STEPHEN F. WILLIAMS, Circuit Judge:

In December 1984 Union Electric Company brought its Callaway nuclear power plant into service. Under traditional ratemaking principles, the resulting inclusion of the plant in Union's rate base would have driven its wholesale rates up by about 75%. In anticipation of the resulting "rate shock," Union and its wholesale customers (the latter working for these purposes under the label Wholesale Defense Group (or "WDG")) agreed on certain features of a possible solution. In ruling on the rates Union filed to reflect the new plant, the Federal Energy Regulatory Commission disregarded these agreements. See Union Electric Company, 40 FERC ¶ 61,046 (1987) ( "Order"), modified somewhat on petitions for rehearing, Union Electric Company, 41 FERC ¶ 61,343 (1987) ( "Order on Rehearing"). Most of the wholesale customers have since settled, but two, the Missouri cities of Malden and Jackson, press the appeal; we refer to them either as the Cities or the WDG. They attack not only FERC's disregard of the Union-WDG agreements but also its unusual decision to impose an off-peak demand charge on a special class of off-peak users. Finally, Union attacks FERC's decision to adjust the allowable rate of return on equity on the basis of non-record facts of which it took official notice but to which it did not permit Union a full chance to respond.

We reverse on all of these issues and remand the case to FERC for further proceedings.

I. FERC's Treatment of the Union-WDG Agreements.

Union and WDG entered into agreements relating to three aspects of the general problem of how to alleviate the expected rate shock. As to two, the parties embodied their accord in an agreement settling a prior rate dispute. In consideration for WDG's dropping a rate challenge, Union agreed that in the looming rate case, anticipated as a result of the Callaway plant, it would propose to FERC two specific devices. Both would have accelerated certain credits and thereby have offset the impact of the plant's costs in its early years. Their accord on the third issue, deferring the effective date of the new rates, took place while the present case was pending before the administrative law judge. The Commission not only rejected all three agreed-on proposals, but gave the parties' agreements little or no weight. This was error.

The judicial requirement that the Commission give substantial weight to contracts between utilities and their customers started with United Gas Pipe Line Co. v. Mobile Gas Service Corp., 350 U.S. 332, 76 S.Ct. 373, 100 L.Ed. 373 (1956), and FPC v. Sierra Pacific Power Co., 350 U.S. 348, 76 S.Ct. 368, 100 L.Ed. 388 (1956). The parties had entered into contracts setting a precise rate (e.g., 10.7 cents per Mcf in the Mobile case), and the Supreme Court held that the Commission could override those rates only if it took on and satisfied the burden of showing that they were unjust or unreasonable. See § 5 of the Natural Gas Act, 15 U.S.C. § 717d (1982); § 206 of the Federal Power Act, 16 U.S.C. § 824e (1982). Here, of course, the parties' agreements do not set a rate; they merely seek to resolve elements of the methodology that should govern the rate decision.

But our precedents have extended the pro-contract policy of the Mobile-Sierra doctrine well beyond the circumstances of its origin, and have applied a weaker but broader variant to agreements falling short of setting a precise rate. In ANR Pipeline Co. v. FERC, 771 F.2d 507, 519 (D.C. Cir. 1985), for example, the parties had agreed on a particular cost-allocation methodology ("incremental costing" rather than average or "rolled-in"). While recognizing that such agreements were "not binding on FERC or dispositive of the question of the legality of rates," we held that the Commission was "not justified . . . in cavalierly disregarding private contracts." Id. See also Tennessee Gas Pipeline Co. v. FERC, 824 F.2d 78, 82 (D.C. Cir. 1987) (requiring the Commission to give weight to the parties' agreement on authorized levels of firm service, and thus on allocation of risk of under-use). The policy favoring contracts encompasses settlement agreements, see Cities of Bethany, Bushnell, Cairo, etc. v. FERC, 727 F.2d 1131, 1139 (D.C. Cir. 1984); Cities of Newark, New Castle Seaford, Delaware v. FERC, 763 F.2d 533, 546 (3rd Cir. 1985), including ones where the parties agree merely on a single methodological issue and leave the others for the Commission, see City of Chicago v. FPC, 385 F.2d 629, 640 (D.C. Cir. 1967).

The agreement of Union and WDG on accelerating certain offsetting benefits was stated in their March 1984 settlement of a prior case involving WDG's "price squeeze" claims. In the first of the relevant provisions, the parties agreed that Union would propose to amortize certain fuel credits, due from Westinghouse Corporation as a result of a breach of contract lawsuit, more rapidly than under traditional ratemaking procedures — over two years, rather than over the 20-year life of the contract:

The Company agrees that in its FERC rate case filing which includes Callaway I in rate base that it will propose to amortize the full Westinghouse fuel settlement over two years in a manner consistent with its recent retail rate filing in Missouri (Case No. ER84-168).

