Villages of Chatham and Riverton, Illinois v. Federal Energy Regulatory Commission, Central Illinois Light Company, Intervenor

Opinion for the Court filed by Circuit Judge McGOWAN. McGOWAN, Circuit Judge: The villages of Chatham and Riverton, Illinois, seek direct review in this court of certain determinations by the Federal Energy Regulatory Commission (FERC) that were made in its consideration of a wholesale electric power tariff filed by Central Illinois Light Company (Cilco) under section 205 of the Federal Power Act, 16 U.S.C. § 824d (1976). As part of its proposal of higher rates, Cilco, in accordance with Commission regulations, submitted projections of its expenses and the villages’ demand for electricity. The villages challenge several of those estimates, as well as the Commission’s decision that the utility did not have excess generating capacity. For the reasons appearing below, we vacate and remand for further consideration the Commission’s adoption of the projections made by the utility of the villages’ coincident peak demand, but we affirm the Commission with respect to Cilco’s estimates of purchased power credits and income tax expense. We also find that substantial evidence supports the Commission’s conclusion that the utility did not have excess reserve generating capacity. I On July 29, 1976, Central Illinois Light Company filed with the former Federal Power Commission 1 a tariff increasing *26 rates for wholesale electric service to three customers, the Villages of Chatham and Riverton, Illinois, and the Corn Belt Electric Cooperative. See Central Illinois Power Company Initial Decision on Rate Increases, No. ER76-819, at 1 (Jan. 24, 1979) [hereinafter referred to as Initial Decision]. Cilco stated that the proposed tariffs would increase wholesale rates to Chatham by 68% or $333,000, to Riverton by 68% or $195,000, and for the Corn Belt Cooperative by 48% or $187,000. Initial Decision at 2. On August 27, 1976, the Commission suspended the rate increase until February, 1977. In a series of orders, the FPC set the new tariff for hearing and permitted the two municipalities and the cooperative to intervene in the administrative proceedings. Id. Corn Belt did not file exceptions to the Initial Decision, Record at 2428, and is not a party to this direct review proceeding. Following a hearing, the Commission’s administrative law judge issued an Initial Decision on Jan. 24,1979, which ruled upon, inter alia, several of the issues that form the basis of the instant petition for review: (1) Cilco’s reserve generating capacity, (2) the discrepancy between Cilco’s projected and actual purchased power revenues, and (3) Cilco’s projections of the villages’ peak demand. In rejecting the villages’ contention that Cilco had too much reserve capacity and should not be allowed to charge customers for the allegedly useless excess, the ALJ noted that record evidence indicated Cilco’s actual 1977 reserve capacity to be 20.2% of its total generating capacity, a figure “only slightly higher than the margins of 15-20% typically recommended by the Commission.” Initial Decision at 12. The other issues that were resolved by the Initial Decision and that the villages contest in this court were rooted in Cilco’s use of a future test period to estimate its expenses and its wholesale customers’ demand for power, as required by FERC regulations, 18 C.F.R. § 35.13(b)(4)(iii) (1980). In addition to filing actual data for calendar year 1975, or period I, Cilco proffered expense and demand estimates for period II, a test period comprising the twelve months from July 1, 1976 to June 30, 1977. Record at 38. By the time of the hearing, May 3 — 4, 1978, actual figures for period II were available, see id., and the villages used them in an effort to cast doubt upon the reliability of Cilco’s estimates. Record at 89-97. With respect to the amount of revenues Cilco obtained from the sale of electricity to other utility companies, the utility estimated period II revenues from the sale of power to other utilities, or purchased power credits, of $3.4 million. Initial Decision at 16. The villages noted that in the calendar year 1977, Cilco actually received just under $10 million from these sales and urged an increase in this figure. Id. The villages took an interest in this because Cilco treats these revenues as credits to the production expenses recovered from its customers. See Record at 857 (Cilco cost-of-service study for period II). The ALJ rejected the villages’ challenges to the $3.4 million figure by stating that (1) there was “no evidence that the projection was unfounded when made,” (2) “[t]he basis of the estimate is not being challenged as deficient,” and (3) the difference between actual and projected revenues was “not substantial,” especially in view of Cilco’s understatements of various production expense items. Initial Decision at 16. The most significant surviving dispute concerns the estimates of the villages’ peak demand that Cilco used to assign responsibility for demand-related costs, such as those incurred in the construction of generating facilities. 