially overhauled federal regulation of natural gas prices by adopting the NGPA. In § 110 it provided for severance taxes, using the words of Opinion No. 699 but adding the phrase “imposed on the production of natural gas.” The only directly relevant legislative history is a line in the Conference Report to the effect that “The term ‘State severance tax’ is intended to be construed broadly.” H.R.Conf.Rep. No. 1752, 95th Cong., 2d Sess. 91, reprinted in 1978 U.S. Code Cong. & Admin.News pp. 8800, 9007. For several years after its adoption producers and purchasers evidently assumed that the Commission’s earlier classification of the two states' taxes would carry over to the new regime.
In 1983, however, Texas gas producers asked that they be allowed under § 110 to add that state’s ad valorem tax to their rates. The present petitioners in turn asked FERC to overrule Opinion No. 699-D’s grant of add-on status for the Kansas tax. FERC consolidated the petitions and denied both. Sun Exploration and Production Co., 36 FERC ¶ 61,093 (1986). The Texas producers invoked the single fragment of legislative history, but the Commission responded by saying that to fall within § 110 an “ad valorem tax still must be one based ‘upon ... production.’ ” 36 FERC ¶ 61,093 at 61,224. In answering the parties’ claim that its treatment of the two taxes was inconsistent, the Commission said they were to be distinguished in terms of “the degree to which production factors are taken into account.” Id. at 61,224. We will shortly consider the Commission’s effort to show that its results fitted that idea. Petitioners sought rehearing of the Kansas aspect of the decision, without success. Northern Natural Gas Co., 38 FERC ¶ 61,062 (1987).
Before reaching the merits, we must address a claim, raised by Mobil Oil and other producer-intervenors in support of FERC, that the present challenge is in essence an invalid collateral attack on FERC’s 1974 disposition of the Kansas tax in Opinion No. 699-D, which they say became “final in 1976, upon the denial of certiorari in the appeal of the Opinion No. 699 series of orders.” Mobil Brief at 3 (emphasis in original). They characterize FERC’s more recent orders not as substantive dispositions, but as mere denials of petitioners’ requests to “reopen a long-closed record.” Id. at 4.
As the NGPA provided a statutory standard for the first time and thus seemingly introduced a new legal question (the one fundamentally before us today), this would seem an unpromising defense under any circumstances. Here, FERC’s own decision to formally reconsider the Kansas and Texas taxes makes it easy to reject the theory. As this court held recently in Association of American Railroads v. ICC, 846 F.2d 1465, 1473 (D.C.Cir.1988), where the ICC asserted that the petitioner’s challenge to a long-standing agency practice was time-barred under the Hobbs Act, once “the agency has opened the issue up anew ..., its renewed adherence is substantively reviewable.” See also State of Ohio v. EPA, 838 F.2d 1325, 1328 (D.C.Cir.1988).
II.
In evaluating the Commission’s handling of the Kansas tax, we start with the recognition that a tax nominally on property may be functionally identical to a production tax. If a state sought to capitalize the annual production (or revenue) enjoyed by each producer by multiplying it by a single fixed figure, the tax would plainly be similar enough to a production tax to qualify under § 110. Neither Kansas nor Texas does that.
We have already noted the way in which different expectations of aggregate future *80recovery cause the PWFs to vary radically. Other rules also drive the Kansas tax away from a pure production tax. Kansas gives its assessors general instructions to use actual production figures only to the extent that these figures are helpful in ascertaining a property’s value. Appraisal Guide i, 1, 13, J.A. 84, 87, 99. It enjoins its assessors to adjust from the Guide itself where “necessary ... to comply with the constitutional law of equality and the statutory requirement of assessing said property at thirty per cent of its fair market value in money.” Appraisal Guide i, J.A. 84.