Stipulation and Agreement Between Union Electric Company and the W-3 Defense Group in Docket Nos. ER77-614, ER81-450 and ER83-646 2 (March 20, 1984) ( "Stipulation"); see Order, 40 FERC at 61,135 (identifying traditional approach).

The second relevant clause also involved accelerating and compressing the benefits of a credit. It is worded somewhat obliquely:

The Company agrees to include in its forthcoming FERC rate case filing which includes Callaway I in rates, a rate phase-in proposal using the same methodology as that proposed by the Company in its recent Missouri case (ER-84-168). The Company agrees to meet with its WDG customers regarding the relative timing and magnitude of the FERC and Missouri retail cases.

Stipulation at 2-3. Pursuant to the clause Union proposed to amortize certain Callaway deferred income taxes over three years, which Union said was the period proposed in its Missouri filing. Union Electric, 35 FERC ¶ 63,076 at 65,258 (1986) ( "Initial Decision"). Standard ratemaking principles would have spread the benefit over the life of the plant, estimated for these purposes at ten and a half years. Initial Decision, 35 FERC at 65,256.

The Commission first tried to escape the implications of these provisions altogether by engaging in some artful interpretation. It focused on the literal truth that they called on Union merely to "propose" the agreed upon approaches to the Commission. From this it inferred that their force was spent once Union made its proposal. Order, 40 FERC at 61,135; Initial Decision, 35 FERC at 65,258; FERC Br. at 13-14, 15. Though in the enforcement of the Mobile-Sierra doctrine this court defers to Commission contract interpretations if "amply supported both factually and legally," Ohio Power Co. v. FERC, 744 F.2d 162, 166 (D.C. Cir. 1984) (inside quotation marks omitted), we cannot find that standard satisfied here. When rates are at issue, it is always true that the parties propose, FERC disposes. It can set aside even contracts within the core of the Mobile-Sierra doctrine if it finds that they "adversely affect the public interest." FPC v. Sierra Pacific Power Co., 350 U.S. 348, 355, 76 S.Ct. 368, 372, 100 L.Ed. 388 (1956). Thus the two clauses seem entirely likely to have represented an agreement that the Westinghouse credits and deferred tax benefits should be accelerated as provided, subject to the possibility that the Commission might override them for adequate reasons.

Where the language of an agreement presents an ambiguity, it is appropriate for a court or the Commission to consider the available extrinsic evidence. See Ohio Power Co., 744 F.2d at 168. Here that suggests that WDG and Union meant the provisions to carry forward beyond Union's filing. First, under the Commission's narrow view WDG secured hardly anything in exchange for abandoning its prior claim. It is true that in the absence of any objection filed by a party or Commission staff, Union's "proposed" terms would likely take effect. But this scenario was implausible. A number of issues remained to be resolved in the Callaway case (including whether the construction of Callaway was imprudent and would be excluded from the rate base entirely), and WDG itself must have been preparing to object to Union's filing. We have in the past resisted a Commission interpretation that would have "completely destroy[ed] the provision being interpreted," Tennessee Gas Pipeline v. FPC, 504 F.2d 199, 202 (D.C. Cir. 1974), and we find its view here scarcely more persuasive. Of course, under some circumstances parties may bargain for a trivial benefit. The rates agreed on in the Union-WDG settlement may have fully satisfied WDG's claims, so that the promises to "propose" may have been mere icing; or WDG's claim itself may have been flimsy to start with. But the Commission has made no such argument. Second, both parties in fact advocated the substance of the proposals before the ALJ and the Commission. Union may have pursued that course simply because it thought the accelerations were advantageous. But even if so, that cuts both ways: it suggests that Union would have been happy to obligate itself to the accelerations, subject only to the Commission's inescapable authority to override.

At oral argument Commission counsel suggested that there is an industry practice of entering into agreements that merely set an approach on the Commission's plate, with no intention that it enjoy any sort of Mobile-Sierra presumption. In effect the Commission may have read the agreements as saying: "We want the Commission to consider this proposal, but it need not state any justification if it wishes to disregard it. Nor should the Commission have any burden on judicial review to justify such disregard." But the record contains no reference to such a practice, much less evidence as to how parties word such an arrangement. On remand the Commission may wish to supplement the record with such evidence. For now, we merely decide that the Commission's interpretation of the settlement agreement is unreasonable on this record. Accordingly we view the parties' proposals as entitled to full consideration by the Commission.