2 To allocate the wholesale *27 customers’ proper share of these capacity-related expenses, Cilco estimated the peak demand of these customers at the time when its entire customer base was demanding the most power. See Record at 49, 873. Since this figure represents the customers’ peak at the time of the system peak, the parties refer to it as coincident peak demand. See FERC Br. at 7 & n.2. A higher coincident peak demand leads to greater capacity costs allocated to those customers, who therefore must pay more to cover their increased share of costs. The villages argue that they have been allocated more than their share of capacity costs because the utility’s estimates of their coincident peak demands are too high. At the hearing, the villages had introduced the actual coincident peak demand figures for period II, which were significantly lower than those forecast by Cilco. The Commission, before this court, admitted that “where CILCO projected a growth in the coincident peak demand of the Villages of 15.9 percent, between Period I and Period II, the actual growth was 7.05 percent.” FERC Br. at 8; see note 7 infra. The villages raised a number of arguments intended to demonstrate that the Commission should use the actual, rather than the projected, figures. While we discuss some of them in the context of arguments made on this appeal, we note that the ALJ’s initial opinion, in its six-paragraph discussion of the coincident peak demand figures, addressed only two of them. The Administrative Law Judges, mentioning that the villages objected “primarily on the basis of two points,” Initial Decision at 16, said, in response to one objection, that the utility’s failure to forecast demand growth for the wholesale customers on the same basis as for its retail customers did not render the estimates unusable for rate-making. Cilco forecasts retail growth by breaking down its customers into classes, such as residential, commercial, and industrial, and projecting demand growth by class. Id. at 16-17. It did not do so in the case of the villages, explaining that it had no statistics on the retail breakdown of the villages’ customers, because it did not meter those customers. Id. The ALJ accepted this explanation, stating that the utility’s estimates were “based on the best information available at the time.” Id. at 17. The Initial Decision also dismissed the villages’ objections to the utility’s forecast that the entire Cilco system would peak between 4 and 7 p. m. Id. They had observed that the actual figures for period II and prior years showed a system peak falling at other times in several months, but the ALJ concluded that the projection was “quite accurate,” because the actual system peak fell between 4 and 7 p. m. in eight of the twelve months of 1977. Id. Having considered and rejected the villages’ two objections, the ALJ credited Cil-co’s coincident peak demand estimates because the villages had “not demonstrated that the Company’s method of making demand and energy projections was unreasonable,” and that the “deviation from actual experience was not substantial enough to warrant a spot adjustment to the estimates.” Id. The ALJ did not explicitly determine that Cilco had established the reasonableness of its estimates, as the utility was required to do by section 205(e) of the Federal Power Act, 16 U.S.C. § 824d(e) (1976). See Part II infra. Both Cilco and the villages excepted to various aspects of the ALJ’s decision. In their brief on exceptions, the villages argued, inter alia, that the ALJ erred in finding that Cilco had no excess reserve capacity, that he should have adjusted the purchased power credits from the estimated to the actual figure, and that he should not have accepted Cilco’s coincident peak demand estimates. Record at 2304-09, 2310-20. The villages asserted the unreasonableness of these figures on the grounds that (1) the actual figures were considerably different, (2) Cilco made no effort to ask the villages about the composition of their loads or projected growth among various custom *28 er classes and (3) the actual system peak often occurred either before 4 or after 7 p. m. Id. at 2314-20. In their exceptions, the villages for the first time asked the Commission to lower Cilco’s income tax expense figure from 48%, the rate prevailing when the tariff was filed, to the new rate of 46%. Record at 2322. In Central Illinois Light Company: Opinion and Order on Application for Rate Increase, Opinion No. 81, No. ER76-814 (March 20, 1980) [hereinafter referred to as Opinion No. 81], the Commission refused to reduce the rate, citing its policy of “declining to make post-period II adjustments, ...” 3 The Commission upheld without comment the ALJ’s disposition of the purchased power credits and coincident peak demand estimates. Opinion No. 81 at 22. The villages petitioned to rehear certain issues, including the use of those estimates. For the first time, the villages argued that the utility’s estimates were implausible on their face because the estimated coincident peak demand figures were higher in many months than the estimated noncoincident peak demand numbers. Record at 2472-73. Since a customer’s noncoincident peak demand is its greatest demand for electricity at any time within a chosen period of time, see FERC Br. at 24 n.7, it is impossible for that figure to be less than the customer’s demand when the system peaks, although it is at least theoretically possible for noncoin-cident peak demand to equal coincident peak demand. 