Edwards, Chief Judge.

Petitioner Williston Basin Interstate Pipeline Company ("Williston Basin") seeks review of multiple orders of the Federal Energy Regulatory Commission ("FERC" or "Commission") in connection with a general rate increase filed by Williston Basin under § 4 of the Natural Gas Act ("NGA" or "Act"), 15 U.S.C. § 717c. The Commission found that Williston Basin had not satisfied its burden of demonstrating that various components of its proposed rate increase were lawful, and it therefore ordered certain adjustments to Williston Basin's filing. In this petition for review, Williston Basin takes issue with the Commission's findings insofar as they concern the rate of return on common equity, ad valorem tax expense, throughput projection, depreciation allowance, and cost of long-term debt. The Public Utilities Commission of South Dakota, Montana Consumer Counsel, and Montana Public Service Commission ("State Agencies") have intervened in support of the Commission's position.

We find that Williston Basin's challenges to the Commission's depreciation and cost of long-term debt determinations are plainly without merit, and, therefore, warrant no discussion. The Commission's decisions require no amplification on these two issues. However, for the reasons provided below, we grant Williston Basin's petition for review and remand to the Commission for further proceedings on the issues related to the rate of return on common equity, ad valorem tax, and throughput.

I. Background A. Regulatory Framework

This case involves the Commission's authority, pursuant to the NGA, to regulate "the transportation of natural gas in interstate commerce." 15 U.S.C. § 717(b) (1994). Section 4(a) of the Act requires that rates charged by natural gas pipelines within the Commission's jurisdiction be just and reasonable. See id. § 717c(a). Consistent with this mandate, pipelines must file all proposed rates with the Commission for a determination as to their reasonableness. See id. § 717c(c). The pipeline bears the burden of demonstrating that a proposed rate change is reasonable. See id. § 717c(e). The Commission may suspend the operation of a proposed new rate for up to five months pending a reasonableness determination. See id. If the Commission fails to reach a determination before the end of the suspension period, it must allow the filed rate to go into effect subject to an ultimate decision, which may be made retroactive. See id.

B. Commission Rate-Setting Practices

The Commission sets pipeline rates by dividing revenue requirements by projected demand to attain a dollar-per-unit-of-service figure. To begin, the Commission sets a pipeline's basic costs by totaling operation and maintenance expenses, depreciation, and taxes, including ad valorem taxes. As it is ordinarily impossible for a pipeline to know at the time of filing what its actual costs will be during the effective period of the filed rates, the Commission has adopted a "test period" approach for this stage of rate making. Under this approach, a pipeline submits data in support of its rate proposal that reflects actual experience over the most recent twelve consecutive months (the "base period"), adjusted for changes that are known and measurable with reasonable accuracy at the time of filing, and that will become effective within nine months after the last month of actual experience (the "adjustment period"). See 18 CFR § 154.303(a)(4) (1998). (Separate test period regulations govern rate setting in the electric utility context. See id. § 35.13.) Under certain circumstances, the Commission has discretion to make adjustments in light of actual, post-test period data. See Exxon Corp. v. FERC, 114 F.3d 1252, 1263 (D.C. Cir. 1997). For the most part, however, the Commission develops rates using the representative cost data available at the time of filing. The test period underlying the rates in this case consisted of a twelve-month base period ending January 31, 1992 and a nine-month adjustment period ending October 31, 1992.

Next, the Commission adds to this basic cost of service figure a reasonable profit, computed by multiplying the rate base by the rate of return. See Boston Edison Co. v. FERC, 885 F.2d 962, 964 (1st Cir. 1989). The rate base, which is not at issue in the present case, represents "total historical investment minus total prior depreciation." Id. (internal quotation omitted). The rate of return, which is very much at issue in the present case, represents a weighted average of the costs of the three elements comprising the pipeline's capital structure: long-term debt, preferred stock, and common equity. See North Carolina Utils., Comm'n v. FERC, 42 F.3d 659, 661 (D.C. Cir. 1994). The cost of common equity is frequently, as it is here, a point of contention in rate making. Nepco Mun. Rate Comm. v. FERC, 668 F.2d 1327, 1335 (D.C. Cir. 1981).

To calculate a pipeline's rate of return on common equity, the Commission first develops a "zone of reasonableness," which gauges returns experienced in the industry, ordinarily by reference to a proxy group of publicly-traded companies for which market data is available. North Carolina, 42 F.3d at 661-62. To arrive at this zone of reasonableness, the Commission favors a discounted cash flow ("DCF") model, which projects investor growth expectations over the long term by adding average dividend yields to estimated constant growth in dividends over the indefinite future. The premise of the DCF model is that the price of a stock is equal to the stream of expected dividends, discounted to their present value. Once the Commission has defined a zone of reasonableness in this manner, it then assigns the pipeline a rate within that range to reflect specific investment risks associated with that pipeline as compared to the proxy group companies. See id. at 661. This figure, combined with the long-term debt and preferred stock figures, represents the overall rate of return used to calculate the pipeline's profit allowance.

In the final rate-making step, the Commission divides the total revenue requirement-cost of service plus reasonable profit-by the total demand. Demand corresponds with throughput volume on the pipeline system, which, like cost of service, is computed by reference to a test period. See Exxon Corp., 114 F.3d at 1263-64. This calculation yields the per-unit price necessary to cover the pipeline's revenue requirement, which, in turn, represents a reasonable price that the Commission will permit the pipeline to recover. See Boston Edison, 885 F.2d at 964.

C. Commission Proceedings

The procedural history of this case, which spanned more than five years and spawned six Commission orders and two administrative law judge ("ALJ") decisions, does not bear exhaustive recitation here. To put the relevant issues into context, we need only summarize the Commission's determinations, as relevant to the rate of return on common equity, ad valorem tax, and throughput issues.