Apart from deprecating the contract, the Commission justified disregarding it solely by noting the interests in "minimiz[ing] departures from traditional ratemaking and accounting procedures," Order, 40 FERC at 61,134, and in "avoid[ing] intergenerational cross-subsidization," id. at 61,135. Both interests are real ones, to be sure, but their invocation rings hollow in this context. Any phase-in plan, including the one adopted by the Commission, will require deviations from traditional practices. The Commission's own plan phases the increase in over a ten-year period, taking some revenues that under traditional practice would be collected in the first three years and locating them instead in the next seven. Moreover, any plan that deviates from traditional practice, including the Commission's, will by definition produce intergenerational inequities — at least if one measures from the baseline of traditional practice. The Commission may be able to support its rejection of the parties' agreed-on proposals on the basis that its plan creates less intergenerational inequity, and that for some reason the incremental tilt resulting from the parties' plan is not justifiable despite their preference. But it has offered no such incremental analysis.

Another concern may account for the Commission's disregard of these agreements — its interest in developing uniform solutions to a recurrent problem. At an earlier stage in the proceedings the Commission had identified the present case as its "first opportunity to consider a phased-in rate increase," and had said that "it is essential that this Commission's actions on [the phase-in proposals] result in a consistent and predictable framework for dealing with them." Union Electric Co., 28 FERC ¶ 61,399 at 61,742 (1984). We fully recognize this interest, and agree that on an adequate record it may overcome private agreements. But if the Commission's decision rests on that concern, it must say so, and must articulate its reasoning in trading that interest off against the parties' entitlement to agree on ratemaking methodologies.

While the case was pending before the ALJ, Union and WDG agreed on a third issue — to delay the effective date of the wholesale rate increase to April 9, 1985, so as to make it coincide with the date of the retail rate increase. Initial Decision, 35 FERC at 65,248. As a quid pro quo for this delay, WDG agreed that Union could continue to build up its "allowance for funds used during construction" ("AFUDC"), i.e., the cost of the capital required for construction, even after Callaway was placed in commercial operation. On the effective date, the additional AFUDC would be placed in the rate base and earn a return and be subject to depreciation. Thus deferral of the increase would be paid for by slightly higher rates later. The Commission rejected the effective date agreement, observing that it would vary the Commission's standard ratemaking practice, under which, it said, AFUDC is never to be accumulated after a plant is placed in commercial operation. Order, 40 FERC at 61,131-32. See Kentucky Utilities Co. v. FERC, 760 F.2d 1321, 1329 (D.C. Cir. 1985) (sustaining Commission's standard view).

Though the agreement on this issue was not technically in the form of a contract or a settlement, it is subject to the pro-settlement policies of Mobile-Sierra. The Commission should not ignore the fact that all parties to the litigation (other than the Commission staff) agreed on the point.

Here the Commission's explanation for rejecting the parties' approach is particularly weak. It originally explained that "once the construction period ends and service begins, the `allowance for funds used during construction' must also cease." Order, 40 FERC at 61,131. But later it denied the existence of any such rule as applied to AFUDC for rate purposes. In an order reviewing Union's "compliance filing" (i.e., filing of rates pursuant to the earlier orders), the Commission allowed Union to accrue AFUDC for ratemaking purposes until January 15, 1985 — nearly a month after the commencement of commercial operation of Callaway on December 19, 1984. Union Electric Company, 47 FERC ¶ 61,427 (1989). There the Commission said that its rule against accumulation of AFUDC after a plant started service was merely "standard accounting practice," and that the Commission was "not obligated to extend the accounting ruling to rates." Id. at 62,360.

As FERC accounting rules exist primarily if not exclusively as a component of its ratemaking, this seems pure mumbo-jumbo. It also completely contradicts the case on which the Commission originally relied, Kentucky Utilities, which observed that the Commission's accounting and rate rules represented a single policy decision as to the allocation of risk between investors and ratepayers. 760 F.2d at 1328. Even if the distinction is a real one, it argues for giving effect to the Union-WDG agreement. Rates were the subject of the parties' agreement, not disembodied accounting principles.

Finally, we note that the parties' proposals and the Commission's phase-in plan are largely substitutes for each other, not complements. If the Commission on remand decides to adopt the proposed acceleration of the Westinghouse credits or the deferred tax benefits, it will likely need to remove all or part of its own special phase-in solution. Otherwise the rate structure would tilt too much of the burden onto late-period users.

II. The Off-Peak Demand Charge Imposed on Peak-Shaving Customers

A demand charge — a sum payable by a utility customer regardless of its total purchases in a year — is characteristically based on the customers' demand at the peak, and allocated among the customers in proportion to their shares of that demand. The charge usually covers all of an electric utility's fixed or "capacity" costs. See Wisconsin Michigan Power Company, 31 FPC 1445, 1453 (1964). They are assessed to the peak-period users because it is peak demand that determines how much a utility will invest in capacity. As FERC said in the very order under review,

[C]ost responsibility reflects the fact that rendering service during periods of peak usage is the most critical factor from a capacity planning perspective. In other words, those creating the critical need for power are assigned the cost responsibility for all capacity-related costs.

