MORGAN, Justice.
This is a rate case which arises from the July 17, 1975, filing of a notice and application for a general increase in electric rates by the Northwestern Public Service Company (company). The proposed rates would have raised company’s electric revenues by approximately $8,450,000 per annum. At the hearing staged before the Public Utilities Commission (PUC), the South Dakota Electric Consumers, a consortium of seven municipalities (cities), intervened and actively participated in the proceedings. The PUC issued its initial decision on September 27, 1976, purportedly allowing company approximately $4,800,000 per annum revenue increase. Company appealed this decision to the circuit court which remanded the case to the PUC for a rehearing. The rehearing was limited to various filings of information. The rehearing decision affirmed the initial decision except for a minor correction of a previous error in calculation. Company then appealed the rehearing decision to the circuit court which affirmed the PUC and dismissed company’s appeal, which action was then appealed to this court. We affirm in part, reverse in part, and remand for further proceedings.1
The rate-making power is vested in the PUC under SDCL 49 -34A. The statutes place the responsibility on the PUC to balance the need of the utility for adequate revenue with the protection of the public from unjust or unreasonable rates. To.this end it is neither a consumer advocate nor a utility advocate. The burden of proving that the rate sought is just and reasonable is on the utility.
Our scope of review has been discussed many times. Most recently it has been thoroughly reviewed in South Dakota Public Utilities v. Otter Tail, 291 N.W.2d 291 (S.D.1980), so that we will not go into such detail in this opinion.2 Succinctly stated, we (1) determine whether the PUC’s order viewed in light of the relevant facts and of the PUC’s broad regulatory duties abused or exceeded its authority; (2) examine the manner in which the PUC has employed the methods of regulation which it has itself selected and determine whether each of the order’s essential elements are supported by substantial evidence; and (3) determine whether the order may reasonably be expected to maintain financial integrity, attract necessary capital, and fairly compensate investors for the risks they have assumed, by providing appropriate protection to the relevant public interest both existing and foreseeable. Permian Basin Area Rate Cases, 390 U.S. 747, 88 S.Ct. 1344, 20 L.Ed.2d 312 (1968).
Before discussing the various issues, we make some general observations. The PUC has adopted the “cost of service” method of rate making. This method entails four steps as follows: (1) Properly determine company’s rate base, i. e., investment devoted to public service; (2) determine a fair and reasonable rate of return; (3) multiply the base [(1) above] by the rate [(2) above]; and (4) add to company’s cost of operations [465]*465referred to above (including taxes and depreciation). To assist in the computation of the steps above, a historical test year is adopted.3 The data from this year must be adjusted as to the cost of operations and the rate base to reflect changes which will be in effect subsequent to the historical test year. The adjustments in this case were considerably complicated by the effect of the Big Stone Power Plant (plant) on company’s operations when it became operational.4 This plant was erected by a consortium of three power suppliers: company, Montana Dakota Utilities Company, and Otter Tail Power Company, each of whom share proportionately in the output of the plant and in the expenses of the construction, maintenance, and other costs of operations. Company, which had formerly been only a purchasing company, became a generating company when the power plant went on line. This event occurred during the course of the test year. While company does not dispute, for the purposes of this appeal, the PUC’s determination in step (2), that a fair and reasonable rate of return would be 9.23%, they argue that because of errors by the PUC in the determination of the rate base in step (1) and the costs of operation in step (3), the net result was to allow a return of only 7.31%.
The disputed issues fall into two categories: (1) Those relating to rate base, of which there are three: (a) disallowance of CWIP; (b) the deduction of “negative working capital”; and (c) the determination of AFUDC. (2) Those related to cost of operations, of which there are three: (a) the determination of power supply costs; (b) deduction for ad valorem taxes on CWIP; and (c) the exclusion of annual payroll and pension expenses. We will treat them in that order.
DISALLOWANCE OF CWIP ON WHICH NO AFUDC WAS TAKEN
We first examine company’s argument that the PUC erred in excluding $303,607 from rate base, the value of certain “construction work in progress” (CWIP) at the end of the test year for which the in-service date was imminent. CWIP can represent a considerable investment by the utility. Yet under the PUC’s rate-making theory the newly constructed asset does not become a part of company’s base until it is actually in use. To compensate company for the investment during the interim the value of the investment is categorized as allowance for funds used during construction (AFUDC), and company is permitted to capitalize this investment. Contrary to the generally accepted concept of capitalization, i. e., record it as an asset and depreciate it out, the term as used in rate making means to calculate a rate of return representing the cost of the capital, considering the various capital components, which rate is applied to the investment figure and the result included in the rate base. While it may sound as though capitalization of AFUDC and the alternative of including CWIP in rate base while the work is under construction are equivalent methods of compensation, this is not so as far as cost recovery from the ratepayers. The capitalization of AFUDC matches the cost reflected in utility rates with the benefits concurrently received by ratepayers, whereas inclusion of CWIP in the base would require current ratepayers to pay for construction that will result in service in the future.