4 The Commission denied the request by taking no action within the thirty days allotted by statute, 16 U.S.C. § 825/ (a) (1976). FERC Br. at 2. The villages petitioned this Court for review of the Commission’s various decisions, and Central Illinois Light Company moved for and was granted leave to intervene before this Court in support of the Commission’s position. II Although the use of projected expense and demand figures in ratemaking has been upheld by this Court, American Public Power Association v. FPC, 552 F.2d 142 (D.C.Cir.1975), we cannot allow the practice to subvert the Federal Power Act’s requirement that “[a]t any hearing involving a rate or charge sought to be increased, the burden of proof to show that the increased rate or charge is just and reasonable shall be upon the public utility, ...” 16 U.S.C. § 824d(e) (1976). In issuing the regulations that required the submission of period II data, the Commission was careful to point out that nothing in its new rules should be construed as an effort to relieve the petitioning utility of its statutory burden of proof. It stated: “[W]e will not approve rates based on unsubstantiated cost estimations. The burden will be on such companies to establish the validity and ac- *29 curacy for each of their cost estimates.” Filing of Electric Service Tariff Changes, 50 F.P.C. 125, 127 (1973), aff’d sub nom. American Public Power Association v. FPC, 522 F.2d 142 (D.C.Cir.1975) (regulations now codified at 18 C.F.R. § 35.13(b)(4)(iii) (1980)). Relying on that language, the Seventh Circuit has upheld the Commission’s decision to reject utility cost estimates in light of the actual figures. Public Service Co. v. FERC, 575 F.2d 1204 (7th Cir. 1978). The utility had estimated purchased power costs of $15 million for a future test period in which the utility had, according to the actual results later available to the parties, incurred no purchased power costs and had sold over $2 million of electricity. 575 F.2d at 1215. Rejecting the utility’s argument that adjusting the $15 million figure to zero would violate the Commission’s policy against spot adjustments, id. at 1216, the Commission adopted the actual results, id. at 1215-16. Judge Sprecher, in upholding the Commission, stated PSCI [the petitioning company] contends that the FPC’s actions violate its long-standing policy against making ‘spot’ adjustments. This argument overlooks ... regulations [that] require the filing utilities ‘to establish the validity and accuracy for each of their cost estimates.’ Since the regulations require each cost estimate to be valid, it is difficult to understand how the FPC could avoid making some ‘spot adjustments’ when a cost estimate is determined to be plainly wrong, such as in this case. Id. (citation omitted). In a later case, the Seventh Circuit upheld the Commission’s refusal to adjust purchased power credits in light of actual figures when the utility had introduced evidence, on which the Commission was entitled to rely, indicating that the actual figures were based in'part on events unlikely to recur. Indiana Municipal Electric Association v. FERC, 629 F.2d 480 (7th Cir. 1980). The court intimated that the future test-year estimates could be upheld if shown to be reasonable when made, even though the estimate was less than half the actual figure. 629 F.2d at 483-84. Concluding that the utility and the Commission had adduced evidence tending to show that the actual figures were aberrational and that the utility’s forecast had a logical basis, the court said that the customers had failed to show the error of relying upon the estimates. Id. The court said it did not think that period II estimates should be rejected merely because they varied from the actual results, since that would supplant period II ratemaking with historic-cost ratemaking contrary to Commission regulations. Id. at 482-83. A panel of this court has had occasion recently to consider whether the Commission could accept future test-period estimates that varied from actual results. Indiana and Michigan Municipal Distributors Association v. FERC, 659 F.2d 1193 (D.C. Cir. 1981). The Commission said that the estimates could be used because the utility had demonstrated that they were a reasonable extrapolation from comparable data available when the estimate was made. Id. at 1198. The court said that if the utility demonstrates that its estimates were reasonable when made, it becomes incumbent upon those contesting the estimate to proffer evidence tending to establish that its use in ratemaking would lead to “an unreasonable result.” Id. at 1198 n. 14. Finding that the only evidence for that proposition was the disparity between actual and estimated results, id., the court affirmed the Commission’s decision to credit the projections. We can infer from these decisions a basis for judicial review of test-year rate-making. The utility must demonstrate that its estimates were reasonable when made, either by explaining the chain of reasoning that it employed to arrive at its projections, or by comparing its estimates with actual data, and so establishing their accuracy. If the utility proves that its estimates were reasonable when made even though at variance with the actual results, the Commission may nonetheless credit those estimates, *30 a decision further substantiated by evidence indicating that the actual figures are based upon unusual or unique occurrences. And of course the Commission is obligated to consider contrary evidence introduced by customers or other parties, as well as the magnitude of the disparity between expectation and reality. We now turn to the challenged estimates in the instant case to determine if the utility has carried its statutory burden of proof “to establish the validity and accuracy for reach of [its] cost estimates,” 50 F.P.C. at 127. Ill A. Ciko’s Estimate