Order, 40 FERC at 61,141 (emphasis in original). As capacity is determined by peak demand, the capacity is in a sense free during the off-peak: "the capacity would be there whether or not the off-peak user made demands on it." Alfred E. Kahn, I The Economics of Regulation: Principles and Institutions 101 (1988). Thus the principle of matching a customer's rates with its share in the causation of costs implies that the utility should charge for off-peak service only the costs immediately referable to that service, such as the costs of energy and actual wear on the machinery. See generally id. at 87-103.

The matching of rates with costs contributes to efficient use of the service. It gives all peak-period users an incentive to consider and employ all alternatives less costly than the plant that their demand would necessitate — reducing their peak consumption (where it is less valuable than the resources needed to satisfy it), constructing alternative supply facilities, or (for some types of services) installing storage capacity. As Professor Kahn has noted, "[t]he only economic function of price is to influence behavior." Alfred E. Kahn, Applications of Economics to Utility Rate Structures, Public Utilities Fortnightly, Jan. 19, 1978, at 13, 15 (emphasis in original). Cf. James C. Bonbright, Albert L. Danielsen, and David R. Kamerschen, Principles of Public Utility Rates 95-101 (2d ed. 1988). Union's rates prior to those at issue here provided just such an incentive to its customers. Responding to it, Malden and Jackson invested in their own small electric generation facilities. The facilities enabled them to "peak-shave," thereby reducing Union's need to invest in capacity and securing (presumably) net savings for the Cities by reducing or eliminating the demand charges assessed against them.

When the Commission is persuaded that off-peak users are in part responsible for some of a utility's fixed costs, it has several remedies. Occasionally, it will include part of an electric utility's fixed costs in the energy charge, i.e., the charge for electricity actually taken, regardless of timing. See Minnesota Power Light Co., 14 Fed. Power Serv. 5-680, 5-695-6 (1978) (allocating part of the investment in two hydroelectric plants to the energy charge where the plants would be partially unavailable during peak winter months); Minnesota Power Light Co., 1 FERC ¶ 63,051 at 65,384-85 (1977) (similar). Normally, however, if the Commission believes that use outside a peak under one definition is partially responsible for a utility's fixed costs, it will simply redefine the peak. See El Paso Electric Company, 10 FERC ¶ 63,008 at 65,032-34 (1980) (calculating peak demand using 12 coincident peaks, one for each month, where the lowest monthly peak was still 71% of the maximum system peak). Users of electricity in secondary peak periods (e.g., peaks outside the summer months of air conditioning demand) are partially responsible for the fixed costs of a utility since, without these secondary peaks, the utility would probably satisfy summer peak demand with less expensive equipment that has higher energy costs. See id. at 65,033-34; cf. re Consolidated Edison Co. of New York, 8 PUR4th 475, 483 (1975) (noting the existence of a significant secondary peak, "absent which the company could use different and cheaper generation modes.").

In the present case FERC sought to impose a higher portion of the fixed costs on off-peak users by adopting a new billing form for the demand charge. It is based on the greater of (1) the highest demand established during the peak hours or (2) one-half the highest demand established during off-peak hours. Initial Decision, 35 FERC at 65,263. The second prong of this new billing form effectively creates an off-peak demand charge that applies not to all customers, but only to those customers whose highest off-peak demand exceeds twice their highest peak demand.

The Cities' quarrel is not with the Commission's assigning some fixed costs to off-peak use, although such a rate structure would be a change from the Commission's prior practice for this utility. See Reply Br. for Pet. WDG at 13 n. 7; and see Missouri Utilities, 10 FERC ¶ 61,297 at 61,600 (1980) (rejecting ALJ's ruling that allowed this utility to collect 25% of its fixed costs through an energy charge). Nor is their quarrel with the Commission's definition of peak use. In fact the Commission in this case approved a narrower definition of "peak use" than is its usual practice and than was its practice for this utility. Rather the Cities object to the Commission's imposing this particular charge solely on users whose off-peak consumption is high relative to their peak use, and to its doing so in a way that deviates sharply from its prior practice, on which the Cities relied in investing in generating facilities.

While Commission decisions in such a technical area are entitled to broad deference, see, e.g., Permian Basin Area Rate Cases, 390 U.S. 747, 767, 88 S.Ct. 1344, 1360, 20 L.Ed.2d 312 (1968), we find here both a complete absence of reasoned decisionmaking and an unexplained swerve from its prior practice. See, e.g., Atchison, Topeka Santa Fe Ry. Co. v. Wichita Bd. of Trade, 412 U.S. 800, 808, 93 S.Ct. 2367, 2375, 37 L.Ed.2d 350 (1973); Secretary of Agriculture v. United States, 347 U.S. 645, 653, 74 S.Ct. 826, 831, 98 L.Ed. 1015 (1954); Greater Boston Television Corp. v. FCC, 444 F.2d 841, 852 (D.C. Cir. 1970).