In its application company opted to treat the $303,607 as a capital asset and include the entire sum in its computation of base rate, rather than capitalize it as AFUDC, and the PUC excluded it from company’s rate base. In doing so the PUC stated in its rehearing decision that it is proper to exclude all CWIP from rate base.
[466]*466This court has previously quoted with approval Columbus Gas & Fuel Co. v. Public Utilities Commission of Ohio, 292 U.S. 398, 54 S.Ct. 763, 78 L.Ed. 1327 (1934), 91 A.L.R. 1403, wherein the Supreme Court said: “ ‘There will be no need in the computation of the rate base to include the . . . value of (assets) not presently in use, unless the time for using them is so near that they may be said, at least by analogy, to have the quality of working capital. . . . ’ ” Application of Northwestern Bell Telephone Co., 78 S.D. 15, 30, 98 N.W.2d 170, 178 (1959) (emphasis added). The PUC made no finding as to the imminence of the use of the CWIP in question. The circuit court, after quoting the correct rule from Columbus Gas & Fuel Co., supra, noted that other than a conclusory statement by a company witness there was no showing that the CWIP in question was soon to be in service, and that company had therefore failed to meet its burden in showing imminent service. We agree with the determination of the circuit court and hold that although the PUC was in error as to its narrow statement of the rule, the error was nonprejudicial for the reason that there is insufficient evidence in the record to sustain company’s argument that the use of the CWIP was in fact imminent.
DEDUCTION OF NEGATIVE WORKING CAPITAL
The next issue that goes to the rate base „ is the action of the PUC in adopting cities’ working capital allowance, actually a negative allowance, thereby diminishing the rate base by $1,292,435. This relates to the accruals on company’s books of account for taxes payable and tax collections payable but not due, which actually are the equivalent of funds contributed by ratepayers on which company is not entitled to earn a rate of return by inclusion in the base as working capital. While it would seem to be a fairly routine adjustment, we instead find three divergent views, with company and cities at opposite extremes and the PUC staff in between.
In the first instance company did not claim any working capital allowance in its rate base, asserting that such a small allowance was needed that it did not justify the expense of an analysis to support the claim. Cities, however, recommended an allowance for working capital, but then claimed that because certain accruals in company’s books for tax liabilities were ratepayer-contributed funds, and because the average of such accruals for the test year exceeded their proposed allowance, company’s rate base should be reduced by such excess. In arriving at the figure proposed by cities, they used a standard formula hereinafter referred to as the “FPC Formula,” as used before the Federal Power Commission, which was applied to company’s operation and maintenance expenses for the test year to arrive at a working-capital allowance from which figure were deducted what cities claimed were the accruals for tax liabilities, which amounted to ratepayer-contributed funds. While company acknowledges this court’s holding in N. W. Pub. Serv. v. Cities of Chamberlain, Etc., 265 N.W.2d 867 (S.D.1978), which approved treating accrual taxes as rate base deductions, company contested cities’ computation on two principal bases: (1) That the accruals on the books represent bookkeeping accruals and were not the equivalent of funds provided by ratepayers available to company for use as working capital; and (2) that the “FPC Formula” already takes into account the various accruals so the deduction amounted to an improper double-counting deduction.
A review of the record discloses that this item was a battle of the experts with testimony supporting all views, and in that respect this court, claiming no expertise in rate making, normally gives credence to the PUC decision. Here we are faced with the PUC’s total disregard of the computations of its own staff which fall midway between company’s and cities’. Staff urged that the cities’ computation was in error because the calculations failed to properly impact the plant’s effect on working-capital requirements. It is undisputed that the effect of sales of surplus capacity from the plant very dramatically increased operation and [467]*467maintenance expenses which resulted in an increase of working-capital requirement after application of the FPC Formula. Neither of the PUC decisions indicates that there is any error in the staff computation.
As to company’s principal argument that the book accruals do not represent actual cash available for company’s use as capital, company has failed to present any evidence to prove such discrepancy. Company has the burden of justifying their rate base. Company started out ignoring working-capital allowance as too insignificant to warrant the cost of analysis. Company was certainly aware of the negative impact of the accruals from its previous experience and should have been prepared to meet the issue with actual figures. Cities and the PUC used the figures available from company’s books and we are not inclined to say they had to go further. The PUC had the evidence before it and made their decision. It is not for this court to weigh the evidence. There appears to be substantial evidence on the whole record to support the PUC’s decision.