of the Villages’ Coincident Peak Demand The villages had raised a number of arguments in their attempt to discredit Cilco’s estimates of their coincident peak demands. Before turning to the responses made by the Commission and by Cilco, we shall dispose of one objection first made by the Commission at oral argument. Counsel for the Commission argued at that time that the villages could not rely upon the discrepancy between noncoincident and coincident peak demand estimates because they had failed to bring that objection to the attention of the Commission during the administrative proceedings. This waiver argument depends upon the proposition that an argument based on previously-adduced evidence and first raised in a petition for rehearing is not brought forward in timely fashion. It is a proposition that we are unable to accept. The text of the statute indicates that a party to a rate proceeding may propound novel arguments based upon evidence already in the record as late as that party’s petition for rehearing and still preserve them for review in the courts of appeals. The statutory provision for judicial review of FERC orders under the Federal Power Act states: “No objection to the order of the Commission shall be considered by the court unless such objection shall have been urged before the Commission in the application for rehearing unless there is reasonable ground for failure so to do.” 16 U.S.C. § 8251(b) (1976). This court, in construing an identically-worded provision of the Natural Gas Act, 15 U.S.C. § 717r(b) (1976), a statute whose administration is also entrusted to the Commission, has stated that any argument brought clearly to the attention of the Commission by the party’s petition for rehearing has been preserved for review in a court of appeals. Belco Petroleum Corp. v. FERC, 589 F.2d 680, 683 (D.C.Cir.1978). Allowing the villages to rely on arguments first made to the Commission on rehearing is in keeping with the purpose of this language limiting the scope of judicial review. In Rhode Island Consumers’ Council v. FPC, 504 F.2d 203, 212 (D.C.Cir.1974) (citations omitted), Judge Leventhal explained that the provision was intended ... to insure that the Commission has an opportunity to deal with any difficulties presented by its action before the reviewing court intervenes. This exhaustion requirement comports with the general function of judicial review to insure that the agency has ‘taken a “hard look” at the salient problems.’ The Commission, squarely presented with the argument that the company’s estimates were absurd because they projected coincident peak demands above noncoincident peak demands for several months, possessed an adequate opportunity to deal with this alleged difficulty. By denying sub siientio the villages’ request for rehearing, it chose not to avail itself of that opportunity. We cannot endorse that decision. For the last third of the test period, Cil-co’s estimates of coincident peak demand (CP) were significantly higher than its estimates of noncoincident peak demand (NCP). The May estimates for Chatham indicated a CP of 5,376, 2,376 Kw or 79% above its NCP of 3,000. See Record at 872-73. The April projections for Riverton were a CP of 2,439 Kw, 639 Kw or 36% above its NCP of 1,800 Kw. See id. Between February and May, 1977, the coincident peaks were projected to exceed the noncoincident peaks by 172 to 573 Kw. See *31 id. This disparity cannot be said to be so trivial as to escape the scrutiny of a reviewing court, especially when the Commission has made no effort in the administrative proceeding to supply an explanation. Cilco, in its brief as intervenor before this Court, states that this theoretically impossible gap between coincident and noncoinci-dent peak demand figures should not trouble either the Commission or this court because the variance might be explained by its use of calendar months in projecting coincident demand and billing months for noncoincident demand. These billing months, Cilco states, “may or may not be coincident with calendar months.” Interve-nor’s Br. at 24-25. Cilco also made that argument, in the same words, to the Commission. See Record at 2409-10. The villages in their reply brief argue that any lack of synchrony between billing and calendar months fails to explain the disparity. The Commission does not state in the record whether it had considered the possible difference in time periods. Leaving aside whether Cilco’s inability to determine with any degree of certainty on what basis it bills the villages should or can be taken to undercut the plausibility of this argument, we think that this matter is one properly to be decided by the Commission in the first instance, 5 and that upon remand the parties are free to restate their cases. It is to be hoped that the Commission will then decide the matter on the record so that we can be sure the parties have had the benefit of the administrative scrutiny of rate increase petitions that the Federal Power Act sought to provide. 