PRETTYMAN, Associate Justice.

This is a rate case and is before us on a petition to review and set aside an order of the Federal Power Commission. Petitioner is a natural gas pipeline company.

Petitioner's first point relates to the 6% rate of return found by the Commission to be reasonable. It claims that this finding does not accord with the precepts of fair play, because, it says, the whole hearing procedure was upon an assumed 6½% rate of return, and a 6% rate was first mentioned in the principal brief of Commission counsel before the Commission. It further says that in eleven prior natural gas cases since the Natural Gas Act was passed, 6½% was allowed, and that the general financial picture as to utilities has not changed since those cases. It further says that the finding as to the rate of return is not based upon substantial evidence and that the Commission did not consider the evidence of petitioner on the point.

The order of investigation which inaugurated the proceeding, and likewise the order setting the hearing, recited that the inquiry would concern petitioner's rates and charges. This was sufficient notice that the rate of return would be considered. At the hearing, both the Commission staff and the company introduced evidence upon the matter. That produced by the Commission staff included voluminous economic and statistical data. That evidence showed that the price of long-term money generally, and similarly such costs to utilities, including natural gas companies, had declined in the period preceding the test year 1943 used in the case at bar. The earnings-price ratios of common stocks of natural gas companies held by the public were, so far as this evidence showed, in some cases up and in some cases down between 1937 and 1943, and no general pattern in that respect is discernible. Those ratios varied in 1943 from 7.29% to 29.71%, and the trend between 1937 and 1943 varied, among companies, from a decline of four points to an increase of eighteen points.

We have examined the eleven cases to which petitioner refers. Four of them were consent orders. Two companies had common stock only. One had $8,000,000 of 5½% debentures outstanding against a rate base of $48,000,000, the balance being represented by common stock. Another had about half its rate base represented by long-term debt of which the cost was 2.88%, and a little less than a fourth represented by preferred stock at 5.86%. In another, the Commission based its 6½% allowance upon a theoretical capitalization of 40% bonds at 3½%, 20% preferred stock at 5¾%, and 40% common stock at 8%. All of those cases were decided in 1943 or earlier and rested upon data antedating that year. The great differences between the financial circumstances in those cases and in this create a wide difference between the overall rate of return allowable in so far as the court is concerned.

Under the rule laid down by the Supreme Court in the Hope Natural Gas Company case, the court is restricted in its review of a Commission rate of return allowance to a test of the end result of the order and, of course, the adequacy of the findings and the sufficiency of the evidence supporting the findings. About half of the capital of this petitioner is represented by 2½% long-term notes and the other half by equity capital. From the standpoint of the cost-of-capital rule, the 6% rate of return allowed would meet the obligation of the 2½% notes and allow about 9½% on the common stock and surplus. The record does not furnish any other statistical test of the end result of the allowance on the equity capital. The average yield of electric utilities on common stock for 1943 was found to be 7.3%, and the evidence shows that natural gas companies are regarded by the public as less desirable and therefore require higher yields. But petitioner does not point to any evidence of the extent of the margin between the two industries in common stock yield requirements. Petitioner asserts certain risks in its business but gives us no statistical measure of those risks by which to test the conclusion of the Commission.

Upon this evidence we cannot say that the rate of return allowed by the Commission was beyond the limit of its power, either as unreasonable, insufficient, or unsupported by substantial evidence.

Petitioner's next point relates to the determination of certain costs of the company's regulable business. Its business consists in part of the sale of natural ags to public utilities for resale, and in part of sales to industrial consumers. The former part is subject to regulation by the Federal Power Commission; the latter is not. In order to determine fair and reasonable rates for those sales which are under its jurisdiction, the Commission must of course, determine the costs involved in those sales. This necessitates an allocation of costs as between those sales which are subject to this regulation and those which are not.

The regulated sales in this case, being the sales to utility companies for resale, are easily identified. The problem is to ascertain the costs incurred prerequisite to such sales, and so to be borne by those customers. This is a question of fact. Expenses (using that term in its broad sense to include not only operating expenses but depreciation and taxes) are facts. They are to be ascertained, not created, by the regulatory authorities. If properly incurred, they must be allowed as part of the composition of the rates. Otherwise, the so-called allowance of a return upon the investment, being an amount over and above expenses, would be a farce. Costs incurred for specific sales are easily assigned to them. But since many supplies are purchased, salaries and wages paid, expenses incurred, and facilities used to serve all customers, it is necessary to apportion such costs in order to ascertain the costs applicable to certain customers. A number of methods are available. One is the demand-commodity method.

There is nothing new or novel about the demand-commodity formula. It has long been used, by both utilities and regulatory authorities, in the composition of rate structures. Customers desire different types of service. If the costs necessitated by the several services differ, different rates are justified, if not required. Functional analyses of costs are therefore made. The cost of each class of service is considered to be the composite of the costs of its functional elements. The basis of the demand-commodity formula is the difference between costs which occur by reason of required plant and equipment capacity and costs which occur directly in the handling of the gas. The company must have the capacity to supply certain demands when made. That capacity must be available whether or not it is being used at any particular moment. Thus, such costs do not vary from time to time but, generally speaking, continue constant, or substantially so. They are demand, or capacity, or fixed costs. Other costs are incurred only when, as and if gas is being made, transported or sold. They relate to the commodity itself. They are commodity, or volumetric, or variable costs. They obviously vary with the sales.

There are three steps in the employment of the demand-commodity method of finding the costs necessitated by the type of service afforded individual users. The first step is the ascertaining of the individual dollar amounts of the various items of cost, i.e., depreciation, taxes, cost of gas, engineering, etc. This is rarely controversial, since it is a routine accounting operation. Second, it must be determined for each item of cost whether by its nature it is a demand cost or a commodity cost, or if not classifiable wholly in either of these categories, the proportions thereof to be assigned as demand and as commodity. The third step is the apportionment of total demand cost and of total commodity cost to each customer or class of customers — in the instant case, to customers comprising petitioner's regulable business and to those constituting the non-regulable business.