DETERMINATION OF AFUDC
The last rate base issue we deal with is the disposition of such AFUDC as company did capitalize and include in their computations of rate base. As previously noted, capitalization in this instance means to determine a rate representing the estimated cost of capital calculated from the weighted costs of the various capital components, i. e., common equity, stocks, bonds, etc., which rate is then multiplied by the investment in construction work, thus arriving at the AFUDC to be capitalized.
The PUC staff and cities both disagreed with company’s calculation. The disagreement came as to the treatment of income tax deductions for construction related interest expense. One of two alternative methods is used in the industry: normalized treatment or flow-through treatment. The selection of method dictates how the income tax deduction is handled. Under the normalized treatment the cost-of-debt component in the AFUDC rate is calculated net-of-tax deduction for interest. This is a lower rate which in effect passes the benefit of the tax deduction to ratepayers over the life of the newly constructed property. The alternative, flow-through treatment, uses the interest deduction in calculating the income tax expense included in the cost of services thereby reducing that component, leaving AFUDC to be calculated on a gross basis. The benefit from the interest deduction is “flowed through” to current ratepayers and is not available for deferral for the benefit of future ratepayers.
The method selected by the PUC was normalized treatment. Cities disputed the validity of company’s computation of net-of-tax rate as arbitrary and argued that the rate used was a gross rate. Staff urged that a net-of-tax rate results in a disservice to the ratepayers and recalculated the AFUDC rate at a gross rate calling for a flow-through treatment. The PUC decided to reject company’s computation as not a net-of-tax rate and then also rejected the recommendation of its own staff and adopted cities’ view. In its findings of fact the PUC stated that company had not proven its AFUDC computation was net-of-tax and went on apparently to treat it as a gross rate. However, no findings of fact or conclusions of law were made to that end.
This dispute was again a battle of the experts. In fact, what should have been a routine accounting exercise of determining the various costs of capital invested turned into a donnybrook with company saying “ ’tis” and cities saying “ ’taint.” Staff apparently took off in a third direction, suggesting that gross base should be used, but nevertheless asserting that company’s proposed rate was not net-of-tax. Each side can point to inconsistencies in the expert testimony of the other. The rehearing decision fails to shed any light on the problem. The PUC merely adopted cities’ argument.
It is not for this court to weigh the evidence. We must rely on the PUC’s expertise. We cannot say that on this issue they were arbitrary or capricious; nor do we agree with company’s contention that contradictions in the record render the evi[468]*468dence unsubstantial. This is true only where there is clear evidence on the other side. Universal Camera Corp. v. Labor Bd., 340 U.S. 474, 71 S.Ct. 456, 95 L.Ed. 456 (1951). In this case evidence on both sides was contradictory. We hold, therefore, that there appears to be substantial evidence on the whole record to support the PUC’s decision.
POWER SUPPLY COSTS
In the category of operation and maintenance expenses, company’s most spirited attack is launched against the PUC’s decrease in power supply costs which deleted more than $3,200,000 from company’s claim for those costs. Again, PUC adopted cities’ computation entirely without regard for the showing of their own staff. Company argues that the decision is not supported by substantial evidence and is arbitrary and capricious. We agree.
Much of the disparity in this instance arises from the going on line of the Big Stone Plant as previously mentioned and the conversion of company from a purchasing company to a generating company at the end of the test year. As 32.5% owner of the plant, company need no longer purchase power from other sources for resale since the plant provides all the power that company requires plus additional off-peak power which can be sold to other electrical utilities. All parties agree that the plant should be included in the test year rate base because of its obvious impact on company’s rates. Inclusion of the plant in rate base, however, requires a corresponding pro for-ma adjustment in the form of a credit to power supply costs for the test year to reflect the additional income company will receive from the sale of off-peak energy. The controversy relates to the magnitude of the adjustment which depends on a determination of how much electric power the plant can generate and how much of the surplus generated power can be sold.