6 The villages also raised other objections to the coincident peak demand estimates that require further Commission examination. They introduced evidence before the Commission showing that whereas Cilco had forecast their demand growth to be twice the system average, see Record at 804, 874 (Cilco projections), the actual figures indicated that the villages’ appetite for electricity increased only about one-third faster than the entire system’s. 7 See Record at *32 1016 (figures submitted by villages). They also pointed out that the same data indicated that Cilco had overestimated by 125% their coincident peak demand for electricity, but had overestimated the system load growth by only 50%. Record at 2314. The Commission’s sole comment on these figures was that the “deviation from actual experience was not substantial enough to warrant a spot adjustment to the estimates.” Initial Decision at 17. Ordinarily this exercise of administrative judgment is not of the sort this court will set aside. But because we suspect the Commission’s conclusion was predicated on its faulty assignment of the burden of proof with respect to Cilco’s period II data, we cannot allow it to stand. We said earlier that the burden was on the utility to justify each and every period II cost estimate. One way it could do so would be to demonstrate the congruence of its projections with actual results. To the extent that the Commission was deciding that it had done so, we think that FERC was incorrect. Disregarding disparities of this magnitude to arrive at the judgment that the actual data validates the projections is not consistent with the statutory allocation of the burden of proof to the utility, an allocation that the Commission said, in its order requiring test-year data, that it would respect. Unable to rely upon actual data to support the period II estimates, the utility was thus charged with the task of justifying the methodology that it employed to forecast demand. The villages placed upon the record evidence, which the Commission credited, indicating that Cilco forecasted growth for the villages without breaking down their retail sales into classes. See Part I supra. The AU refused to concede that this difference in method rendered Cilco’s estimates vulnerable, because “statistics regarding the demand of each of those classes is not obtainable from company records,” and because the projections were “based upon the best information available at the time, and [are], therefore, reasonable.” Initial Decision at 17. Cilco, and the Commission as well, apparently thought that section 205 of the Federal Power Act would be satisfied by relying upon information in the company’s files, no matter how inadequate. By Cilco’s own admission, it never asked the villages for information or consultation in load forecasting, explaining that nothing in the record showed that consulting the villages would have produced useful information. Record at 2409. There was also nothing in the record preventing Cilco from surveying the villages’ patterns of load growth on their own, as they had done with their fifty largest industrial customers, Record at 34. The utility’s lack of curiosity about a matter in which it bears the burden of proof cannot redound to its benefit. The most significant obstacle to upholding the Commission’s findings with respect to Cilco’s demand estimates was Cilco’s failure at any time to articulate the reasoning employed to arrive at estimates that, as we have explained above, were (1) contradicted by actual results, (2) theoretically impossible and (3) based upon data a good deal more impoverished than those relied upon by the utility to forecast its retail demand growth. The initial decision attached the proposition that the utility had no better information to the conclusion that the estimates were reasonable by means of the verbal rivet “therefore,” but we fail to perceive any necessary connection between the two. Estimates based upon the best information may still be unreliable if the methods used to generate them from the data *33 are faulty. We are unable to determine what methods were employed In creating the period II data here and, more crucially, whether the Commission discharged its responsibility to examine the relevant methodology. 8 Cilco argues before this court that it “forecast the Villages’ demands and energy usage separately through the sound exercise of their judgment and knowledge of the Villages’ historic use characteristics,” Intervenor’s Br. at 19, but nowhere adverts to the record support for that conclusion. The record is no more illuminating. Company witness John F. Weisbruch, Jr., responding to the villages’ effort to substitute actual data, explained that the “estimates were the best available to Cilco at the time and were quite reasonable.” Record at 65. The Commission apparently credited this testimony, stating that “[mjunicipalities have not demonstrated that the Company’s method of making demand and energy projections was unreasonable.” Initial Decision at 17. This statement reveals the basic misapprehension underlying the Commission’s treatment of the instant dispute that requires this Court to vacate this aspect of its decision. It is not the customers that must shoulder the burden of proving the company’s period II estimates unreasonable. It is the petitioning company that, in the Commission’s own words, must “establish the validity and accuracy for each of [its] cost estimates.” 