Although FERC contends that "there is no established Commission policy against permitting [demand] billing for such [off-peak] service," Brief for Commission at 25, it does not point to a single case in which it actually imposed a charge of the sort approved here. The only cases the parties have brought to our attention reject similar charges. See Kansas Power Light Co., 30 FERC ¶ 63,028 at 65,100-02 (1985) (rejecting a proposal by the Commission staff to impose an off-peak demand charge for non-interruptible, partial requirements municipal customers); Southern California Edison Co., 26 FERC ¶ 63,098 at 65,358-59 (1984) (rejecting a proposal by some municipal customers to include off-peak demand charges in the rate structure); Kentucky Utilities Co., 15 FERC ¶ 61,002 (1981) (refusing to impose demand costs on interruptible customers). So far as the parties here are concerned, the Commission has rejected a similar proposal in Missouri Utilities Co., 10 FERC at 61,600 (refusing to allow part of the demand charge to be assessed as the equivalent of an energy charge). Since that decision, the most salient change in circumstances is one of corporate form — Missouri Utilities was then a subsidiary of Union Electric but has since been merged into Union. The Commission seeks to distinguish the case in the following terms:

Missouri Utilities purchased almost all of its power from Union, its parent. Missouri Utilities' demand charge to its municipal customers was designed to track the Union rate, so that a customer which caused Missouri Utilities to incur a purchased power charge would pay a demand charge. Union charged Missouri Utilities for purchased power only if such power was purchased during [peak] months. In the instant case, the issue involves installed capacity costs rather than purchased power costs.

Order, 40 FERC at 61,140.

This appears to be no distinction at all. The Commission expressly asserts that the Missouri Utilities demand charge "was designed to track the Union rate." Thus the Commission evidently sought, before the merger, to treat the Cities essentially as if Missouri Utilities were not there. It is mysterious why removal of the corporate formality, treated as such by the Commission, should justify a serious shift in rate design. Thus, even if the Commission had occasionally ordered an off-peak demand charge for others, it would have to explain its decision to do so here. See Secretary of Agriculture v. United States, 347 U.S. at 651-53, 74 S.Ct. at 830-32 (shift in policy applied to special situations requires explanation even though agency has already applied the alternative policy elsewhere).

The Commission advances two reasons for varying its prior practice, but we find each of them insufficient. First, the Commission purported to offer a cost justification. It observed:

The characteristics of system peak load (size, duration, etc.) affect a utility's choice of generation mix (base, intermediate and peaking units). Base load units, by definition, operate nearly continuously. Consequently, capacity costs associated with base load plants are incurred during both peak and off-peak periods.

Order, 40 FERC at 61,141. But to the extent that this is true it is universally so. If a utility's use of its base load plant for supply of off-peak customers justifies an off-peak demand charge, FERC should impose one universally, or nearly so. Conceivably the passage alludes to some cost characteristic that is special to Union, but FERC has revealed none. Even such a justification, however, would not necessarily explain FERC's singling out those customers who consume more during off-peak than peak periods; if off-peak use leads to capacity costs, then the Commission should either redefine the peak or find some other way of making all users off the current peak share the burden of off-peak costs proportionally to off-peak use. In any event, the Commission has not even begun to identify special factors that call for an unusual burden on off-peak use.

To the extent that it does offer a special reason, FERC appears to lapse into gallows humor. As we noted above, FERC in the very order under review acknowledged that peak usage is critical in capacity planning and that therefore peak users are usually "assigned the cost responsibility for all capacity-related costs." Order, 40 FERC at 61,141 (emphasis in original). Inexplicably, the Commission then reversed field and concluded that it was "unreasonable for customers to avoid responsibility for capacity-related costs by [peak-shaving]." Id. Rate allocation in accordance with cost of course aims at giving accurate price signals, with the expectation that customers will respond, thereby increasing the efficiency of resource use, i.e., reducing waste of resources. Here the Commission appears to be saying that if customers respond to the incentives it has created in exactly the intended way, it will change the rules so as to deny them the resulting advantages. There may be special circumstances where a related theory would work. If peak-period prices had the effect of shifting the peak, adjustments would be necessary (perhaps redefinition of the peak or reduction in the costs allocated to the peak). See, e.g., Re Consolidated Edison, 8 PUR4th at 488 ("Another problem relating to use of peak responsibility concerns the possible shifting of the peak if a full peak responsibility method is utilized to allocate costs."). See also Kahn, I Economics of Regulation at 91. But nothing of the sort is argued here. The Commission may set the rules, and it may change them, but it may not change them simply because they enabled someone to thrive.