The apportionment of commodity cost among the customers is relatively simple; the proportion of this cost which must be borne by any customer or class of customers is that which the customer's use of gas is of the total use of gas during the period for which the costs have been determined. The commodity cost chargeable to a customer bears the same ratio to the total commodity cost for the period as the amount of gas consumed by him bears to the total quantity of gas sold by the utility to the aggregate of all of its customers during that same period.

Apportionment of demand cost among customers is a more elusive problem. The end result must be such that each user will shoulder his share of the cost of providing and maintaining plant capacity. Each customer should pay for the capacity necessitated by his service demands. Various methods of apportionment are available, and these vary considerably in complexity and in the merits of results achieved. One which is not prohibitively complicated and which yields equitable results in many situations has been approved by the Supreme Court in the Colorado Interstate case. The initial step in its employment is the ascertainment of that day within the test period on which the gas used over the entire system was at its maximum or peak. Then the total demand costs are assigned to individual users accordingly as the total gas sold on the peak day was taken by each. Or, stated differently, the demand cost assignable to each m.c.f. (thousand cubic feet) of gas sold on the peak day may be found by dividing demand cost by gas sold; then the demand cost to be borne by any customer or class of customers will be the product of m.c.f. of gas used by him, or them, on the peak day and the demand cost carried by each m.c.f.

The Commission has wide power in the selection of formulae for the ascertainment of costs. As the Supreme Court has twice remarked "it is much easier to reject formulas presented as being misleading than to find one apparently adequate", and the commodity-demand formula has been approved by that Court. It is not the function of the courts to select formulae in these matters.

It is also true, and the Supreme Court has held, that judgment and discretion must control in the allocation of costs, because the matter is not an exact science. The discretion which must be exercised is that of the Commission. Congress has confided that function to it. At the same time, Congress has forbidden arbitrary action and has imposed upon the courts a duty of review in that respect. Arbitrary action, if it means anything, means action not based on facts or reason. The discretion and judgment confided in the Commission must be exercised upon facts and for reason. The duty of review imposed upon the courts requires that the facts be found and the reasons stated. Otherwise, the courts cannot determine whether a given action is or is not arbitrary.

The Congressional provisions extend to complicated, difficult matters as well as to simple questions. The courts cannot evade their responsibility merely because the subject matter is obscure. And neither can they be required to probe the minds of the agency for unfound facts or unexpressed reasons. The coordination of the two functions of administrative discretion and judicial review requires that the facts upon which the discretion is exercised, and the reasons, be clearly and completely stated. When the matter is complicated, the necessity is greater.

Petitioner's first point in regard to allocation concerns the Commission's computation of the percentage of demand costs applicable to regulable sales. It has been noted above that petitioner's business consisted of direct sales to industrial customers and of sales to other utilities for resale to ultimate consumers, and that only the latter were within the jurisdiction of the Commission. Among the industrial customers were some who contracted for gas on a firm basis and who had a right and priority to receive their gas whenever they wanted it. The contracts of others provided that their supplies of gas would be interrupted or cut off whenever it was necessary to do so because of the demands of the firm customers. These interruptible customers, then, were able to receive gas only at such times as there existed an excess of capacity beyond that required to meet firm customers' demands.

The Commission made several computations. The winter season 1943-44 was the period of time from which the data was drawn. The Commission found the percentage of regulable sales to total sales on the following days to be:

System peak day — February 14, 1944 44.98% Firm gas peak day — January 7, 1944 50.79% Firm gas peak day — December 16, 1943 54.55% Firm gas peak period — December 14-17, 1943 53.81% ______ Average ................................ 51.03%

Then the Commission made a computation based on the average of the four days in the peak period December 14-17, 1943. The last computation in the above table was on the total of the four days. The daily deliveries to firm customers during these four days averaged 123,840 m.c.f. The Commission used as the capacity of the line 135,000 m.c.f., and then computed the regulable sales (68,485 m.c.f.) on that average day to be 50.73 per cent of the system capacity.

Based on all the foregoing data, the Commission determined that 51 per cent of demand costs were properly assignable to regulable sales.

Petitioner makes several contentions in regard to the foregoing determination. (A) It says that it was error for the Commission to give any consideration whatever to the coincidental, or system, peak day. It says that the use of such day does not test to the utmost the interruptibles' liability to curtailment, and that neither does it indicate accurately the relative long-time demands of different firm customers. The company has much in its favor in this contention. The use of "peak responsibility" for the allocation of demand costs has been severely critiziced and, we are told by one authority, generally abandoned. It fails to give effect to load factors, diversity factors and other mathematical expressions of utility characteristics which are fundamental in a precise distribution of costs. But, as we have already said, it is not for the courts to select formulae in these matters, and, moreover, the use of the system peak day (provided the figures are accurate) was approved in the Colorado Interstate case, supra. Upon that authority the Commission might have used the percentage shown for that day (44.98%), which would have been greatly to petitioner's disadvantage. It did not do so. And the ultimate 51 per cent used by the Commission was otherwise indicated as reasonable by the average day of the peak period. The use of the system peak day for the distribution of demand costs may well be a proper subject for further examination by the Commission and the courts in a case in which the facts indicate inaccuracy in the result thus reached.

(B) Petitioner says that it was error for the Commission to use January 7, 1944, as a firm gas peak day. It says that December 16, 1943, was the correct day. But the elimination of that data on January 7, 1944, does not materially affect the result of the Commission's computation, because it happens that the percentage shown for January 7th (50.79%) is about the average (51.03%), and so its elimination does not materially change the average. Moreover, there is some dispute in the evidence as to which was the firm peak day, the company witnesses first using January 7th and then changing to December 16th. The Commission, it will be noted, used both days in its calculation, although, as we have said, its use of January 7th had no effect on the result.