Cities calculated the credit to company’s power supply costs using data contained in a letter written by the plant’s superintendent, Mr. Johnson, to an official of one of the other co-owners of the plant. This letter (hereinafter sometimes called the Johnson letter) was written on March 26, 1975, before the plant was operational. The letter is concerned with, as appears under “Subject” at the top of it, operating and maintenance costs. The letter contains Mr. Johnson’s prediction of the plant’s operating and maintenance costs assuming the plant operated “for the full year in 1975 at an unrestricted load.” The unrestricted load assumed by Mr. Johnson was 337 days at 396 MW with a heat rate of 10,500 BTU/KWH. Using these figures, Mr. Johnson concluded that the plant could produce 3,202,800 MWHs of generation. Not only did cities adopt the data in the Johnson letter to calculate the plant’s generating capacity, it used the same figures to compute company’s power pool sales, that is, the amount of excess power available for sale to other electric companies in the MAPP pool. Cities’ power pool sales estimate further assumed that 100% of the plant’s surplus output could be sold.
Staff disagreed with cities’ approach. As pointed out by the staff witness, cities’ calculations did not take into consideration the peaks and valleys of electric usage. Demand for power is greater during some seasons than others; in fact, during certain hours of the day. Common sense dictates that a plant the size of Big Stone would not be built merely to meet the present power needs, but rather to handle expanded future needs. Staff and company urged a less simplistic approach based on existing contracts and historical sales data. Company points out that power sales are a matter of supply and demand, but while the Johnson letter may give an indication of the available supply, it does not address the factor of demand.
Another point of attack by company on the computation of credit for wholesale sales is described as a mismatch in the use of two different costs of fuel. Computation based on the Johnson letter resulted in the figure of 5.41 mills per KWH. A November 1975 actual figure of fuel costs of 6.89 mills per KWH was established in the rec[469]*469ord. The deductive process of computation was first to determine the fuel costs per KWH of electricity generated; second to determine the fuel costs per KWH of the power that is wholesaled- and then to deduct the latter from the former which results in the costs to be passed on to the retail customers as an add-on in the operation and expense category.
Cities’ witness used the lower figure of 5.41 mills in his calculation of the total cost of power generated, but used the higher actual cost figure in his calculation of costs of wholesale power to be deducted. Company urges that this results in a mismatch, and we agree. On appeal, counsel for the PUC urges that the November cost figure is not accurate, that it is too short a period to reflect the true cost. In this he may be correct, but he fails to explain why it is so inaccurate that it cannot be used in one step but accurate enough to be used in the second. Whichever figure is used it should be used in both steps, otherwise they are comparing apples with oranges.
By the time the rate increase hearing began in April of 1976,. the plant had been in operation for nearly a year. Company presented evidence of its actual experience with the plant during that year which showed that the power production prediction contained in the Johnson letter was highly overestimated. Company’s evidence also showed that sales of 100% of excess power to other electric companies was not being realized as forecast by cities, but that only 65% of the excess output was being sold. The assumptions on which the Johnson prediction was based were neither verified nor proven to exist. The PUC adopted all of cities’ calculations and rejected all of the evidence presented by company, ignoring company’s actual experience with the plant for nearly a year which had elapsed between the test year and the hearings. We conclude that the PUC’s decision in this regard was arbitrary, since the result was a drastic underestimation of company’s power supply costs and a concomitant overestimation of the wholesale revenues which would stem from the operation of the plant.
“The purpose of using a test year is to establish with a reasonable degree of accuracy the revenue and expenses that a utility will experience during the period when the new rates will be in effect.” N. W. Pub. Serv. v. Cities of Chamberlain, Etc., supra, at 879. It is well-settled, however, that when reliable evidence of actual experience is available, it should supplant evidence of a purely theoretical and predictive nature. This principle was recognized and applied by this court in N. W. Pub. Serv. v. Cities of Chamberlain, Etc., supra, at 878 and 879. This principle was enunciated in West Ohio Gas Co. v. Public Utilities Com., 294 U.S. 79, 55 S.Ct. 324, 79 L.Ed. 773 (1935), wherein the United States Supreme Court held that the public utilities commission acted arbitrarily by using predictions of income and expenses based on test-year data and ignoring available evidence of actual post test-year earnings.
We think the adoption of a single year as an exclusive test or standard imposed upon the company an arbitrary restriction in contravention of the Fourteenth Amendment and of “the rudiments of fair play” made necessary thereby. The earnings of the later years were exhibited in the record and told their own tale as to the possibilities of profit. To shut one’s eyes to them altogether, to exclude them from the reckoning, is as much arbitrary action as to build a schedule upon guesswork with evidence available.