Williston Basin is a natural gas company that operates a pipeline system within the states of Montana, North Dakota, South Dakota, and Wyoming. On April 30, 1992, Williston Basin filed tariff sheets with the Commission in order to implement a proposed general rate increase under § 4 of the NGA, to be effective on June 1, 1992. On May 29, 1992, the Commission accepted Williston Basin's filing, suspended the rates until November 1, 1992, and made the increase subject to refund, various conditions, and the outcome of a hearing on cost-of-service and throughput issues. See Williston Basin Interstate Pipeline Co., 59 F.E.R.C. ¶ 61,237 (1992). Subsequently, on September 30, 1992, Williston Basin filed a superceding rate increase to reflect firm service conversions, from sales to transportation, on its system. Because this filing relied on the same cost of service and allocations as the earlier rate case, the two proposals raised several identical issues. The Commission accepted the revised rates, suspended them until November 1, 1992, and consolidated the new filing with the pending proceeding. See Williston Basin Interstate Pipeline Co., 61 F.E.R.C. ¶ 61,129 (1992). In the meantime, Williston Basin filed revised tariff sheets in connection with its restructuring pursuant to FERC Order No. 636. By order dated February 12, 1993, the Commission set various issues in the restructuring proceeding for hearing, to be addressed in the ongoing proceedings in the 1992 dockets. See Williston Basin Interstate Pipeline Co., 62 F.E.R.C. ¶ 61,144 (1993). The Commission thereafter permitted Williston Basin to implement its restructuring as of November 1, 1993, subject to certain conditions. Consequently, the orders under review affect Williston Basin's rates from June 1, 1992 through December 31, 1995, the effective date of its next rate case.

On July 19, 1994, following an evidentiary hearing on the matters raised by the Commission, the presiding ALJ issued an initial decision in these proceedings, finding, with respect to the issues relevant here, that Williston Basin had adequately supported the ad valorem tax and throughput components of its rate proposal, but had failed to justify the return on common equity element. See Williston Basin Interstate Pipeline Co., 68 F.E.R.C. ¶ 63,007 (1994). On July 25, 1995, the Commission issued an order affirming in part and reversing in part the ALJ's decision. The Commission concluded that Williston Basin had not met its burden as to any of these issues. See Williston Basin Interstate Pipeline Co., 72 F.E.R.C. ¶ 61,074 (1995) ("July 1995 Order").

With respect to the rate of return on common equity, the Commission focused on the appropriate data to be used for the dividend growth rate in the DCF model. In its filing, Williston Basin had proposed a return on equity of 15 percent, based on five-year earnings forecasts published by the Institutional Brokers Estimate System ("IBES") for the relevant proxy group companies. Rejecting this single-stage approach to the dividend growth estimate, the Commission relied instead on the two-stage approach articulated in several recent rate cases. See Williston Basin, 72 F.E.R.C. at 61,376. The Commission looked in particular to Ozark Gas Transmission System, 68 F.E.R.C. ¶ 61,032 (1994), where it had recognized that exclusive reliance on short-term growth projections is inconsistent with the DCF model, which assumes dividend growth for an indefinite period of time. The Commission concluded, therefore, that Williston Basin's dividend growth projection must reflect estimates of both long-and short-term growth. See Williston Basin, 72 F.E.R.C. at 61,376. Because the Commission found that the proposals before it lacked sufficient evidence of long-term growth rates, it took official notice of Data Resources, Inc./McGraw Hill ("DRI") projections for retail gas consumption and prices, which had been used to represent long-term growth in Ozark, and averaged those data with the IBES five-year projections. See id. When added to the average dividend yields for the proxy companies, these data produced a zone of reasonableness for the rate of return on common equity of 10.97 to 13.43 percent, from which the Commission adopted the midpoint of 12.20 percent for Williston Basin. See id.

With respect to the ad valorem tax expense, the Commission found that test period principles precluded Williston Basin's proposal to include in its cost estimate increased amounts that it anticipated owing to Montana and South Dakota in connection with various plant additions. See id. at 61,363. The Commission found that, although the plant additions occurred during the test period, the effect of these additions "is not known and could not be measured with reasonable accuracy during the test period." Id. The Commission reasoned that, because "[t]he determination of the exact ad valorem tax effect is a local matter involving local valuation and tax assessment procedures," the projected adjustment to ad valorem tax liability was too speculative. Id. The Commission also attempted to distinguish a previous Williston Basin rate proceeding, on the ground that the adjustment permitted there was not speculative. See id.

Finally, with respect to throughput, the Commission rejected, also on test period principles, Williston Basin's proposal to reduce the projected volume to reflect two major bypasses to its pipeline system. See id. at 61,382-83. At the time of its filing, Williston Basin had expected these bypasses to occur during the nine month "adjustment" portion of the test period; however, the bypasses in fact occurred during the four months following the close of the test period. See id. Thus, although Williston Basin's estimates were reasonable when made, the Commission relied instead on the most updated actual data for the test period that was available before the rates took effect. See id. According to the Commission, where the bypasses were known not to have occurred during the test period, and where the actual time that they would occur could not have been known then, the fact that the bypasses did subsequently occur could not be considered. See id.

Williston Basin sought rehearing as to each of these issues. By order dated July 19, 1996, the Commission addressed further, but declined to rehear, the ad valorem tax and throughput issues. See Williston Basin Interstate Pipeline Co., 76 F.E.R.C. ¶ 61,066 (1996) ("July 1996 Order"). However, the Commission granted rehearing on the issue of the long-term growth factor to be used in calculating the rate of return on common equity. Commenting that the data of which it took official notice in the previous proceeding was not widely available, the Commission concluded that:

"[t]he parties need an opportunity to cross-examine the proponents of using the DRI data, or any other long term growth projection, to determine whether the projections are properly used. At an evidentiary hearing, parties will have the opportunity both to present their own testimony concerning the appropriate data to use in projecting long term growth and to ascertain the basis of any other party's reliance on the DRI or other data."