(C) Petitioner says that the Commission erred in using 135,000 m.c.f. per day as the line capacity in the last calculation made by it, as above stated. But there was evidence to support the figure. At least two witnesses testified to it, and an exhibit showed that on the peak days of March 3, 1943, and February 14, 1944, the 135,000 figure was actually exceeded in the days' deliveries.

(D) Petitioner says that the Commission erred in that it averaged and gas delivered to firm industrials and to resale customers for the four days December 14-17, 1943, but did not average the total deliveries. It did not compute the relationship between sales for resale and total deliveries. Instead it computed the relationship between sales for resale and line capacity. Thus petitioner says that the denominator used to compute the percentages was erroneous. It seems to us that an allocation of capacity costs to interruptible customers is reasonable and can reasonably be the excess of capacity over the actual firm peak. It need not necessarily be only the amount of actual deliveries to interruptibles upon the day or in the period when firm deliveries are at the peak. The excess capacity above actual peak use by firm customers was available for interruptible customers, although it was likewise available for unexercised demand. The method does not seem to be the most scientific possible, and there is an inconsistency between this use of capacity and the allocation among firm customers on actual gas taken in a peak time. But, even so, this assignment to interruptibles does not seem to us to be beyond the limits of reasonableness. The Commission did not confine itself to it. There is a difference of opinion upon the point. The staff witness French in the case at bar supported the ratio of relative use rather than the relationship to capacity. But our inquiry is ended if we find that the method used by the Commission is clearly stated, clearly supported by evidence, and reasonable. Even though other computations be also proper, or even better, we cannot for that reason reject the method used by the Commission. In the present case the Commission's result was supported by the other computations in which actual use, rather than capacity, was the denominator.

(E) Petitioner argues against any allocation of demand costs to interruptible customers. But it seems to us proper that they be required to bear the costs of such part of the capacity as is available for them over and above that required to meet actual peak requirements of the firm customers; in other words, that interruptibles should bear such part of capacity costs as are incurred for them. The company (by original inclusion or by subsequent addition) may have provided in its facilities some capacity for the purpose of serving interruptible customers. And, even if the capacity merely meets the maximum potential demand of firm customers, that potential demand may not in practice actually materialize; in other words, the firm customers may not actually use the full capacity required to meet their potential rights to demand, and thus interruptible customers may always be using some plant capacity. That being so, such customers ought in fairness to bear some capacity costs. There was evidence in this record that the company had actually installed some capacity to care for interruptible customers.

In its petition for review in this court, the company says that the percentage of demand costs applicable to regulated sales should be 53 per cent instead of 51 per cent. The narrow margin of the alleged error is an indication of the close approach to accuracy achieved by the Commission even in petitioner's view.

The allocation of demand costs among the several customers is at best an indefinite and unsatisfactory process. No completely scientific method or formula has been devised so far as we are informed. The Commission's findings and conclusions upon the subject are complete and clear. The figures used are supported by ample evidence. The final figure (51%) was indicated by both the average of system peak day, firm peak day and firm peak period, and the average day of the firm peak period measured against capacity. We think it was within the limits of the proper exercise of judgment by the Commission.

Petitioner's next contentions relate to the Commission's classifications of various costs as demand costs or as commodity costs. Before examining these points in the present record, it is necessary that we note what was done in these respects in the Canadian River case, to which reference has already been made.

The decision of the Supreme Court in that case is referred to in the Commission's opinion in the present case as authority for the method here used in allocation of costs. In that case the Commission adopted certain exhibits presented by its staff concerning cost allocation, which it said followed principles long recognized and widely accepted. In its opinion it said:

"Therein costs are divided essentially into two groups, fixed and variable. Fixed costs are largely joint costs which do not vary with volume of sales. The total amount of such costs is largely proportional to the maximum demand on the system or system capacity. Accordingly, these costs have been allocated basically in proportion to each customer's responsibility for the peak day demand. Variable costs are largely those that vary proportional to output or volume of sale. Accordingly, these costs have been allocated in proportion to volume of gas purchased by each customer."

The Canadian River Gas Company was a production, transmission and distribution company. The petitioner in the case at bar is a transmission company, except for insignificant distribution costs, so that only the transmission data in the prior case is pertinent here. The Commission staff in that case made the following allocations of present interest:

Volumetric Capacity Costs Costs Operation of Compressing System (Labor and Supervision) $ 51,173.27 $ 51,173.27 Depreciation 290,036.61 — of Compressing Station Equipment 48,953.15 Ad Valorem Taxes 193,777.60^fn14 Corporate Taxes 8,103.68 Labor Taxes 11,835.73 Return on Transmission Plant 331,327.14 331,327.14

In his testimony the staff witness explained the basis for the items. He based his allocation of compressor station labor upon his judgment and the fact that "a station that is shut down but available for use carries a skeleton crew which is generally about half of the crew that is necessary to operate the station's full load." His division of the return, one-half to volumetric and one-half to capacity, was upon the basis that one-half of the return constituted the risk element in transmission, and that the risk element should be borne in proportion to the volume handled.

As we have indicated, the Supreme Court approved the action of the Commission in that case.

In the case at bar the staff of the Commission presented a study of the allocation of costs, supported by voluminous data and a statement that careful consideration had been given recognized methods of cost determination, customs of the industry, and special engineering features of the particular property, all to the end that the allocations not result in a distortion of costs either to the benefit of or to the discrimination against any group of customers. That study and its results bore the approval of the Chief of the Division of Rates and Research, and the Chief of the Bureau of Accounts, Finance and Rates, of the Commission. It included two methods, one a demand-commodity allocation and the other a straight commodity basis. In the former, the study for 1943 allocated to demand $2,450,744.23 and to commodity $4,704,729.44 of costs, which included expenses, depreciation, taxes and return on the rate base. The study was presented by the Commission's senior rate investigator on the witness stand, and the recommendations on controversial items were probed on cross examination. He testified that the method used was that used by the Commission in other cases and approved by the Supreme Court in several decisions.