294 U.S. at 81 -82, 55 S.Ct. at 325, 79 L.Ed. at 776 (citations omitted).
We find the Court’s reasoning in that case equally applicable to the instant case when cities and the PUC relied almost entirely on the data contained in the Johnson letter. That letter was nothing more than a prediction based upon certain assumptions which were never verified or proven to exist. Company, on the other hand, presented concrete evidence of actual post test-year experience with the plant, which was in sharp contrast to the Johnson letter data. We conclude that the PUC’s reliance in this case on the speculative data presented by cities and its refusal to con[470]*470sider company’s evidence of actual results and the recommendation of its own staff was arbitrary and that the decision on this issue was not supported by the evidence.
AD VALOREM TAXES RELATED TO CWIP
The next item of cost in dispute is the PUC’s exclusion of $57,944 from company’s cost of services, representing the amount of ad valorem taxes related to CWIP. Company sought to include ad valorem taxes on CWIP in its cost of services because of the bookkeeping burden of capitalizing small amounts. The PUC ruled, however, that such taxes had to be capitalized and made the deduction. Company does not dispute the PUC’s ruling regarding method; however, it asserts that the amount deducted greatly exceeds the amount that should have been deducted.
The figure was arrived at by one of the expert witnesses for cities, purportedly from information furnished by company. Company disputes his basis for computation and denominates it as guesswork. Company had estimated total ad valorem taxes of over 1.5 million. When broken down the valuations upon which the taxes were computed were found to contain some 2 million dollars plus in the CWIP category. Under the PUC’s ruling such could not be included. Company does not dispute the method of computation but only the basic figure of 2 million dollars plus. Nowhere does company point to a more accurate figure; instead it refers to a “de minimus” amount. The figure appears to have been taken from company’s computation for tax deduction. The PUC had the expert testimony upon which to base its decision, and we find that to be substantial evidence and will not overturn it.
EXCLUSION OF INCREASE OF ANNUAL PAYROLL AND PENSION EXPENSES
The PUC excluded $355,017 of increased payroll and pension expenses from company’s cost of service, having found that the increased salary expense resulted in increased productivity.
It is settled that the PUC is allowed to exclude increased payroll expense where it is shown to be offset by increased productivity. N. W. Pub. Serv. v. Cities of Chamberlain, Etc., supra.
At the first hearing the testimony of members of the PUC’s staff indicated their beliefs that such would be the case here. The report upon which staff relied showed that between 1965 and 1974 company’s per-unit cost of producing power went only from 5.36 mills per KWH in 1965 to 5.72 mills per KWH in 1974, with fluctuations in between. Company does not dispute the report, they simply argue that this increase of .36 mills for KWH over a nine-year period is a huge increase and that the PUC’s finding is without basis. On rehearing staff apparently completely changed its position, as summarized in the findings of fact proposed to the PUC on the rehearing, as indicated by the following: “We have reviewed the record of this proceeding and are inclined to modify our decision and order in this respect.... Initially our review of staff exhibit No. 9 does not conclusively indicate a trend of decreasing wage and salary costs per KWH sold and upon reconsideration, we conclude that this expense appears to be justified on record.” And further “these expenses appear on the record to be somewhat justified, and since the commission recently approved a comparable adjustment in Northern States Power, F-3062, October 6, 1976, where the factual allegations were similar, we conclude in the sound exercise of our discretion in this matter, that it is proper to recognize a pro forma wage and salary adjustment of $355,-017.”
In spite of this stance of their own staff, in the rehearing decision the PUC stated: “The evidence before the commission fully supports this finding that increased productivity has, in fact, offset payroll and related increases.” They make no reference whatever to their own staff’s change of posture, nor do they suggest any distinction between this case and the Northern States Power case cited by staff.
[471]*471On appeal counsel for the PUC states in his brief that the staff witnesses’ testimony is fully supported on the record. Company’s witness attempted to controvert portions of staff witnesses’ testimony but was unsuccessful. On the basis of staff witnesses’ testimony and detailed study, the PUC was fully justified in finding that increased productivity would offset the payroll and related expenses claimed by NWPS in this proceeding. But nowhere does he allude to or explain the impact of the change of the posture of staff as indicated above.
We hold, therefore, that upon the face of the record the determination by the PUC on rehearing was wholly unsupported by substantial evidence and was arbitrary and capricious.
SUMMARY
We affirm the PUC rehearing decision with respect to disallowance of CWIP, deduction of negative working capital, determination of AFUDC, and determination of ad valorem taxes on CWIP. We reverse with respect to the determination of power supply costs, and the disallowance of increased payroll and pension expenses.
The case is remanded to the circuit court with directions to remand to the PUC for further action consistent with this opinion.
WOLLMAN, C. J., concurs.
FOSHEIM, J., concurs specially.
DUNN and HENDERSON, JJ., concur in part and dissent in part.