Williston Basin, 76 F.E.R.C. at 61,390 (footnote omitted). Accordingly, the Commission ordered a hearing for "the sole purpose of determining the appropriate long term growth rate to be applied" in the two-stage DCF analysis approved by the Commission in the July 1995 Order. Id. On October 8, 1996, following a hearing in which all parties presented testimony on this issue, the ALJ rendered a decision, essentially adopting the approach taken by the Commission in its previous order-i.e., use of DRI data for the retail gas commodity. See Williston Basin Interstate Pipeline Co., 77 F.E.R.C. ¶ 63,001, at 65,006 (1996).

By order dated June 11, 1997, the Commission reversed the ALJ, concluding "that a projection of long-term growth for the specific pipeline companies in the proxy group or for the pipeline industry as a whole cannot reasonably be developed based on available data sources." Williston Basin Interstate Pipeline Co., 79 F.E.R.C. ¶ 61,311, at 62,388 (1997) ("June 1997 Order"). The Commission found that Williston Basin's proposal, which advocated the sole use of IBES five-year earnings forecasts in the DCF model, was at odds with the two-stage DCF approach announced in Ozark, as well as the approach to long-term growth used by large investment brokerage houses. See id. at 62,388. Furthermore, it determined that the FERC staff's approach, which used DRI projections of growth in retail gas consumption as its basis for determining long-term growth in pipeline earnings, was also deficient, for there was no reason to assume the necessary correlation between gas commodity and gas transmission revenues. See id.

Based on the evidence presented at the hearing, the Commission abandoned the industry-specific approach to long-term growth estimates and adopted, as an alternative, "the long-term growth rate of the economy as a whole, as measured by the gross domestic product." Williston Basin, 79 F.E.R.C. at 62,387. The Commission provided four reasons for its decision to use economy-wide growth estimates: first, the record showed that, as companies reach maturity, their growth rates approach that of the economy as a whole; second, it is reasonable to predict that, in the long run, a regulated firm will grow at the rate of an average firm in the economy, because regulation will moderate profitability in good and bad economic periods; third, whereas the record did not show that investors rely on the approaches suggested by the parties in determining long-term growth, there was evidence that two large brokerage firms, Merrill Lynch and Prudential-Bache, use the long-term growth of the economy in conducting DCF analyses for investment purposes; and fourth, the FERC staff witness in this case, and witnesses in other cases, have used the long-term growth of the economy to confirm the results of their analyses conducted using industry-or firm-specific estimates of growth. See id. at 62,389-90.

After relying on investment houses to support its shift to an economy-wide approach to long-term growth, the Commission declined to adopt the particular long-term growth models used by Merrill Lynch or Prudential-Bache. The Commission acknowledged that the use of gross domestic product ("GDP") differed from the methodologies of the investment houses, but explicitly made this choice, because it found that the three-stage approaches used by these firms demanded more "involved" calculations, which depended on "the exercise of subjective judgment." Id. at 62,390. Although no party had discussed or advocated GDP data at the hearing, an exhibit to the FERC staff's testimony contained, as background, estimates of long-term GDP growth from both DRI (5.37 percent) and Energy Information Administration ("EIA") (6.33 percent). The Commission averaged these estimates to yield a long-term growth figure of 5.85 percent and an adjusted zone of reasonableness of 10.5 to 12.96 percent. See id. Accordingly, the Commission ordered Williston Basin to use the midpoint of 11.73 percent-ironically, a figure even lower than that reached in the July 1995 Order-in its compliance filing. See id.

Williston Basin once again sought rehearing, reiterating its opposition to the use of any long-term growth projections in the DCF analysis, and challenging on multiple grounds the Commission's adoption of the GDP as the long-term growth factor. See Request for Rehearing of Williston Basin Interstate Pipeline Company ("Rehearing Request"), reprinted in Joint Appendix ("J.A.") 201-27. On October 16, 1997, the Commission rejected these challenges in the final order under review, reaffirming the two-stage methodology underlying its rate of return determination and defending its choice of GDP as the long-term growth factor to be used in the DCF analysis. See Williston Basin Interstate Pipeline Co., 81 F.E.R.C. ¶ 61,033, at 61,174-77 (1997) ("October 1997 Order"). This petition for review followed.

II. Analysis A. Standard of Review

We review FERC orders under the Administrative Procedure Act's ("APA") arbitrary and capricious standard. See Union Pac. Fuels, Inc. v. FERC, 129 F.3d 157, 161 (D.C. Cir. 1997); 5 U.S.C. § 706(2)(A) (1994). Our role in this context is "limited to assuring that the Commission's decision-making is reasoned, principled, and based upon the record." Pennsylvania Office of Consumer Advocate v. FERC, 131 F.3d 182, 185 (D.C. Cir. 1997) (citations and internal quotation marks omitted). To this end, we examine the orders on review to ensure that the Commission has considered the relevant data and "articulate[d] . . . a rational connection between the facts found and the choice made." Association of Oil Pipe Lines v. FERC, 83 F.3d 1424, 1431 (D.C. Cir. 1996) (citations and internal quotation marks omitted).

B. Rate of Return on Common Equity

We begin with the most vigorously contested issue in this case: the rate of return on common equity. Although all parties to the proceeding have embraced the DCF model for calculating Williston Basin's return on equity, they dispute the appropriate methodology and data sources for determining the dividend growth rate to be used in this analysis. In the orders below, the Commission adopted a two-stage growth factor, using IBES data for the short-term growth rate and GDP data for the long-term growth rate. As we view it, Williston Basin's litany of challenges to this determination boils down to two core arguments: (1) the Commission erred in requiring the DCF analysis to include a long-term growth factor at all; and, (2) even assuming that some long-term growth factor was appropriate, the Commission improperly adopted the two GDP figures that happened to be in the record but were not discussed at the hearing.