III. Official Notice

In deciding on a utility's permissible charges, FERC computes its "rate base," here greatly expanded by Union's addition of the Callaway plant, and then allows a reasonable rate of return on that base. The rate of return is a composite derived from the cost of the firm's debt and equity capital. Since shares of common stock do not carry a fixed interest obligation, their cost — in essence what investors demand by way of expected future income — is calculated by a process of inference. One especially pertinent point of comparison is the return on United States government fixed-income obligations, which is viewed as a "risk-free" rate of return. (Of course it is not literally risk-free. A change in the expected rate of inflation will immediately change the real value of a debt obligation framed in nominal dollars.)

Here the Commission addressed a problem created by the two-year lag between the ALJ's development of a record, which ended August 15, 1985, see Initial Decision, 35 FERC at 65,237, and the issuance of the Commission's own decision on July 20, 1987. During this period, the yield on 10-year Treasury bonds fell 4.09%, from 11.42% (averaged from December 1984 to May 1985) to 7.33% (averaged from November 1986 to April 1987). On this basis, the Commission reduced the ALJ's figure for the equity rate of return by the same amount. Later, in response to Union's argument in its petition for rehearing that the Commission could and should have taken into account two later months, making the updated Treasury rate 7.77% (rather than 7.33%), the Commission reduced the adjustment from 4.09% to 3.65%. Order on Rehearing, 41 FERC at 61,925-27. Thus, it ultimately approved a rate of 12.15%, rather than the 15.80% rate recommended by the staff and found reasonable by the ALJ.

The Commission has before wrestled with the problem of establishing an equity rate of return that reflects current capital costs while fulfilling its duty to rely on record evidence. As the Commission earlier described the problem:

We face conflicting goals: on the one hand, when the rate will be effective prospectively, we try to establish a rate of return on common equity that reflects current capital costs; on the other hand, we must base our decision on record evidence.

Orange Rockland Utilities, Inc., 45 FERC ¶ 61,252 at 61,753 (1988). Before this case, the Commission followed what it characterized as a "consistent" course in "striking a balance between these conflicting goals." Id. Though it made last-minute adjustments to the return on equity to reflect new data on interest rates, it made the adjustments only within a range of reasonable equity rates based on record evidence. Id.; see also Public Service Company of New Mexico, 17 FERC ¶ 61,123 at 61,245 (1981); Public Service Company of Indiana, Inc., 7 FERC ¶ 61,319 at 61,710 (1979). In recent cases, the Commission has updated the rate of return by exactly the change in the 10-year Treasury bond rate, up or down, but always cutting the adjustment off where necessary to keep the rate within the range found reasonable on the record. See, e.g., Orange Rockland Utilities, Inc., 44 FERC ¶ 61,253 at 61,952-53 (1988), aff'd on rehearing, 45 FERC ¶ 61,252 at 61,752-53 (1988); South Carolina Generating Company, Inc., 43 FERC ¶ 61,217 at 61,562-54 (1988), aff'd 44 FERC ¶ 61,008 at 61,038-39 (1988); New York State Electric and Gas Corp., 37 FERC ¶ 61,151 at 61,377-78 (1986).

Union argues that the Commission's refusal here to let it challenge the use of the Treasury bond data violated its rights under the Administrative Procedure Act, specifically 5 U.S.C. § 556(e) (1988), and under the due process clause of the Fourteenth Amendment. We agree that the Commission's extension of its earlier approach violated the APA. We do not reach the constitutional claim.

Section 556(e) provides:

When an agency decision rests on official notice of a material fact not appearing in the evidence in the record, a party is entitled, on timely request, to an opportunity to show the contrary.

The section must be interpreted in light of pre-APA decisions involving due process challenges to official notice, most notably Ohio Bell Telephone Co. v. Public Utilities Commission of Ohio, 301 U.S. 292, 57 S.Ct. 724, 81 L.Ed. 1093 (1937). There the Ohio Public Utilities Commission adjusted the value of the utility's property downward, for ratemaking purposes, to reflect the Great Depression, which had begun in the middle of the ratemaking. The Court found no difficulty with the commission's taking notice of the depression as such, or of the general decline in market values as "one of its concomitants." Id. at 301, 57 S.Ct. at 729. (Stagflation had not yet appeared on the scene.) But it objected to the commission's use of the data, for the general decline did not show "[h]ow great the decline has been for this industry or that, for one material or another, in this year or the next." Id. Moreover, the Ohio commission manifested a "deeper vice" by never disclosing the particular evidence on which it relied. Id. at 302, 57 S.Ct. at 729. Thus the party against which the officially noticed facts were used had no opportunity to "see the evidence or hear it and parry its effect." Id. The Court observed that "[official] notice, even when taken, has no other effect than to relieve one of the parties to a controversy of the burden of resorting to the usual forms of evidence. `It does not mean that the opponent is prevented from disputing the matter by evidence if he believes it disputable.'" Id. at 301-02, 57 S.Ct. at 729-30 (citations omitted).