On the items now in controversy the staff made the following allocations:

Demand Commodity

Transmission Supervision Engineering $ 44,528.83 Compressor Station Labor 297,738.10 Depreciation 731,863.34 $ 71,336.33 Taxes other than Federal Income 291,271.71 36,450.88 Return on Rate Base (at 6½%) 445,660.00 445,660.00

The company presented two allocation studies. One of them agreed with the Commission staff on compressor station labor, while the other assigned 25 per cent of that item to commodity because of the casual labor involved. The company studies assigned the great majority of depreciation and taxes other than federal income tax to demand, thus agreeing with the Commission staff, but they made direct charges to customers of the balances, instead of charging them to commodity.

The Commission rejected all studies on these controverted items, except as to the return on the rate base. It allocated as follows:

Demand Commodity

Transmission Supervision Engineering $ 11,702.00 $ 32,827.00 Compressor Station Labor 297,738.00 Depreciation 401,600.00 401,600.00 Taxes other than Federal Income 158,730.00 158,730.00 Return on Rate Base (at 6%) 411,378.00 411,379.00

Before considering the several controverted items, we note the action of the Commission in respect to the demand-commodity formula itself. The formula and its factors were recited. Then the Commission referred to the fact that certain costs — depreciation, taxes, and return — are "principally" dependant upon capacity. But it concluded that to assign these costs exclusively to demand "because they are proportional to plant and do not vary with the annual volume of gas sales" would be "inequitable' and "manifestly error". Therefore, the Commission concluded "From the evidence and the exercise of informed judgment, we find that the following classification of cost of service is reasonable and proper:". In sum, the Commission did not allocate costs according to the demand-commodity formula as heretofore understood, but according to its own definitions and its informed judgment.

It seems evident that formulae have no validity without definition of the factors. A formula is a combination of defined factors which yields a stated result. If the factors comprising the formula be not used, the formula is not used.

The Commission was not bound to the established demand-commodity formula, even though an undeviating pursuit of an established path would lend itself to facility of understanding. It chose in this case to depart from the established concepts; that is, from the established definitions of the factors which comprise the formula. So it cannot say that its allocation is proper merely because the named formula is proper. When the Commission purports to act upon "the exercise of informed judgment," apart from established criteria, it must make clear and complete the findings and the reasons which lead to its conclusion. These costs, as we have said, are facts; they are to be ascertained, not merely allowed by grace. The critical consideration in the present case is whether the classifications the Commission made "From the evidence and the exercise of informed judgment" were within the permitted bounds of its discretion. The court cannot determine that question unless the findings and the reasons upon which the classifications were actually made are clear enough and complete enough to be understood and judged.

In the case at bar, the Commission defined "demand costs" as costs which "vary with the demands made on the pipeline's capacity". This is a clear departure from recognized definition and is most ambiguous. The formula theretofore established by the Commission and approved by the courts had, as we have seen, distinguished between fixed, or capacity, costs, predominantly proportional to the size of the plant and largely independent of the annual Volume of gas sold, and commodity costs, which vary with the annual volume of gas. The Commission now says that demand costs vary with the demand. We are at a loss to understand what is meant. If the statement means that demand costs as initially incurred are in amounts fixed by demand requirements, such costs are fixed and do not "vary"; that was the former, and approved, definition. If the statement means that this class of costs varies from year to year or month to month, or among customers, according to the demand made in each period or by each customer, it is not accurate; and, moreover, we fail to perceive any essential difference between that and a commodity cost. Demand made is for gas to be delivered, and that certainly is the volume of gas delivered. The demand which fixes the minimum to plant capacity is the right to demand, adjusted by a diversity factor. This is the meaning of "demand" in the recognized demand-commodity formula. It does not vary with demand actually made.

The meaning of the new definition of "demand cost" is not sufficiently clear for us to test its validity. Where, as here, the Commission rejects all formulae used in the testimony and those heretofore followed by it and the courts, the nature of the new proposal must be clear enough for the courts to exercise the function of review imposed on them by the statute.

In the course of its discussion of the formula, the Commission says, "Mississippi now contends that all taxes, depreciation and return should be classified as `demand' charges," etc. The statement may be literally correct, but it is hardly objective. The Commission was repudiating the recommendations of its own staff. An objective statement would have said so, and not have cast the repudiation in the form of a mere denial of a company claim.

Petitioner does not challenge the propriety of the use of a demand-commodity formula. Its contention is that in classifying costs as between "demand" and "commodity", the Commission completely ignored the facts and so manipulated the costs as to reach an erroneous determination of the costs of the sales subject to its regulation.

The importance of the classification is apparent. The Commission found, as we have seen, that about 51 per cent of the sales during peak periods were to utilities for resale; i.e., were regulable sales. And it found that about 32 per cent of all sales were to those utilities. It, therefore, found it reasonable to allocate 51 per cent of the demand costs to regulated sales and 32 per cent of commodity costs to those sales. So that if a particular cost were classified as a demand cost, 51 per cent of its amount would be included as a cost in the computation of these regulated rates; whereas, if it were classified as a commodity cost, 32 per cent of it would be included. An error in the classification would result in a material distortion of the costs of regulated sales. This is what the petitioner says occurred. The court is limited to the inquiry whether the action of the Commission was arbitrary or was based upon facts and reason. Since the classification of items of cost was the direct control over the amount included in the costs of regulated sales, the Commission could not be arbitrary in that classification, any more than it could be in a simple refusal to include a proper cost.