1. Use of a Two-Stage Dividend Growth Factor in the DCF Model

One of Williston Basin's principal concerns is the Commission's decision to follow Ozark and other Commission precedent and include a two-stage growth factor in its DCF model. Williston Basin suggests that the applicability of the DCF methodology is unjustified for lack of clarity in FERC precedent. We have a very different view of the matter. The Commission squarely addressed the application of its new policy to the particular context of Williston Basin's ongoing proceeding: following a hearing devoted expressly to long-term growth issues, the Commission entertained and rejected Williston Basin's arguments on this point, explaining in full its decision to require a long-term growth estimate in conformity with the Ozark methodology. See Williston Basin, 77 F.E.R.C. at 65,005; Williston Basin, 79 F.E.R.C. at 62,388; Williston Basin, 81 F.E.R.C. at 61,173-76. In short, whatever questions Williston Basin had regarding the use of some two-stage growth factor in the DCF model were answered by FERC.

Williston Basin, for its part, was intractable in its position that the Commission should rely exclusively on the short-term IBES forecasts in projecting dividend growth. Indeed, when the Commission established a hearing for the sole purpose of determining the appropriate long-term growth rate, Williston Basin proposed no objective measure of long-term growth, arguing instead that long-term growth was irrelevant, and that, even if it was relevant, IBES five-year data was the best estimate thereof. See Rehearing Request at 21-22, reprinted in J.A. 221-22. This tactic proved to be fruitless, for the Commission reasonably decided to adhere to its two-stage DCF model after concluding that it properly applied in this context. See Michigan Wis. Pipe Line Co. v. FPC, 520 F.2d 84, 89 (D.C. Cir. 1975) ("There is no question that the Commission may attach precedential, and even controlling weight to principles developed in one proceeding and then apply them under appropriate circumstances in a stare decisis manner."). Thus, to the extent that Williston Basin's arguments on this score reflected efforts to skirt or modify, rather than comply with, the Commission's preferred DCF policy, the Commission acted reasonably in rejecting them.

In summary, we find that the question of whether the DCF model must incorporate some long-term growth factor was clearly raised, considered, and resolved by the Commission. We conclude, therefore, that Williston Basin is not entitled to yet another opportunity to oppose the application of that policy to this rate case.

Our inquiry does not end here, however, for a critical issue remains with regard to the Commission's implementation of its two-stage growth projection-specifically, the appropriate weight to be given to the short-and long-term data in this model. In performing the DCF analysis in this case, the Commission averaged these data, relying on the general approach used in prior proceedings. The Commission supported this method by explaining that it lacked the information necessary to predict the duration of the short and long terms, as well as the rate at which growth would transition to maturity. See Williston Basin, 81 F.E.R.C. at 61,176. As a result, the Commission decided "to give [these periods] equal weight" in applying the "well-accepted constant growth model . . . to determine an average constant growth over time." Id.

During the pendency of this appeal, however, the Commission shifted course, finding in the context of a different proceeding that short-term growth projections should receive a two-thirds, rather than one-half, weighting in this analysis. See Transcontinental Gas Pipe Line Corp., 84 F.E.R.C. ¶ 61,084, at 61,423 (1998). The Commission concluded that:

"While determining the cost of equity nevertheless requires that a long-term evaluation be taken into account, long-term projections are inherently more difficult to make, and thus less reliable, than short-term projections. Over a longer period, there is a greater likelihood for unanticipated developments to occur affecting the projection. Given the greater reliability of the short-term projection, we believe it is appropriate to give it greater weight. However, continuing to give some effect to the long-term growth projection will aid in normalizing any distortions that might be reflected in short-term data limited to a narrow segment of the economy."

Id. In other words, the Commission essentially found that the method of averaging short-and long-term projections used in this case gave undue weight to the long-term data.

Because Transcontinental appears to reflect a significant shift in Commission policy with regard to the DCF analysis, we conclude that the Commission is obligated to reconsider the application of that policy to Williston Basin. See Panhandle E. Pipe Line Co. v. FERC, 890 F.2d 435, 438-39 (D.C. Cir. 1989). In Panhandle, the Commission had rejected the pipeline's tariff sheets, based in part on the agency's policy against "capacity brokering." While the matter was on appeal to this court, the agency revised its policy, determining that capacity brokering should be considered on a case-by-case basis. See id. at 438. We held that "[w]hen an agency changes a policy or rule underlying a decision pending review, the agency should immediately inform the court and should either move on its own for a remand or explain how its decision can be sustained independently of the policy in question." Id. at 439 (citation omitted).

Notwithstanding the admonishment in Panhandle, Commission counsel contended at oral argument that Transcontinental does not require a remand in the present case. Rather, according to counsel, Transcontinental has no bearing on this case, because Williston Basin never discussed how the growth factors should be weighted in the DCF model. We reject this view as too simplistic. While preserving the basic two-stage approach of Ozark, the Commission in Transcontinental explicitly determined that long-term growth projections can be unreliable and therefore should be given a lesser weight in the DCF model. See Transcontinental, 84 F.E.R.C. at 61,423. Similarly, Williston Basin, although it did not propose a re-weighting of the growth projections per se, relied in large part on the shortcomings of long-term data in advocating sole reliance on the IBES data. See Rehearing Request at 13, reprinted in J.A. 213. Clearly subsumed within the argument that the long-term data should receive no weight is the argument that the long-term data should receive a lesser weight.

Commission counsel also attempted to distinguish Panhandle by characterizing that case as involving a "reversal," rather than a mere "revision," of Commission policy. We find this argument equally unavailing. For one thing, in Panhandle, we referred to the intervening policy change as a "revision," see 890 F.2d at 439, which belies the suggestion that the relevance of that case is limited to instances in which an agency makes an about-face. Moreover, the instant case itself implicates important policy matters that have concerned the Commission in multiple rate adjudications over the course of the past half decade. See, e.g., Northwest Pipeline Corp., 79 F.E.R.C. ¶ 61,309 (1997); Williams Natural Gas Co., 77 F.E.R.C. ¶ 61,277 (1996); Panhandle E. Pipe Line Corp., 71 F.E.R.C. ¶ 61,228 (1995); Ozark Gas Transmission Sys., 68 F.E.R.C. ¶ 61,032 (1994). While the Commission's paramount change in policy was its incorporation of a two-stage growth rate in the DCF model, its determination of the appropriate weights to be assigned the various growth projections is central to any application of this policy. On this score, even the Commission conceded that a re-weighting of the short-and long-term growth factors would have a substantial impact, in dollar terms, on Williston Basin's rates.