We read Ohio Bell as establishing two prerequisites for use of official evidence. First, the information noticed must be appropriate for official notice. Second, the agency must follow proper procedures in using the information, disclosing it to the parties and affording them a suitable opportunity to contradict it or "parry its effect."

We have no difficulty with FERC's taking notice of a change in the rate on 10-year Treasury bonds. It is true that judicial notice is generally limited to "matters of common knowledge," id. at 301, 57 S.Ct. at 729, but official notice is broader, allowing an agency in addition to notice "technical or scientific facts that are within the agency's area of expertise." McLeod v. INS, 802 F.2d 89, 93 n. 4 (3rd Cir. 1986). The logic behind this expanded scope of official notice is that since "administrative agencies necessarily acquire special knowledges in their sphere of activity," certain highly technical facts "may become, to the administrators, as obvious and notorious `facts' as facts susceptible of judicial notice are to judges." Administrative Procedure in Government Agencies: Report of the Committee on Administrative Procedure, Appointed by the Attorney General 71 (1946) ( "Attorney General's Report"). As the FERC decisions mentioned above reflect, the Commission has long taken notice of the interest rates on Treasury bonds for ratemaking purposes. This court has previously noted this practice and approved of it in passing, and at least one circuit has expressly held it proper. See Mississippi Industries v. FERC, 808 F.2d 1525, 1568 (D.C. Cir. 1987); Boston Edison Co. v. FERC, 885 F.2d 962, 966-69 (1st Cir. 1989). As we said in Mississippi Industries, FERC's taking official notice of such interest rates is not troubling because "such information is not typically subject to dispute." 808 F.2d at 1568.

The Commission's procedures in using the Treasury interest rates for inferences on the cost of equity, however, did not adequately protect Union's right to "parry [their] effect," i.e., to challenge the Commission's inference. In Ohio Bell, the Court had no difficulty with the Ohio commission's taking notice of the Great Depression, but invalidated its use of that fact (plus undisclosed additional material) for inferences about the value of Ohio Bell's assets. Because of variations from industry to industry, the commission could not draw such inferences automatically. Ohio Bell was entitled to an opportunity to raise objections and attempt to "parry [the] effect" of the data relied on. Here we have precisely the same problem: the Commission apparently assumed a linear relationship between the trend for 10-year Treasury bond rates and that for Union's cost of equity capital. Union raised substantial objections to the official notice and was therefore entitled to an opportunity to dispute the Commission's findings. See Market Street Railway Co. v. Railroad Commission of California, 324 U.S. 548, 562, 65 S.Ct. 770, 777, 89 L.Ed. 1171 (1945) (a hearing on officially noticed evidence must be granted so long as the requesting party can make a good showing that it can contest the evidence).

Section 556(e)'s assurance that parties must have "an opportunity to show the contrary" encompasses a chance not only to dispute the facts noticed but also to "parry [their] effect," i.e., to offer evidence or analysis contesting the Commission's inferences. The Attorney General's Report on the APA confirms this reading, saying that the section entitles parties "not only to refute but, what in this situation is usually more important, to supplement, explain, and give different perspectives to the facts upon which the agency relies." Attorney General's Report at 72. See also Administrative Procedure Act: Legislative History 32 (1946) (adopting approach of the Attorney General's Report on official notice issue).

Here, as we have seen, Union was allowed a partial opportunity "to show the contrary." Union contended that two later months should be used in calculating the yield on Treasury bonds, with a resulting reduction in the adjustment from 4.09% to 3.65%. The Commission agreed and adopted Union's data. By hearing Union's complaint and responding to it reasonably, the Commission satisfied § 556(e) on this point.

The Commission did not, however, seriously address Union's broader attack on its use of the Treasury bond rates. In its request for rehearing Union submitted an affidavit asserting that there was "no generally-accepted financial theory which supports the Commission's assumption that a Company's cost of common equity capital varies linearly with the yield on ten-year U.S. Treasury bonds." Given the existence of variation among companies, as in Ohio Bell, the point is scarcely disputable; indeed the words of the Union affidavit are echoed in later decisions of the Commission itself. See South Carolina Generating Company, Inc., 44 FERC ¶ 61,008 at 61,039 (1988) ("[I]t has not been established that there is a one-for-one correlation between fluctuations in Treasury bond rates and fluctuations in the cost of utility equity capital."); see also Utah Power Light Company, 44 FERC ¶ 61,166 at 61,554-55 n. 50 (1988) (to the same effect). Moreover, Union did not rely merely on the general truism of company-to-company variation. It offered an analysis suggesting that in the same period there had been a marked increase in its stock's "beta," a commonly used measure of volatility (and therefore risk). Given Union's showing that the variation among companies was of potential importance in this case, its right to an opportunity to "parry [the] effect" of the noticed Treasury rates encompassed a right to have its proof either accepted or appropriately refuted.