The principal item to which petitioner objects is the allocation of the cost of compressor station labor in the amount of $297,738. The Commission classified it as a commodity cost and, therefore, allowed 32 per cent of it, or $94,680.68, as a cost of regulated business. If the item had been classified as a demand cost, 51 per cent, or $151,846.38, would have been allowed as a cost of regulated business.

We first examine the evidence in support of the Commission's findings, guided in our examination by the Commission's brief. It directs our attention to the testimony of three witnesses, Shuttleworth, Comfort and Simonds. The first-named was assistant treasurer of the company in charge of accounting. The testimony cited to us related to an exhibit showing a computation of net income for a hypothetical test year. It showed adjusted figures for 1943 and then indicated increases or decreases by items to produce a schedule for a so-called normal or probable postwar year. This test year was supposed to represent the operations of the company after the increases in volume of sales which had come from war business had been eliminated. A decrease in gas revenues was indicated. In respect to compressor station labor in this hypothetical test year, it was stated, both on the exhibit and in the testimony, that it was estimated that there would be a reduction in labor cost in approximately direct proportion to the reduction in sales volume. The statement was not addressed to the problem of allocation but purported to relate to a variation in the total year's expenses. No factual data in support of the bare conclusion was submitted.

The testimony of the witness Simonds which is cited to us by the Commission was directed to a computation of the costs of furnishing gas under a special contract to a customer known as the Crossett Lumber Company. Simonds included in the cost of this gas labor costs in the proportion of Crossett gas to total gas. He commented, "To a degree, the treatment of the $1904 [labor] is not consistent with the general treatment of Compressor Station Labor", but "the amount involved was, I thought, negligible, and rather than recast it, it was left that way." Pressed on the point in cross examination, he repeated that the treatment was inconsistent but that the amount involved was negligible. Thus, that part of Simonds' testimony upon which the Commission relies on this point, is not his testimony as to allocation of costs, but was on another matter stated by him to be a deviation from the principles of allocation because of its negligible amount.

The Crossett transaction was described by the witness French, of the Commission staff, as nothing more than a transportation arrangement by which the company bought earmarked gas for resale to the Crossett Company. He made a separate treatment of it, different from his treatment of all other business. The Commission, in a footnote to its opinion, treated the item the same way.

The testimony of the witness Comfort cited to us in the Commission's brief related to an estimate of the costs of operating certain new equipment then being installed. He said that the same kind of engine was already in operation in another station and that he had taken the cost of operating that station "and modified that expense by the different amount of labor that would be required at Perryville [the new equipment] — they wouldn't require quite as much labor or supervision as is given at Crossett [the old equipment] — and also modified by the amount of time that it will operate at Perryville as compared to the operation at Crossett." We fail to find in this testimony, which is the only citation to Comfort's testimony given us by the Commission, any reference to cost of labor being proportional to sales.

We are told by the Commission that

"The fact that compressor station labor is proportional to the volume of gas compressed is shown by the following data (App. 974, 978):

Compressor station Cost per Year MCF sales labor Mcf

1940 .............. 36,532,523 $181,753 0.50c 1941 .............. 42,321,165 234,872 .55c 1942 .............. 44,012,110 263,629 .60c 1943 .............. 45,190,533 297,738 .66c"

But the table indicates to us the opposite of what the Commission says it shows. The percentage of variation of m.c.f. sales and of compressor station labor as shown by the quoted table are as follows:

Percentage of Percentage of increase in sales increase in labor Year over prior year over prior year

1941 ........... 15.8 29.2 1942 ........... 4.0 12.2 1943 ........... 2.7 12.9

The variations are obviously wholly different, instead of being the same. The labor does not vary in proportion to the gas sold, so far as the quoted table shows. Moreover, it would seem, in the absence of explanation, that if the labor costs were proportional to sales, the cost per m.c.f. would be the same. On the contrary, the table shows an increase in the cost per unit (m.c.f.) sold. The Commission says that consideration of wage increases and overtime shows that the average cost per m.c.f. is almost exactly proportional to the gas sales. No such computation appears either in the record or in the brief. Only the dollar amounts of average monthly earnings and of wages per day appear in the record, and we are supplied with no factor by which those dollar amounts can be reduced to a common denominator for purposes of comparison.

In its brief the Commission tells us that "the labor cost of operating the engines is proportional to the volume of gas pumped because the operators work when the engines are operating." The generality of the stated premise seems to us no support for the conclusion. Of course, the operators work when the engines are operating. But even assuming that the statement means that the operators do not work when the engines are not operating, and that their pay is in proportion to their "work" in that sense, we are given no reference to testimony in the voluminous record which would support that statement. On the contrary, we find testimony that "regular" employees, in contract to casual labor, were 55.62% of the total field employees in 1943, which would seem to indicate that more than half the field employees were on some sort of regular pay roll. We also find the direct testimony of one witness who said that "I know that when an engine is shut down they do not lay off the man", explaining that some leeway was given for casual labor. It seems to us that reasonable assumption would be to the same effect. This company had firm commitments to furnish large quantities of gas upon demand. Obviously the compressor stations must be in readiness to meet those obligations. Such readiness must require a certain amount of labor on the job. All labor pay, all maintenance, and all consumption of supplies could not possibly cease just because no gas was actually being pumped for a period, so long as such large firm obligations were outstanding.

In its order the Commission finds, "Mississippi's President Chief Accountant and other allocation witnesses all testified that the Compressor Station costs shown above are approximately proportional to the volume of gas compressed." The statement is hardly candid or objective. Mississippi's president and chief accountant were not "allocation witnesses"; they did not testify on the subject of allocation. There were three "allocation witnesses", i.e., witnesses who had made special studies of the allocation of costs and who testified directly on the matter. They were the witnesses French, Simonds and Miller.