Thus, we find that, in light of the Commission's recent refinement of its two-stage DCF model, Williston Basin may be entitled to a re-calculation of its rate of return on common equity. Accordingly, we remand this matter to the Commission so that the agency can reconsider whether the IBES five-year projections advocated by Williston Basin should receive a greater weighting in the DCF analysis, and, if so, to implement this change. Cf. NLRB v. Food Store Employees Union, Local 347, 417 U.S. 1, 10 n. 10 (1974) ("[A] court reviewing an agency decision following an intervening change of policy by the agency should remand to permit the agency to decide in the first instance whether giving the change retrospective effect will best effectuate the policies underlying the agency's governing act."); National Fuel Gas Supply Corp. v. FERC, 899 F.2d 1244, 1249-50 (D.C. Cir. 1990) (referring, in another context, to the "general principle that an agency should be afforded the first word on how an intervening change in law affects an agency decision pending review").

2. Adoption of GDP as the Long-Term Growth Factor

Bearing in mind our earlier conclusion that the Commission properly required Williston Basin's rate of return on common equity to reflect long-term, as well as short-term, growth expectations, we turn now to the Commission's particular selection of the GDP for that purpose. According to Williston Basin, the Commission's July 1997 Order adopting GDP as its measure of long-term growth was a "bolt from the blue"-an unexpected outcome that was untested at the hearing and unsupported by the record. Thus, Williston Basin's contentions reduce essentially to a claim of inadequate notice concerning the possibility that the Commission would reach the result that it did, as well as several subsidiary claims challenging the result itself.

In its July 1996 Order, the Commission established a hearing for the purpose of determining the appropriate long-term growth factor to be used in the DCF model. See Williston Basin, 76 F.E.R.C. at 61,390. In particular, the Commission found that the parties "need[ed] an opportunity to cross-examine the proponents of using the DRI data, or any other long term growth projection, to determine whether the projections are properly used." Id. (footnote omitted). Following this hearing, however, the Commission shifted tack. Notwithstanding the fact that it had summarily adopted the DRI data in its July 1995 Order, that the ALJ had accepted the DRI data after considering the parties' positions at the hearing, and that the DRI data had been used in Ozark, the Commission determined to use instead an economy-wide projection based on GDP data. Moreover, notwithstanding its earlier position that a hearing was needed concerning the suitability of DRI data for the two-stage DCF model, the Commission refused Williston Basin's request for such a hearing on the use of GDP data.

It is well-established that "[a] party is entitled . . . to know the issues on which decision will turn and to be apprised of the factual material on which the agency relies for decision so that he may rebut it. Indeed, the Due Process Clause forbids an agency to use evidence in a way that forecloses an opportunity to offer a contrary presentation." Bowman Transp., Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 288 n. 4 (1974); see also Hatch v. FERC, 654 F.2d 825, 835 (D.C. Cir. 1981) (same); United Gas Pipe Line Co. v. FERC, 597 F.2d 581, 586-87 (5th Cir. 1979) ("The law will not tolerate . . . after-the-fact, in fact retroactive, imposition of standards," especially where there is "no evidence either to support or justify" the new standard.). Our present concern centers, then, on whether the Commission's order setting the long-term growth matter for hearing provided Williston Basin with adequate notice of the issues that would be considered, and ultimately resolved, at that hearing. In particular, we question whether Williston Basin had reason to know that an economy-wide projection based on GDP data was at issue and, also, whether the Commission's judgment on this score followed logically from the testimony and other evidence adduced at the hearing. In addition, we question whether substantial record evidence supports the actual GDP figure adopted by the Commission for use in calculating Williston Basin's rate of return on common equity.

As we perceive it, the Commission's decision progressed in two relatively distinct steps: first, the Commission expanded the scope of its long-term growth factor from the natural gas industry to the economy as a whole, as reflected in the GDP; and second, the Commission adopted the average of two GDP estimates contained in a record exhibit as the long-term growth factor to be used for the newly-defined DCF model in this case. The Commission's first step-its decision to adopt an economy-wide approach-reflected a well-reasoned and supported outgrowth of the matter under consideration, namely, the appropriate long-term growth factor to be used in the DCF analysis. The Commission established the hearing in broad terms, inviting the parties both to advocate the appropriate data to be used in general, and to challenge the use of DRI data in particular. See Williston Basin, 76 F.E.R.C. at 61,390. Moreover, the testimony adduced at the hearing demonstrated that major investment houses used an economy-wide approach to projecting long-term growth, that such an approach was supported by practical economic considerations, and that existing industry-specific approaches imperfectly reflected investor expectations and made unfounded economic assumptions. See Williston Basin, 79 F.E.R.C. at 62,388-90. Finally, whether or not it is true, as Commission counsel now suggests, "that GDP [is] virtually synonymous with the economy as a whole," Brief for Respondent at 42, we have little doubt that GDP is among the most commonly used and widely available measures of economy-wide growth. In short, we are convinced that FERC's decision to expand the scope of its long-term analysis reflected a reasoned progression from the issues set for hearing, and that the data informing that decision was in the record and discussed at the hearing.