50 F.P.C. at 127. A petitioning utility’s bare assertions that its methods and forecasts are “reasonable” or the “best available” are not sufficient to shift the burden of persuasion onto those objecting to the new tariff. It would be a diffident utility indeed that lacked the confidence in its case to make such statements, but for the Commission to credit them without any evidence indicating what data and which methods were employed would be to abjure its statutory responsibilities to the consumers of electric power. The test year estimates of the villages’ coincident demand were not shown by the filing company to be either reasonable in result or based upon sound methodology. We remand this aspect of the case to the Commission for whatever further proceedings it thinks relevant to a decision upon the reliability vel non of these projections that is consonant with the dictates of the Federal Power Act. 9 B. Ciico’s Estimate of Purchased Power Credits The Commission was also confronted with a sizeable disparity between Ciico’s projected purchased power credits and the actual figure for period II. The immediately preceding discussion suffices to justify our inability to endorse the ALJ’s acceptance of the Cilco estimate on the grounds that there was “no evidence that the projection was unfounded when made,” or that “[t]he basis of the’ estimate is not *34 being challenged as deficient.” 10 The final ground, however, is firmer: that the estimates of purchased power credits and purchased fuel costs cancel each other out. Cilco had projected a fuel expense in period II of $53.4 million; the actual figure for calendar 1977 was $84.4 million, $28.8 million above the 1976 calendar year cost. Record at 857, 2850. The purchased power credits are charged against the overall production expenses, the most substantial portion of which is the cost of fuel. Record at 857. Therefore, evidence of record indicated that the utility’s net production expenses would not decline if actual rather than projected figures were used. We think that, in deciding whether to substitute actual for estimated numbers, the Commission retains some discretion to look beyond the challenged line item to related figures and to determine that, if actual numbers were substituted on a uniform basis, the result would not favor the party that had objected to the period II figures in the first instance. When, as here, the Commission explicitly makes the determination and can produce substantial evidence to support it, we fail to see the harm in upholding their decision to rely upon test-year estimates. 11 Accord, Indiana Municipal Electric Association v. FERC, 629 F.2d 480, 485-86 (7th Cir. 1980). C. Normalization of Taxes at the 48% Rate The Commission’s Opinion No. 81 declined to adjust the 48% rate adopted by the utility to calculate its tax expense, which it could then collect from its customers, to the actual rate of 46%. Although these tax expenses were normalized, they present the same issue as the utility’s other estimates. 12 See note 3 supra. The Commission, citing its policy against spot adjustments to estimates in light of actual data, allowed Cilco to use the 48% rate in calculating tax expense. There is no difficulty in concluding that Cilco’s estimate was reasonable when made and based upon reasonable methodology, for the utility merely assumed that the corporate income tax rate on the books when it prepared its rate increase filing would prevail in period II. Therefore, the Commission was faced with a situation in which an estimate, reasonable when made, was at variance with the actual results, not in period II, but in later years. In such a situation, the courts have given the Commission considerable latitude to balance various policy considerations, such as fairness to consumers, and the possible corresponding unfairness to regulated companies inhering in adjusting their estimates only in the customers’ favor, in determining whether to adopt or reject actual data. In both Public Service Co. v. FERC, 575 F.2d 1204 (7th Cir. 1978) and Indiana Municipal Electric Association v. FERC, 629 F.2d 480 (7th Cir. 1980), the Seventh Circuit affirmed the Commission with respect to its decision to accept or reject certain test year estimates in the light of varying actual figures. In the Public Service case, the *35 Commission decided to make the spot adjustment in the petitioning company’s purchased power costs. See Part II supra. On the other hand, when the Commission refused to make a spot adjustment in the Indiana Municipal Electric Association case, the Seventh Circuit again affirmed the Commission’s decision, explaining that “a certain degree of latitude has to be permitted since overstated estimates would almost certainly be balanced by other offsetting understatements.” 629 F.2d at 486. The cases therefore indicate that the Commission has a measure of discretion when confronted with a request for a spot adjustment, a measure that can only be greater when actual results do not vary until well after period II. We understand these cases, and the Federal Power Act itself, to place some boundary upon this discretion. The Commission is not invariably entitled to forbear from adjusting or rejecting faulty estimates when the deviation from reality within the test period is substantial and not offset by other mistaken estimates or when the utility has not explained the basis for its estimates. The Commission’s policy with respect to spot adjustments, if woodenly applied, would prove too much, because it would imply that FERC should never adjust test year estimates, no matter how faulty the reasoning or ludicrous the result. And an overly-rigid application of that-policy would, as explained in Part II supra, stand in opposition to section 205 of the Federal Power Act, which requires the petitioning utility to bear the burden of proof with respect to its claim that its test year estimates can serve as the basis for just and reasonable rates. In this case, however, when the utility’s