The witness French was the member of the staff of the Commission and its senior rate investigator in the instant case to whom we have already referred. As we have seen, in his study of the allocation of costs of service, he assigned the entire cost of compressor station labor to "demand". The Commission rejected in toto the testimony of its own expert on this subject.

The witness Simonds was a consultant presented by the company, who had also made a study of allocation of costs. He allocated 75 per cent of compressor station labor to "demand" and 25 per cent to "commodity". He said that compressor station labor has frequently been considered chargeable wholly to "demand", because it continues regardless of the volume of gas handled, but that under the conditions of this petitioner's operations, referring to the amount of casual labor employed, he had allocated 25 per cent of the item to "commodity". The witness Miller was also presented by the company. He had made a study of the allocation of costs, and used without adjustment the allocation of compressor station labor made by the Commission witness French. This, as we have noted, was an allocation of the whole of the item to "demand".

The question, therefore, is whether the one wholly indirect reference of the witness Shuttleworth can be deemed to be substantial evidence, within the meaning of the rule in support of a flat and unequivocal finding by the Commission that compressor station labor is directly proportional to the volume of gas sold and is, therefore, a "commodity" charge for allocation purposes under the demand-commodity formula. We do not think that it is. He was not testifying as to allocation of costs; three other witnesses testified directly as to proper allocation of costs, and two of them assigned all of this item to "demand" and the other assigned 75 per cent of it to "demand". The witness presented by the Commission staff itself was one of the former two.

It seems to us that the fact that this labor cost varies in proportion to the volume of gas sold, or the fact that it is not so proportionate, is easily susceptible to precise demonstration. Complete data must be available to both the Commission and the company. An important item of fact cannot be sustained by wholly unsatisfactory generalities or by a twisted application of one bit of evidence not directed to the point, in the midst of a contrary concurrence of testimony by those who testified directly upon the subject. The Commission has wide discretion in the solution of problems such as this, but the solution must rest upon the facts. We repeat that the same rules as to substantial evidence and necessary findings apply to technical facts, difficult of ascertainment, as apply to ordinary facts easily found and easily understood. The facts must be found, the findings must be supported by substantial evidence, and the conclusion must in turn be supported by the findings. The allocation of this compressor station labor is not so supported.

Petitioner next contends that the Commission's allocation of the annual charge for depreciation was arbitrary, contrary to the evidence, and unsupported by substantial evidence. The Commission assigned one-half of this annual charge to demand and the other half to commodity. In support of this treatment it said:

"The record does not support the Company's claim that depreciation expense is not related in any manner to the volume of gas sales, and that this item should be considered a `demand' charge. Mississippi is engaged in a wasting-asset business and it has stipulated that the depreciation rate reflects consideration of the associated natural gas reserves. In this case, functional rather than physical depreciation is controlling because the overall quantity and output of the natural gas resources fixes the service life of the pipeline. It is evident that the volume of gas used by interruptible customers does contribute to functional depreciation. Here again it would be improper to allocate depreciation expense exclusively to the `commodity' or `demand' classification, and we concluded that it should be assigned equally to those categories."

It was stipulated before the Commission that the annual depreciation charge was on a straight-line basis and that the rate reflected the average service life of the plant "as measured by the gas reserves of the fields from which it now secures its natural gas supplies."

Annual depreciation charges on a straight-line basis are fixed by two factors, the cost of the depreciable property (less salvage) and its service life in years. The charge is merely the mathematical quotient of the former divided by the latter. It does not vary with the quantity of gas sold or transmitted. The service life of the plant may be fixed by the life of the gas reserves, and thus the total volume of gas to be transmitted during the entire venture, divided by the estimated annual rate of transmission, determines the number of years over which the cost of plant is to be spread. But the total depreciation to be taken is the total cost of plant, less salvage; it is a constant; it does not vary with the annual volume of gas sold. When the years of life have been estimated, the annual charge becomes a constant. It does not vary with the volume of gas sold annually. It is not a commodity cost, as that term is defined by the Commission. The only relationship between volume of gas and straight-line depreciation is the wholly indirect one that the volume of gas measured in years of supply fixes a maximum life for the plant. Moreover, although there are many divergent views as to the nature of depreciation, from the standpoint of the expense accounts the annual depreciation provision from income is universally regarded as, and in fact undoubtedly is, a return to the investor of the cost of plant. Thus viewed, it is not only related to cost of capacity, but in sum total it is the cost of capacity; so it would seem to be the purest of demand charges.

In the Canadian River Gas Company case, supra, the Commission treated depreciation on various equipment in different ways. Transmission depreciation, except on compressor station equipment, was treated as a capacity, or demand, cost. In that case the Commission adopted the allocation studies of its staff.

In the present case the Commission has rejected the depreciation allocation study of its staff. We are not supplied with any clear idea as to why or how the Commission, rejecting the expert evidence before it, simply divided the whole depreciation charge into halves and allocated one half to demand and the other half to commodity. No finding of fact supports the conclusion. We are supplied with no data with which to test its validity.

The Commission says that "functional rather than physical depreciation is controlling", but no finding of fact is offered as a premise to that conclusion, and we are directed to no evidence which would serve as a premise. Functional depreciation, as the term is usually understood, means that caused by inadequacy or obsolescence. In the present case, if the life of the gas reserves is less than the normal physical life of the plant, then obsolescence would seem to fix the limit of its service life. But that would hardly seem true as to all items of equipment, and we have no findings of fact by which to test the validity of the statement. Depreciation is an important item of cost, and an allowance of a correct amount is a right of the company. Upon remand the Commission is instructed to make findings of fact which will support whatever allocation of this item it proposes.