However, we find that the Commission's second step, by which it reached the precise long-term growth estimate of 5.85 percent, lacked adequate support in the record. As discussed above, we do not take issue with the Commission's decision, on a general level, to use GDP data in estimating long-term growth. The problem, in our view, is that there are conceivably a number of estimates of GDP created by different entities and based on different economic assumptions. Yet, FERC, after substantially modifying the scope of its long-term analysis, and without forewarning to the parties, simply teased two GDP figures from the background section to a single exhibit to reach the result here at issue. See J.A. 315; Williston Basin, 79 F.E.R.C. at 62,390. This was a bizarre conclusion to the hearing. It is undisputed that the record in the hearing had been created largely in response to a specific concern over the suitability of industry-specific DRI data for use in the DCF model. No party at the hearing had presented, advocated, or even mentioned the use of GDP data. In light of these circumstances, we find that the Commission neither explained nor supported its choice of the DRI and EIA estimates of GDP contained in the existing record. Accordingly, we remand to the Commission for further proceedings on this issue.

C. Ad Valorem Taxes

Next, we address Williston Basin's proposed ad valorem tax expense, which the Commission rejected as inconsistent with test period principles. In its filing, Williston Basin sought to recover the additional ad valorem taxes associated with plant increases during the test period by applying the effective tax rate for the 1991 year in each state in which it owned property to its total capital investment in those states, as adjusted for additions during the test period. Williston Basin contended that this represented a proper adjustment to reflect changes that were "known and measurable" within the meaning of the Commission's regulations. See 18 C.F.R. § 154.303(a)(4). The plant additions occurring during the test year will, it argued, produce higher tax assessments by the relevant states for the effective period of the rates.

The Commission refused to approve this approach, however, requiring instead that Williston Basin support its filing with the actual ad valorem tax liability incurred during the test period. See Williston Basin, 76 F.E.R.C. at 61,384. In reversing the ALJ on this point, the Commission explained that Williston Basin's proposed tax liability was too speculative:

"While the plant additions occurred within the test period, the effect on [Williston Basin's] ad valorem taxes of the installation of those facilities is not known and could not be measured with reasonable accuracy during the test period in this case. The determination of the exact ad valorem tax effect is a local matter involving local valuation and tax assessment procedures. Further, because of depreciation, existing facilities may generate lower ad valorem tax liability than as reflected in the test period data, thereby offsetting in some unknown way the potential ad valorem tax liability."

Williston Basin, 72 F.E.R.C. at 61,363. According to the Commission, "the actual costs for any expense or tax during the test period generally reflects the best evidence of what the company can expect to incur in the future." Williston Basin, 76 F.E.R.C. at 61,384.

The Commission's ruling on this issue reduces to its basic position that Williston Basin's proposed calculation was "conjecture," because "too many variables" could influence Williston Basin's actual tax liability during the effective period of the rates. Id. Williston Basin counters that it obviated these concerns by assuming the tax rate in effect during the test period and adjusting only that variable-plant balance-for which changes were known and measurable at the time of filing. According to Williston Basin, this approach is consistent with the methodology approved by the Commission in a prior Williston Basin proceeding, Williston Basin Interstate Pipeline Co., 56 F.E.R.C. ¶ 61,104 (1991) ("1991 Order").

In the prior proceeding on which Williston Basin relies, Montana had significantly increased the allocation of Williston Basin's pipeline property to that state, but had allowed a phase-in of the higher allocation percentage over a three-year period, from 1986 through 1988. The Commission apparently found that this phase-in produced known and measurable increases in the allocation on which the taxes that Williston Basin owed Montana were based. On this ground, the Commission approved a methodology reflecting the increases that occurred during the applicable test period, which ended January 31, 1988. This approach involved two steps: first, the 1987 ad valorem taxes paid were divided by the 1986 year-end plant balance to determine the relevant tax rate; and second, that rate was applied to a tax base representing the plant balance as of December 31, 1987. See Williston Basin, 56 F.E.R.C. at 61,382-83. As we see it, the new plant capital at issue in this case is analytically equivalent to the phase-in of property allocation permitted in the earlier Williston Basin proceeding. Therefore, the upward adjustment allowed by the Commission in the 1991 Order to reflect increases in plant balance during the test year is apparently of the same variety proposed by Williston Basin in the present case.

Although the Commission was not strictly bound to follow the methodology approved in the prior Williston Basin proceeding, it was obligated to articulate a principled rationale for departing from that methodology. See Gilbert v. NLRB, 56 F.3d 1438, 1445 (D.C. Cir. 1995) ("It is . . . elementary that an agency must conform to its prior decisions or explain the reason for its departure from such precedent."); National Conservative Political Action Comm. v. FEC, 626 F.2d 953, 959 (D.C. Cir. 1980) (same). In other words, "[r]easoned decision-making requires treating like cases alike." Hall v. McLaughlin, 864 F.2d 868, 872 (D.C. Cir. 1989). The Commission's task on this score was not unduly onerous, for we have held that "[w]here the reviewing court can ascertain that the agency has not in fact diverged from past decisions, the need for a comprehensive and explicit statement of its current rationale is less pressing." Id. Thus, an agency's findings will be upheld, "though of less than ideal clarity, if the agency's path may reasonably be discerned." Greater Boston Television Corp. v. FCC, 444 F.2d 841, 851 (D.C. Cir. 1970). In this case, the Commission failed to satisfy even this relatively forgiving standard.

The Commission purported to distinguish the prior Williston Basin proceeding, holding that:

"the salient finding by the Commission was that the State of Montana had prescribed higher tax rates and that it had allowed a phase-in of those rates. Therefore, there was ample rationale for finding there that the expenses claimed by [Williston Basin] were based on adjustments for known and measurable changes that would occur during the period the rates were to be in effect. Here, this is simply not the case. While the plant additions occurred within the test period, the effect on [Williston Basin's] ad valorem taxes . . . is not known and could not be measured with reasonable accuracy during the test period."

Williston Basin, 72 F.E.R.C. at 61,363. The Commission elaborated on rehearing that "there was nothing speculative in the increase because the same valuation of the properties was used and only the percentage allocation to Montana was changed. That type of change is different than an increase in the value of the property which [Williston Basin] claims in this proceeding." Williston Basin, 76 F.E.R.C. at 61,384.