Moreover, in 1980 Kansas amended its tax to move it (we think) still further in the direction of a property tax. Recognizing that new wells typically produce at disproportionately high rates, it provided that when a well starts production after July 1, the assessor should discount its annualized rate of production by 40% before proceeding to the remaining calculations. Kan. StatAnn. § 79-331(b). Notice it is in fact annualized production that is used; the tax on a property starting to produce December 1 will be far more than one-twelfth that of an equally productive property starting to produce January 1.
On the other hand, we note some rules giving actual current production a greater role than would seem normal for a standard property tax. A well that is capable of production in a major proven field, but is shut-in before production starts, is initially appraised at a simple $20 per foot of depth. See Appraisal Guide at 13, J.A. 99. At least for wells with good prospects, this must understate true value. Once shut-in for two years, the well is valued at the salvage value of the equipment, see Appraisal Guide at 13, 15, J.A. at 99, 101, again a likely understatement. On this issue, Kansas has manifested a focus on production (rather than on recoverable reserves or productive potential) that is likely to generate valuations below market. Accordingly, the tax’s effect on an owner’s decision whether to shut in a well may be very similar to that of a conventional severance tax.
As we have already noted, the Commission in Opinion No. 699-D merely observed that “the bulk of the Kansas ad valorem tax is based upon production factors.” 52 FPC 915 (1974). But it has not systemically analyzed the tax or assessed its operation by reference to any cogent theory of what makes a tax “similar” to a production or severance tax under § 110. It has not identified those features that it believes constitute the production-factor “bulk,” much less the features excluded from the bulk. Its treatment of the 1980 amendment is puzzling. While this seems clearly to turn the tax further away from pure production, the Commission in 1986 asserted in conclusory terms that “the amendment supported] the ... conclusion that the Kansas ad valorem tax is based in substantial part on production.” 36 FERC ¶ 61,093, at 61,224.
Equally striking is the apparent irrelevance of the Commission’s statements in explanation of its decision to treat Kansas and Texas differently. While saying generally that the difference lay in “the degree to which production factors are taken into account,” 36 FERC 11 61,093, at 61,224, it did not identify pertinent distinctions. It made three points. First:
The Texas tax is applied to the assessed cash market value of the remaining recoverable reserves (i.e., real property) in place and unsevered, at the time of appraisal each year. The tax assessment is based on factors used to project the prospective future income (discounted to present value) that a prospective buyer could reasonably expect to derive from the production and sale of the remaining recoverable reserves in place at the time of appraisal____
36 FERC H 61,093, at 61,224 (emphasis in original). The same can be said of the Kansas tax.
Second, the Commission said that in Texas “[ajctual current and prior year production data is used only to arrive at a declining rate of production to project future rates of production from the remaining recoverable reserves in place at the time of the appraisal.” Id. In attempting to differentiate the Kansas tax, it remarked that “while Kansas ... utilize[s] a present worth factor to evaluate gas properties (be*81cause an ad valorem tax is a tax on property) actual current or prior-years’ production volumes or the revenues derived therefrom are a significant factor in the formula used to determine the tax.” Id. at 61,224 (emphasis in original.) The two descriptions seem to vary only in that for Kansas the Commission has italicized “actual” and has tagged the production factors as “significant.”
Third, FERC observed that “Texas has had both a severance tax and an ad valorem tax for a long time, an indication that Texas looked upon the two taxes as separate and distinct taxes.” Id. This mystifies us. The issue before the Commission was one of federal law — classifying the taxes by reference to some vision of Congress’s purpose in § 110. While conceivably Congress might have intended state labels to control, it seems implausible and no one so argues here. If the test is a functional one, it would seem that the taxing state’s outlook on the tax would be relevant (if at all) only as a tie-breaker: in a case otherwise in perfect balance, the state’s viewpoint might suggest how assessors would exercise their remaining discretion, and thus determine the outcome. But this presupposes a careful analysis of how the specific rules of the tax actually function. Further, as the Commission’s modifying phrase (“for a long time”) hints, the distinction is purely historical. By the time of FERC’s decision, Kansas too had a severance tax. Kan.Stat.Ann. § 79-4217. We find it hard to imagine any theory under which the date of a state’s coming to believe its ad valorem tax and its severance tax were distinct could rationally guide the classification under § 110.