Instead of responding to Union on the merits, the Commission dismissed its claim out of hand, saying that "Union should have been aware based upon past precedent that the Commission might consider updating the return on common equity under the circumstances present in order to determine just and reasonable rates to be charged on a prospective basis." Order on Rehearing, 41 FERC at 61,926. As we have seen, the precedent on which the Commission relied was much narrower, involving only adjustments within a range of reasonableness based on the record.

Curiously, the Commission argues before us that this case is not governed by its prior precedents because the ALJ here determined no zone of reasonable rates. That is true in the sense that he merely determined that 15.62% and 15.8% were reasonable on the record, not that those were the limits of reasonableness. But it should be clear that the absence of a predetermined range of reasonableness does not validate the Commission's position. Quite the opposite: before this case the Commission had reconciled its decision to use current data with its duty to rely on record evidence by limiting its adjustment to points within the record-based range of reasonableness. The lack of such a range simply extinguishes the basis of reconciliation. Nor can the Commission claim that 12.15% was within a range that it could have found reasonable on the record. The lowest rate any party even suggested was 13.5%, the rate supported by WDG's witnesses. Initial Decision, 35 FERC at 65,274.

The effect of our decision is quite limited. It does not draw in question the Commission's past practice of making posthearing adjustments within a range of reasonableness previously determined on the record. The First Circuit has recently approved such an adjustment, Boston Edison Co., 885 F.2d at 966-69 (reduction in allowed rates only one-fifth of fall in Treasury bond rates), and we do not disagree. If there should be any sign of the Commission's artificially stretching the range of reasonableness for these purposes, however, courts would obviously respond with careful monitoring.

Nor does our decision prevent the Commission from adopting a systematic updating procedure by rule, which would appear the best long run solution. Compare Generic Determination of Rate of Return on Common Equity for Public Utilities, FERC Statutes and Regulations, Regulation Preambles 1982-85 ¶ 30,644 at 31,346-47 (1985) (establishing a quarterly indexing procedure to update the advisory generic rate of return for utilities). Indeed, the Commission recently suggested that affected parties try to develop a generic solution to the problem of finding a reasonably current equity rate of return. See Allegheny Generating Company, 44 FERC ¶ 61,436 at 62,379-80 (1988). So long as the rule established a legally valid means by which to relate Treasury bond rates changes to individual company equity rates of return, its post-hearing use of changes in the former would be permissible. The principle that an agency must give an opportunity to challenge its inference from an officially noticed fact is inapplicable when the accuracy of the inference "already has been tested fairly during rulemaking, [since] the rulemaking proceeding itself provides sufficient procedural protection." Heckler v. Campbell, 461 U.S. 458, 470, 103 S.Ct. 1952, 1959, 76 L.Ed.2d 66 (1983).

Further, we do not wish to be taken as saying that the Commission may not update one set of data used in a ratemaking without updating all others. Since at least the early 1960s the Commission has allowed electric utilities to update their fuel costs much more rapidly than the other components of their rates. See Amending Regulations to Govern the Filing of Rate Schedules by Public Utilities and Licensees, 38 FPC 850 (1963), codified as amended at 18 CFR § 35.14 (1989). In response to the increased volatility in energy prices during the early 1970s, the Commission adopted a parallel system for gas pipelines. See Purchased Gas Cost Adjustment Provision in Natural Gas Pipeline Companies' FPC Gas Tariffs, 47 FPC 1049 (1972), codified as amended at 18 CFR § 154.301 et seq. (1988). For a brief history of purchased gas adjustment clauses, see Revisions to the Purchased Gas Adjustment Regulations, 52 Fed.Reg. 43,854, 43,855 (1987). Perhaps return on equity equally deserves a fast-track system. If it could be established that individual-company deviations from the trend of risk-free rates are likely to wash out over time, or are too small to be of material consequence in the time spans involved, or can otherwise be handled fairly and systematically, there would be no objection to special treatment of return on equity without special treatment of other costs.

Finally, we note WDG's argument that, although in its view the Commission was correct in reducing Union's rate of return prospectively, it should also have reduced the rate retroactively, i.e., for the period from the addition of the Callaway plant to the rate base to the Commission's final order. The Commission rejected this argument on the ground that the adjustments would entail too great administrative costs; the Treasury bond rate was of course continually changing over the period. We think the selection of a starting date for the application of any updated equity rate of return was in the absence of irrationality a detail entirely up to the Commission.

* * *

Accordingly, we reverse and remand the case to the Commission.