The 1991 Order may indeed be distinguishable on the basis of property valuation, which is a critical step in assessing ad valorem tax liability. Property valuation is performed by individual states in accordance with local practice and, often, the discretion of individual assessors. FERC may be right that valuation was speculative in the present case, because the new plant additions had not yet been assessed by the relevant states. In fact, Williston Basin proposed to rely solely on its own investment in plant facilities, even though the valuation process almost certainly includes consideration of other factors. Thus, the impact of the plant increases here may not have been "known and measurable," as required by the test period regulations. By contrast, at least as we see it, the "plant addition" at issue in the 1991 Order resulted simply from the phased-in allocation of existing plant, which had presumably already been valued. In that case, then, there was nothing uncertain: the Commission took the known Montana tax rate during the test period and applied that rate to the known tax base-a higher percentage of the previously-assessed value of Williston Basin's property-during the test period.

Assuming the facts as we do, and assuming that our reasoning mirrors the Commission's intended reasoning, we think that this distinction may be compelling. The sticking point for us, then, is the extent to which the Commission's orders compel us to make such assumptions. In other words, we are simply unable, on the record as it now exists, to assure ourselves either that this distinction holds water, or that this analysis does, in fact, capture the Commission's reasoning. For example, the Commission described the 1991 Order variably as involving a phase-in of tax rates and a phase-in of property allocation. Yet, tax rates are not at issue here, because Williston Basin voluntarily assumed the tax rate in effect during the test period.

Moreover, even assuming that we have correctly identified the distinction upon which the Commission relied, we are not confident that the record supports this distinction. Our uncertainty derives principally from the lack of clarity in the 1991 Order, and the Commission's failure to explain that order here. Specifically, because we do not know the precise calculations and dollar amounts involved in the prior case, we are not sure that the Commission's prior ruling was based solely upon the phase-in of plant allocation. Indeed, our own rough calculations suggest that approximately $67,000 in disputed ad valorem tax expenses is not explained by the phase-in. The logic of the Commission's holding in the 1991 Order, particularly as it is couched in broad language, might support the inference that this discrepancy reflects additional plant increases of the nature involved in this case. If that is the case, the supposed distinction, based on the uniquely known and measurable character of the phased-in plant allocation, rings hollow.

On its face, the 1991 Order refutes the broad principle on which the Commission relied in rejecting Williston Basin's ad valorem tax expense in this case-namely, that a pipeline may only use taxes actually paid during the test period to support its estimate of taxes in its compliance filing. Particularly in light of this contradiction, we believe that the Commission failed to provide a clear and well-supported explanation of why the methodology used in the 1991 proceeding was not appropriate here. We recognize that the Commission may, in fact, have a persuasive ground for distinguishing this case from the 1991 Order. On remand, then, FERC will have an opportunity to offer a coherent rationale to support its judgment and, also, to show that the cited rationale is supported by the record.

D. Throughput

Finally, we turn to the Commission's decision to reject Williston Basin's proposed throughput volume. In its filing, Williston Basin sought to adjust its base period data to account for decreases in throughput resulting primarily from bypasses of its transmission system by two major suppliers.

At the time of filing, Williston Basin expected these bypasses to occur before the adjustment period ended on October 31, 1992. Thus, it argued that they represented "known and measurable" changes to its actual experience during the test period, which could properly be reflected in its rate filing. See 18 CFR § 154.303(a)(4).

The source of contention here arises from the fact that the bypasses did not actually occur until after the test period had ended. In other words, due to the timing of these rate proceedings, actual adjustment and post-test period data was available by the time the Commission considered the matter. This data showed that the bypasses were not completed during the test period, but were completed very shortly thereafter. Thus, if Williston Basin was permitted to include this adjustment, it would over-recover for three or four months of the rate period commencing November 1, 1992. However, if Williston Basin was not permitted to include this adjustment, it would under-recover for eight or nine months of that rate period (assuming, that is, that it did not file a new rate case to cover that period).

The crux of Williston Basin's position is that the Commission should accept its throughput projection, because the estimate was reasonable when made. The ALJ agreed, concluding that "under established Commission precedent, a test year projection may be set aside only if its is shown to have been unreasonable when made." Williston Basin, 68 F.E.R.C. at 65,069. Both Williston Basin and the ALJ relied chiefly upon Public Service Co. of Indiana, 7 F.E.R.C. ¶ 61,- 319 (1979), aff'd sub nom. Indiana Municipal Electric Ass'n v. FERC, 629 F.2d 480 (7th Cir. 1980), a proceeding in which the Commission accepted an electric utility's test period cost-of-service estimate-even though a particular component of its projection ultimately proved exaggerated-because the estimate was reasonable when made and did not yield unreasonable results. See Public Service, 7 F.E.R.C. at 61,701-02.

The Commission, however, rejected this view in the orders below, holding that whether or not Williston Basin's projection was reasonable when made, "where the pipeline . . . projects an event to occur before the end of the test period, but in fact that event does not become effective within the required time period, the Commission generally requires that event not be reflected in the pipeline's rates." Williston Basin, 72 F.E.R.C. at 61,382. In the Commission's view, the alleged reasonableness of Williston Basin's estimate went only to its compliance with filing requirements under 18 CFR § 154.303. See id. It did not "preclude the Commission from considering updated data in deciding the ultimate question of what rates should be found just and reasonable for the relevant periods," id.; nor did it "endow [the bypasses] with the required characteristics to be allowed as an adjustment." Williston Basin, 76 F.E.R.C. at 61,388. Thus, the Commission refused the proposed adjustment, adopting instead the FERC staff's proposal, which based throughput levels on actual data for the twelve months immediately preceding the effective date of the rates. See Williston Basin, 72 F.E.R.C. at 61,382.