estimate of a 48% rate was reasonable when made and the deviation from the actual post-period II figure is just over 4%, we cannot say that the Commission was without warrant to accept the period II estimate or that rates derived therefrom are for that reason other than just and reasonable. IV The ALJ found that Cilco’s reserve capacity was 20.2%, only slightly higher than the Commission’s recommended margin of 15% to 20%. Given the notorious difficulty in accurately projecting demand for electric power far enough into the future to compensate for the lengthy process of planning and constructing new generating facilities in times of economic and regulatory uncertainty, the 20.2% figure can hardly be regarded as so far off the mark as to require Cilco’s stockholders, rather than its ratepayers, to bear the cost of maintaining an adequate reserve of generating capacity. The villages, in their reply brief, concede that the actual data indicates that Cilco’s generating reserve is not excessive, and withdraw their request to adjust Cilco’s rate base, Petitioners’ Reply Br. at 36-37. For these reasons, we have no difficulty in affirming the Commission’s treatment of Cilco’s reserve generating capacity. V The Commission’s decision to credit Cil-co’s purchased power credit and income tax expense estimates have been affirmed, as has its determination that the petitioning utility did not suffer from excess generating capacity. But we have vacated and remanded FERC’s willingness to rely upon coincident peak demand estimates that are both unsubstantiated and facially inconsistent with other proffered test period data. We do not conclude that the estimates must be rejected; we remand for such further proceedings, including, at a minimum, the drafting of more detailed findings and, if the Commission deems it useful, additional submissions of evidence, that would lead a reviewing court to conclude the Commission’s decision to adopt the utility’s coincident peak demand estimates was contrary neither to the Commission’s own past statements on future test-year ratemaking nor to section 205(e) of the Federal Power Act. It is so ordered. Footnotes [1] The jurisdiction of the Federal Power Commission over wholesale electric power rates was transferred to the Federal Energy Regula *26 tory Commission by 42 U.S.C. § 7172(a)(1)(B) (Supp. II 1978). [2] For the most part, the expenses incurred by an electric utility can be divided into demand or capacity costs, which vary with the peak demand placed upon the system by its customers, and energy costs, which vary with the amount of electricity consumed over time. Thus, because utilities must build generating capacity sufficient to serve their customers at the time of greatest demand, a large portion of demand costs is attributable to plant, including such items as interest and depreciation. In contrast, the largest portion of energy costs is fuel expense, because fuel is consumed to produce *27 each kilowatt-hour the customers use, even when the system is not operating at peak. See P.. Garfield & W. Lovejoy, Public Utility Economics 152-54 (1964). [3] Our disposition of the income tax expense issue is not complicated by the fact that FERC allowed Cilco to normalize these tax expenses. Tax normalization occurs when a utility is allowed to recover from its ratepayers tax expenses that the utility has incurred in that year but whose payment it can defer, under various provisions of tax laws. “Thus, the income tax expense for ratemaking purposes is calculated without giving effect to the statutory provisions ... which the taxpayer may have elected for tax purposes.” P. Garfield & W. Lovejoy, Public Utility Economics 112 (1964). In theory, these normalized taxes are placed into an account which is depleted when the deferred taxes fall due. Id. The Commission had adopted a rule favoring tax normalization vacated by this court in Public Systems v. FERC, 606 F.2d 973 (D.C.Cir. 1979). After the remand, the Commission, while still considering the normalization issue, took the position in particular cases that it would allow utilities to normalize certain tax deferrals subject to whatever new rule it would issue. See Opinion No. 81 at 18-19, 23. The day before oral argument was heard in this case, FERC issued a new final rule requiring the use of tax normalization in various contexts. Order No. 144, No. RM 80-42 (May 6, 1981). Since petitioners raise no issue with respect to the desirability of normalization sim-pliciter, but only the rate used to calculate normalized expenses, the general issue is not before us. [4] Illinois Power Co.: Initial Decision on Proposed Rate Increase, No. -ER77-531 at 14 (June 24, 1980), rev’d in part on other grounds sub nom., Illinois Power Co.: Order Affirming in Part and Reversing in Part Initial Decision (April 10, 1981). [5] Our conclusion is greatly reinforced, if not compelled, by the decision of the Supreme Court in FPC v. United Gas Pipe Line Co., 393 U.S. 71, 89 S.Ct. 55, 21 L.Ed.2d 55 (1968), a case arising under the Natural Gas Act, whose judicial review provision, 15 U.S.C. § 717r (1976), is very similar to that of the Federal Power Act, 16 U.S.C. § 825/ (1976). In United Gas, the Supreme Court reversed the court of appeals’ disposition of an FPC order, explaining: [T]here was ‘no indication of the basis on which the Commission exercised its expert discretion,’ no articulation of ‘any rational connection between the facts found and that choice made.’ On this issue the Commission’s order was vulnerable on rehearing and in the Court of Appeals. But it does not follow that the Court of Appeals . .. should have itself determined [the disposition of certain tax savings]. These questions should have had adequate attention from the Commission in the first instance before being subjected to judicial review. Before the courts can properly review agency action, the agency must disclose the basis of its order and ‘give clear indication that it has exercised the discretion with which Congress has empowered it,’ otherwise the courts are propelled ‘into the domain which Congress has set aside exclusively for the administrative agency.’ 393 U.S. at 72-73, 89 S.Ct. at 55-56 (citations omitted). [6] It is for the same reason that we decline to affirm the Commission on the basis of a contention raised by Cilco, which argues that, because the period I figures also show coincident peaks above noncoincident peaks, the period II figures are not suspect in that particular. In-tervenor’s Br. at 25. The villages reply that some period I coincident peaks were also estimates, and similarly faulty. Petitioners’ Reply Br. at 16-17; see also Record at 802. The parties are free to raise this issue before the Commission. [7] The comparison of percentages of estimated demand growth was performed by the petitioners in their brief at 10. The figures are derived from Cilco tables of actual period I and estimated period II demand. For the entire Cilco system, period 1 peak demand was 760,183 Kw; the period II estimate was 820,250 Kw. Dividing the projected by the actual figure yields a quotient of 1.079 or a projected growth rate of 7.9%. For the villages, period I demand was 6,449 Kw, after power losses were considered, and the period II projection was 7,477 Kw. The same calculations produce a growth rate of 15.9% or about twice as large as the projected system growth rate. Percentages based upon the actual period II results can be found in the Respondent’s *32 (FERC) Brief at 21. These figures rely upon another table supplied by the villages. Record at 1016. The system average of monthly peaks was 760,183 Kw in period I and 800,192 Kw in period II. The difference, 40,009 Kw, is 5.26% of the period I base, and represents the system’s actual rate of demand growth. The villages’ monthly average coincident peak demand grew from 6,449 Kw in period I to 6,904 Kw in period II. The difference, 455 Kw, is 7.055% of 6,449 Kw. The ratio of growth rates (7.055% to 5.26%) is 1.34 or an increment of about one-third. [8] This court has stated that the Commission is obligated to issue findings sufficiently detailed to demonstrate, in the course of judicial review, that it has considered all relevant facts and has brought to bear its expert judgment upon them. Sun Oil Co. v. FPC, 445 F.2d 764, 767-68 (D.C. Cir.1971). [9] We affirm the Commission’s disposition of the villages’ claim that the Cilco estimates were further deficient in that they projected a system peak between 4 and 7 p. m. in period II, whereas for several years the system peak often did not occur during those hours. The ALJ said that the actual 1977 results demonstrated that in eight of twelve months the peak did fall between 4 and 7 p. m., and cited Record page 357. Initial Opinion at 17. Testimony at that page indicates that the Cilco witness states that the system peaked at that time during eight months of 1976, not 1977. See Record at 357. This disparity may be rooted in a typographical or transcription error; it does not seem to be dispositive, inasmuch as period II comprised six months of both years. Nor do petitioners explain the import of the timing issue, except as part of their general attack on Ciico’s coincident peak demand estimates, which we have refused to endorse on numerous other grounds. To the extent that the petitioners have attempted to cast doubt upon Ciico’s noncoinci-dent peak demand estimates by reference to differing figures introduced at a later compliance proceeding, see Petitioners’ Br. at 37-38, their argument, based on. evidence dehors the record, cannot be entertained. [10] If that were so, the villages placed the basis of the purchased power credits estimate in controversy by stating in their brief on exceptions: ... [T]he Presiding Judge rejected the proposed adjustment to CILCO’S cost of service because ‘[T]here is no evidence that the projection was unfounded when made.’ Municipalities submit the contrary is true; there is no evidence in the record to the effect that the estimate was reasonable when made. Record at 2312-13 (footnote omitted). [11] The villages failed to respond to the ALJ’s offset argument either in their brief on exceptions, see Record at 2312-13, or in their petition for rehearing, see Record at 2477. [12] We cannot see what effect normalization has on the petitioners’ arguments that the estimated tax rate should be adjusted in light of actual data. The villages complain that normalization at the erroneous estimated figure would lead to rates that did not reflect the utility’s true costs. We understand that to be their argument with respect to all other estimates of expense items at variance with actual figures. The argument against the tax rate estimate does not gain strength from the fact that the taxes will be paid later at 46%, not 48%. If the taxes were not normalized, the customers’ argument would be equally compelling, in that the utility would be paying a smaller proportion of its income in taxes than it had forecast.