Petitioner next complains that the Commission rejected all evidence of record in allocating supervision and engineering costs. All witnesses on the subject assigned it to demand. The Commission divided it, assigning $21,738 to demand and $46,206 to commodity. It said that these costs "have been prorated on the basis of the supervised expenditures of the respective groups, and this is equitable." The record, so far as it is cited to us in any of the briefs, is barren of evidence upon the item, except for the mere entry to demand in the exhibits. It appears that the difference between the allocation of the witnesses and the allocation of the Commission is the difference between assigning supervision and engineering to plant and assigning it to expenditures which are supervised. Such a difference would rest upon a difference of fact, and if in fact the costs were incurred for the purpose of supervising expenditures, we see no reason why the Commission's choice of the controlling principle is not correct. But if, as a matter of fact, the cost is a fixed cost related only to the plant itself, it would not seem to have commodity characteristics. The facts as to the true nature of the item are easily ascertained and should be found.

Petitioner complains of the Commission's allocation of general and administrative costs upon the basis of supervised expenditures exclusive of gas purchased. This seems to be a reasonable principle and within the discretion of the Commission. The same principle was used by the Commission staff in its testimony.

Petitioner complains of the Commission's allocation of taxes other than federal income taxes, half to demand and half to commodity. The Commission says only that "property taxes" are treated consistently and similarly to income taxes. The major part of these taxes ($210,971.28 out of $302,321.38, according to the exhibit of the Commission's staff) are property taxes. The balance is composed of capital stock, unemployment compensation, old-age benefit, franchise, power, state income and city license taxes. The Commission staff allocated all of them to demand, except capital stock and state income, which it divided equally. The company witnesses allocated this entire item to demand. In the Canadian River Gas Company case, supra, as we have seen, the Commission allocated all these taxes to demand, except that it divided the state income tax equally. We are unable to say, upon the basis of the brief reference to the item in the opinion of the Commission whether its division of this entire item equally between demand and commodity, is or is not arbitrary. The findings are neither complete enough nor clear enough. Certainly, on general principles, property taxes would seem to be a fixed charge and directly related to the plant itself. We do not see how they could vary with volume of gas handled, and neither the evidence nor the opinion of the Commission supplies information to that effect. Property taxes are in no way similar in nature to income taxes, which depend upon profit and are thus a function of both demand and commodity, as the Commission properly found.

Petitioner complains of the Commission's allocation of the rate of return, half to demand and half to commodity. This is an established treatment of this item. We have noted the explanation of it in the Canadian River Gas Company case, supra. It was the method used by the staff witness in the present case. The Commission points out that return is a capital cost and that the investment of capital serves both demand and commodity purposes throughout the year. Partaking of the elements of both, the division is made. The company says that the return is determined by a percentage of the rate base, which is property, and is proportional to the property and not to the volume of sales. Thus, says the company, it falls squarely within the definition of a demand cost and without that of a commodity cost. We think that the basis for the Commission's allocation of this item is clear and complete, both from past history and on the present record, and was within its discretion. The return earned is a function of the volume of sales as well as of demand.

In conclusion upon the matter of allocation of costs, we repeat that the statute imposes upon us certain functions of review of Commission orders. The Supreme Court has prescribed guides for our action in that review. It has repeatedly emphasized, as it said in Colorado-Wyoming Gas Co. v. Federal Power Comm., that the court must know what the findings are before it can give them the conclusive weight provided by the statute. It has emphasized the necessity for adequate findings of the facts upon which the order rests. It has also held that the grounds upon which an administrative order must be judged are those upon which the record discloses that the action was based, and that where the decision of a Commission is explicitly based upon the applicability of certain principles, its validity must likewise be judged on that basis. The Court has held that it is our function to review the record for the substantial evidence said to support the findings.

That the Commission might have adopted some other method or formula for determining the costs of the regulated business is beside the point. In so far as it purported to adopt the demand-commodity formula, our review must be upon that basis. The Commission cannot announce the applicability of a formula and then distort its application by failure to find accurately the factors required by the formula, or by departing from the essential progress of the formula from premise to conclusion. When the Commission announces principles or formulae as applicable, the validity of its order can be determined only by measuring what it does against the principles it announces. This is so not only upon the authority of Securities Comm'n v. Chenery Corp., but because any other course would permit an administrative agency to announce a proper principle and, under that protection, achieve an inproper result by unrevealed considerations wholly apart from the announcement. The prescribed judicial review would be set wholly at naught by any such procedure. In so far as the Commission purported to act upon its own informed judgment, apart from formulae or general principles, its findings and reasons must be clearly and completely shown.

The order of the Commission as to allocation of costs will be reversed and remanded with instructions that the Commission make clear the formula, or method, it uses in ascertaining the costs of the regulable sales, and that it make findings of the facts, based upon substantial evidence in the record, upon which it makes its allocations of compressor station labor, depreciation, supervision and engineering, and taxes other than federal income taxes.

We have carefully examined each of petitioner's contentions in respect to the allowance of operating expenses. These relate to labor costs, past service pensions, contributions to customer companies, rate case expense, and federal income taxes. The Commission has left open the possibility of adjustment in the last two items, in the event that the rate case expense exceeds he $70,000 estimated by it as the probable total, or that federal income taxes be increased by adjustments by the Bureau of Internal Revenue. We find no error in respect to expense allowances.

We have also carefully examined petitioner's contentions in regard to the rate base, including working capital, but we find no error in these respects.

Petitioner urges the invalidity of that part of the order of the Commission which directs it to make effective in total amount whatever reduction of its costs of gas purchased from The Interstate Natural Gas Company is finally approved by the courts upon review of the Commission's order in that case. We need not pass upon that point, as the Supreme Court has granted certiorari and heard argument in the Interstate case, and the rates of that company will shortly be fixed. Since the present case is remanded, that reduction, if any be proper, can be effectuated in such manner and amount as proves proper upon the reconsideration.

Affirmed in part, reversed in part, and case remanded for further proceedings in accordance with this opinion.