We begin our analysis of this issue by recognizing a point that, while seemingly semantic, may bear on the relative merit of the parties' arguments-that is, who sought the "adjustment" in this case? On the one hand, from the Commission's standpoint, Williston Basin asked for an adjustment to its base period data to reflect a decline in throughput that was projected to, but did not, occur during the applicable "adjustment period." Under this view, the Commission's decision was apparently consistent with the test period regulations governing pipelines, which on their face allow only adjustments for changes that will occur before the end of the test period. See 18 CFR § 154.303(a)(4). Not only is it undisputed that the changes in this case did not occur during the test period, but the Commission actually noted that, "[h]ad the bypasses taken place in the test period, . . . the adjustment would have been permitted." Williston Basin, 76 F.E.R.C. at 61,388. However, the Commission found that, because the bypasses did not occur during the test period, and because it could not be known during the test period exactly when they would occur, the use of post-test period data showing that they did occur shortly after that time expired was "too much in the nature of hindsight." Williston Basin, 72 F.E.R.C. at 61,383. Moreover, it determined that the position advocated by Williston Basin would give pipelines an incentive to selectively project only adjustments that would prove favorable to them if they actually occurred — i.e., increases in costs and decreases in throughput, see Williston Basin, 76 F.E.R.C. at 61,388 — which is, in fact, what Williston Basin appears to have done in this case.

On the other hand, however, Williston Basin labels the Commission as the party that sought an adjustment, because Williston Basin wanted to use the estimate it made upon filing this rate case, while the Commission wanted to adjust that estimate to account for actual data during the adjustment portion of the test period. Under this view, the Commission's ruling appears less reasonable, for Williston Basin is quite correct in observing that the Commission in the past has declined to disturb test period estimates that were proven inaccurate in light of later data if those estimates were reasonable when made and did not produce unreasonable consequences. See, e.g., Indiana Mich. Mun. Distribs. Ass'n v. FERC, 659 F.2d 1193, 1198-99 (D.C. Cir. 1981); Public Service, 7 F.E.R.C. at 61,701. In this case, the Commission conceded that Williston Basin's throughput projection was reasonable when made, and did not even attempt to explain why the projection, although it in fact occurred within a short time after the test period, was so erroneous as to yield unreasonable results. Yet, it refused to let Williston Basin's projection stand. Thus, instead of analyzing Williston Basin's claim under the framework of the above cases, the Commission simply ignored them, citing them only insofar as it summarized the parties' arguments, and leaving us to guess as to why they should not apply here.

As with the ad valorem tax issue, we once again find ourselves able to surmise a solid basis for distinction. Here, it is the simple fact that the vast majority of cases espousing the principle of "reasonable when made" involved electric utilities, rather than natural gas pipelines. See, e.g., Public Service, 7 F.E.R.C. ¶ 61,319. Although the Commission employs a test period methodology for setting rates in both contexts, the applicable regulations differ considerably in their treatment of estimates. As noted, the rates for pipelines are based on actual data for a one-year period, as adjusted to reflect known and measurable changes that will occur over the following nine months. See 18 C.F.R. § 154.303. These pipeline regulations do not appear to make use of estimates at all; indeed, they require test period projections to be updated with actual data for the adjustment period as it becomes available. See id. § 154.311(a), (b). By contrast, the rates for utilities are derived from two distinct periods: actual data for the year known as "Period I" and estimated data for the year known as "Period II." See id. § 35.13(d)(1), (2). These utility regulations do not explicitly require that Period II estimates are known and measurable, or that they will in fact occur during the test year. See id. § 35.13(d)(2)(i).

As we interpret them, then, the regulations applying to utilities vest far greater weight in estimates than do the regulations governing pipelines. It is plainly rational to infer from these differences in regulatory context that the "reasonable when made" formulation applies only to a utility's Period II estimates and not to a pipeline's projected adjustments. In short, applying the rule of Public Service comports with the plain language of the utility regulations, but would require the Commission to recognize an exception to the pipeline regulations. The Commission may therefore reasonably have determined that Public Service was inapposite in this context.

This explanation for the Commission's decision would be satisfactory but for two shortcomings. First, although this distinction may seem fairly obvious once recognized, the fact remains that the Commission itself did not articulate, or even allude to, it in the orders below. See American Pub. Transit Ass'n v. Lewis, 655 F.2d 1272, 1278 (D.C. Cir. 1981) (citing SEC v. Chenery Corp., 332 U.S. 194, 196 (1947)). Second, both the Commission and courts have, in the past, essentially ignored this issue, citing test period precedent interchangeably in utility and pipeline cases. See, e.g., Exxon, 114 F.3d at 1263 n. 23; Distrigas of Mass. Corp. v. FERC, 737 F.2d 1208, 1220 (1st Cir. 1984); National Fuel Gas Supply Corp., 51 F.E.R.C. ¶ 61,122, at 61,334 n. 53 (1990). Thus, we have no way of knowing whether the Commission's desired approach is to recognize this broad distinction between the regulations, or to intentionally skate over the differences in the terms of the regulations, intending instead that the test period concept operate identically in the utility and pipeline contexts.

By failing to distinguish the authority on which Williston Basin relied in support of its position, and which at least superficially contravened the Commission's ruling, the agency appeared to "gloss over or swerve from prior precedents without discussion," Greater Boston, 444 F.2d at 852, thereby foregoing reasoned decision-making. It may well be that the Commission had in mind this, or another, rational explanation for its ruling. But as we have noted in the past, "[w]ithout any explicit recognition by the Commission that the standard has been changed, or any attempt to forthrightly distinguish or outrightly reject apparently inconsistent precedent, we are left with no guideposts for determining the consistency of administrative action in similar cases, or for accurately predicting future action by the Commission." Hatch, 654 F.2d at 834-35 (footnote omitted). As such, we must remand to the Commission on this issue as well.

III. Conclusion

For the foregoing reasons, Williston Basin's petition for review is granted in part and denied in part, and the matter is remanded to the Commission for further proceedings.

So ordered.