Before us FERC argues that we should disregard the claims that it distinguished irrationally between the two states. Its basis for the argument is that the decision on the Texas tax was not appealed and so is not before us. This is doubtless true. But the Commission’s dissimilar treatment of evidently identical cases, on the same day, seems the quintessence of arbitrariness and caprice. Courts routinely insist that an agency changing direction over time offer a rational explanation of the shift, Motor Vehicle Mfrs. Ass’n v. State Farm Mutual Automobile Ins. Co., 463 U.S. 29, 41-42, 103 S.Ct. 2856, 2865-66, 77 L.Ed.2d 443 (1983); Greater Boston Television Corp. v. FCC, 444 F.2d 841, 852 (D.C.Cir.), cert. denied, 403 U.S. 923, 91 S.Ct. 2229, 29 L.Ed.2d 701 (1971); we ask less than that in requiring some plausible distinction for simultaneous, apparently contradictory acts.
Of course under Chevron U.S.A. Inc. v. NRDC, 467 U.S. 837, 842-44, 104 S.Ct. 2778, 2781-83, 81 L.Ed.2d 694 (1984), any Commission interpretation of § 110 that is not precluded by the statutory language and traditional methods of statutory construction, and that is reasonable, will control. See also NLRB v. United Food and Commercial Workers Union, Local 23, AFL-CIO, — U.S. -, 108 S.Ct. 413, 421, 98 L.Ed.2d 429 (1987). But we cannot defer to a vacuum. We remand to the Commission for the analysis that so far has been wanting.
In closing we take note of the rather special context of this case. It came to us some 34 years after federal regulation of independent producer prices was initiated by virtue of the decision in Phillips Petroleum Co. v. Wisconsin, 347 U.S. 672, 74 S.Ct. 794, 98 L.Ed. 1035 (1954), and, as it has proved, at a point when statutory change and market forces have brought the system virtually to an end. Many categories of gas are formally deregulated; many others are subject to price ceilings far above what producers can obtain in the market. Regulation holds sway primarily in two quite disparate areas. Many producer-pipeline contracts provide that if the gas is subject to a legal ceiling, that ceiling shall be the price, but that if the gas is deregulated, the market price should control or the parties should renegotiate. See FERC v. Martin Exploration Mgt. Co., — U.S. -, -, 108 S.Ct. 1765, 1768-69, 100 L.Ed.2d 238 (1988). Where the ceiling is above market the effect of continued regulation is thus to enable the producer to secure more than the market price. FERC’s decision that gas eligible for both a regulated and a deregulated category should be treated as deregulated, recently upheld by the Supreme Court in Martin, will partly extinguish this residual effect of *82regulation. At the opposite end of the range of price controls, ceilings are below market, but FERC, by Order No. 451, 51 Fed.Reg. 22,168 (June 6, 1986), appeal docketed sub nom., Mobil Oil Exploration & Producing Southeast, Inc. v. FERC, Nos. 86-4940, et al. (5th Cir.), has sought to create room for price renegotiation. Thus any changes by the Commission may be of limited relevance, and FERC’s task on remand may be about as inviting as having to make costly repairs on a building slated for demolition.
Nonetheless, we think the law requires the Commission to exercise its interpretive authority, to identify the features of the Kansas tax that point toward one classification or another, and to offer sensible distinctions between taxes that it chooses to treat differently.
We note that the Commission denied petitioners' request for a hearing. Whether one is necessary depends on the Commission’s analysis of § 110 and on the alternative means.'by which the Commission may secure the information that proves relevant. The choice of process lies in FERC’s sound discretion.
We grant the petition and remand the case.
. In this court Colorado Interstate is petitioner and Northern Natural intervenor; in the proceedings below they occupied the opposite roles. We refer to them both collectively as petitioners.