On Petition for Rehearing.

In its petition for rehearing the Commission presents one contention in a manner which makes an additional word advisable, lest the problem or the decision upon it be misunderstood.

The subject is the allocation of costs, which is a process of ascertaining certain costs which pertain to the regulable business and which, therefore, must be included in the computation of the allowable rates to be charged consumers. The Commission says that a reviewing court is limited to a consideration of the "end result", the rate, or price, prescribed by the final rate order for the commodity or service. It says that when the court finds this rate to be fair and reasonable its function is exhausted. Upon that premise the Commission seems to contend that the court has no power to examine the composition of the prescribed rate or the several elements which together constitute the rate.

The danger of misunderstanding and confusion arises because the Commission's premise stems from a trite and true generality. It is true that if the rate prescribed by the Commission be fair, reasonable and non-confiscatory, a court cannot disturb it. The questions are: When is a rate fair, reasonable and non-confiscatory? And how does a court determine whether the rate is so or not? The Commission seems to contend that the court must examine the rate per se, as an abstraction, or as a naked economic fact, divorced from the elements of which it is composed and regardless of its effects; and if it appears fair and reasonable upon such examination, it must stand. Such contention is without merit, either upon reason or upon authority, and moreover is impossible of practice.

The character of a rate — fair, reasonable and non-confiscatory as our law knows it, is not an abstract economic concept of the proper price to be paid by consumers for a commodity or service. Our law has never provided that either a company, a commission, or a court can fix a price for a utility commodity or service by an abstract observation or by a comparative evaluation of current prices for other commodities or services, or by any such process. The character of the rate has always been determined, in our law, by its relationship to the sum of a number of components. Those components, principally, are the expenses of the operation, an allowance for depreciation or depletion, and a proper "return" to the company. If the rate yields no more than the sum of these items, it is fair and reasonable. If it yields that sum, it is non-confiscatory. So far as we know, there is no rate case in American reports in which the proper rate has not been determined by a determination of these components and then their combination into a total. And there is no reported court review which was not concerned with one or more of those components. Constitutional restrictions upon government control of public utility rates have been held by the courts to require this sort of process.

The "end result" of a rate is not an intangible characteristic, nor is it the relative status of the rate as a price in the current economy. The "end result" is the ability of the rate to meet the sum of the costs required to conduct the operation in fairness to the consumer and to the company. The matter is nowhere better illustrated than in the opinion in which Mr. Justice Douglas announced the rule of the Supreme Court based upon "end result". The "end result" of which he spoke was described by him in terms of expenses and capital costs. And even when he discussed "end result" in relation to the return to the company, which was the specific subject there under consideration, he spelled it out in terms of the required components of the return, i.e., an amount which will provide service on the debt and dividends on the stock, sufficient to assure confidence in the financial integrity of the enterprise, and to maintain its credit and to attract capital.

It follows from the foregoing that a court can review the fairness, reasonableness and non-confiscatory character of a rate only by reviewing the propriety of the elements of which the rate was composed. No other way is known to our law, or ever was followed by any authority so far as we know. No standard is available for use in any other process of review.

When the court examined the "allocation of costs" in this case, it was examining the "end result" of the prescribed rate, because one phase of the "end result" is that the rate must yield enough revenue to meet all proper expenses. The contention of the Commission that a reviewing court has no power to inquire into a challenged refusal to include certain expenses in the computation of a prescribed rate, is without merit.

Not only do the foregoing general observations apply in the present case, but this allocation involves one of those difficult and delicate balances between federal and state authority, and thus in a peculiar respect requires judicial scrutiny. The federal statute, as was necessary, forbids federal regulation of local intrastate rates. But, obviously, if the federal commission can arbitrarily refuse to allow in its regulated interstate rates the full and true costs of the interstate business, it can present the company with a dilemma. If the uncompensated interstate costs be imposed upon the local intrastate business, the federal authority has, by indirection, control of those rates, even to the extent of making such business so expensive as to be prohibitive. If the uncompensated interstate costs be not imposed upon the intrastate business, obviously the fairness and reasonableness of the prescribed interstate rates are fictions.

The Commission says that nowhere in its opinion did this court say that the resulting rates were unreasonable, unfair or confiscatory. A rate order which does not provide for proper allowable expenses, taxes, depreciation and return, is unfair, unreasonable and confiscatory. We thought it unnecessary to recite that obvious basic premise.

The Commission also now attempts to translate the carefully stated conclusion of the Supreme Court upon the peculiar facts and concessions in the Panhandle Eastern case into a general doctrine applicable to all cases. The Commission says that that decision supports its position here, because the Court there held that no formal allocation of costs was there necessary. That the opinion in that case was based upon facts, agreements and concessions distinctly present in that record was meticulously stated by the Court. The case at bar presents no such record or question.

The Commission says, "The Court, unlike the Commission, is not an expert body * * *." Be that as it may, the Commission, even though expert, is forbidden to be arbitrary. And the courts are directed to prevent it from being so. The Commission cannot, and we do not understand it to claim that it can, shield arbitrariness by averments of its own infallibility, by technical expressions, or by failure to state adequate reasons for its conclusions. This court, in the present case, has not reversed the conclusions of the Commission, except in the procedural sense necessary to a remand. It has remanded the case for clarification where clarity is not present, and for completion where incompleteness now exists. When the findings and conclusions are complete and clear, the court will then, if appropriate proceedings are brought, consider whether the ultimate rulings of the Commission are within the permissible bounds of its power. The Commission says that the court has substituted its judgment for that of the Commission. The court has not done so. That contention in the present case is merely the recitation of a favorite cliche of administrative tribunals whose findings and conclusions have been queried by a court.

The other contentions of the Commission in its petition for rehearing were presented upon its pre-decision brief and argument. We have reexamined them but are of opinion that the petition should be denied.

Denied.