VI. Long-Run Marginal Costs (LRMC)

A. Summary

Since the inception of LRMC ratemaking for gas utilities, there has been an ongoing debate over the appropriate methodology for calculating both customer marginal costs and marginal costs for the demand related functions of distribution, transmission, and storage. That debate continues in this proceeding.

Both ORA and TURN recommend replacing the existing "rental method" for calculating customer marginal costs with the "new customer only" method. While the Commission originally adopted the rental method in its LRMC policy decision, that method has subsequently been replaced by the NCO method for every major gas and electric utility except SoCalGas.

The original LRMC policy decision found that the capital component of the demand related marginal costs for distribution, transmission, and storage should be based solely on the incremental investments needed to meet growth in demand. In its testimony in the 1995 PG&E BCAP, ORA identified several problems with the adopted methodology and recommended modifying it by including not just the investments needed to serve demand growth, but also the investments needed to maintain system reliability. The Commission adopted the proposed "replacement cost adder" as a "necessary refinement" to the existing methodology. (Re Pacific Gas and Electric Co., D.95-12-053, 63 CPUC2d 414, 433.) Both ORA and TURN have recommended adopting the replacement cost adder in this proceeding for both SoCalGas and SDG&E.

In addition to the above policy recommendations, ORA also takes issue with some of the more technical aspects of SoCalGas' marginal cost estimates. Each of these issues is addressed below.

B. NCO/Rental Method

In this case, as in each cost allocation proceeding since 1992, the Commission is faced with a choice between the rental method and the NCO method for calculating customer marginal costs. The two approaches are significantly different in both concept and outcome.

The NCO and rental methods both begin by estimating the cost of installing the service line, regulator, and meter (SRM) at a customer's premises. The NCO method assumes that the SRM facilities that have been installed for existing customers are a sunk cost. Consequently, only the SRM investments for new customers anticipated over the BCAP period are considered in determining marginal customer costs. A second component is then added to the SRM capital estimate to reflect the replacement of existing SRM facilities due to wear and tear. Finally an annual O&M cost is applied to all customers.

The rental method, like the NCO method, begins with an estimate of SRM costs. This estimate is then annualized using a real economic carrying charge (RECC). The resulting "rent" is then charged to all customers. The same O&M component used in the NCO method is also applied to all customers. In essence, the rental method treats all customers as new customers and requires them to pay a rental fee to gain access to the system.

The two approaches result in significantly different marginal customer costs. For example, the marginal cost for SoCalGas' residential customers is $75 under the NCO approach and $120 under the rental approach (1999$). For commercial/industrial (G-30) noncore customers the marginal cost is $5,274 under the NCO method and $8,852 under the rental method. On a cost allocation basis, the rental method allocates $31.5 million more to the core than the NCO method.

The proponents of the NCO method claim that the rental method is based upon an inappropriate theoretical foundation: a hypothetical competitive rental market with no opportunity to pay hookup charges or purchase the equipment. As a consequence, the rental method significantly overcharges customers.

The proponents of the rental method claim that it is the NCO method which is fatally flawed because it is the rate of growth of a particular customer class which drives the marginal cost estimates. As an example, they point to the impact that the NCO method had on the gas engine class following our initial adoption of the NCO method in SoCalGas' last BCAP. Because the NCO method resulted in an 80% increase for this class, we elected to retain the rental method (D.97-08-062). These proponents believe the NCO method is theoretically incorrect, is not based on cost causation, and sends inaccurate price signals.

Until this proceeding, ORA has been a consistent advocate of the rental method. However, ORA has elected to not pursue adoption of the rental method in this case given the long line of Commission precedent stating a preference for the NCO method. As noted below, we have now considered the arguments in favor and against the NCO and rental methods on several occasions and have consistently opted for the NCO method. There is nothing unique in this case justifying a deviation from that long line of precedent.

In the original LRMC decision, we were faced with choosing between the NCO method proposed by TURN and PG&E and the rental method proposed by the other utilities and the Division of Ratepayer Advocates (DRA) (ORA's predecessor). We opted for the rental method observing that it had been in use for electric utilities for the past four years. At the same time, we noted that the NCO method was being actively considered for electric ratemaking purposes in PG&E's then pending general rate case. (Re Rate Design for Unbundling Gas Utility Services, D.92-12-058 47 CPUC2d 438, 463.) In fact, we adopted the NCO method for electric ratemaking purposes on the same day. (Re Pacific Gas and Electric Co., D.92-12-057, 47 CPUC2d 143, 293.) One of DRA's arguments against the NCO method in the PG&E GRC was that it was unstable in that the marginal costs were driven by the rate of growth of a particular class. That argument, which is also being made in this proceeding, was rejected.

The issue was revisited in the 1995 PG&E BCAP with TURN and PG&E recommending a revised NCO methodology. Under the revised method, a component was added to the marginal cost to reflect the replacements of existing SRM facilities due to wear and tear. ORA continued to support the rental method while acknowledging that the proposed revision to the NCO method was an improvement. We adopted the NCO proposal noting that it "provides a better measurement of the future costs the utility will incur to serve its customers and therefore should be adopted." (Re Pacific Gas & Electric Co., D.95-12-053, 63 CPUC2d 414, 437.)

The issue was revisited in SCE's 1996 general rate case. We began our analysis by noting that its goal was to establish marginal costs that simulate pricing in a competitive market. (Re Southern California Edison, D.96-04-050, 65 CPUC2d 362, 403.) We went on to note that:

Parties opposing the NCO method argue that marginal costs should not distinguish between existing and new customers or vary according to the growth rate in new customers within a class. They argue that all customers should see the same per unit marginal costs, consistent with pricing in a competitive market. They point out that other components of marginal cost demand costs do not distinguish among new and existing customers in this manner. In their view, the methodology for calculating marginal customer costs should similarly apply an annualized charge to all customers.

We then proceeded to analyze and reject each of these arguments finding: (1) that the NCO method fully comports with marginal cost pricing theory; (2) the rental method is premised on an assumption concerning opportunity value that does not hold for customer hookups; and (3) the rental method does not produce a competitive price for customer hookups and, in fact, significantly overstates the price that would prevail in a competitive market (Id., pp.403-404) In short, we considered and rejected each of the arguments being made in this proceeding.

Finally, the issue was revisited yet again in SoCalGas' 1996 BCAP with both TURN and SDG&E proposing the NCO method and SoCalGas, ORA, and other intervenors supporting the rental method. The NCO method was again attacked on grounds that the rate of growth was the primary driver of the allocation and that small, rapidly growing customer classes could experience rate volatility. We adopted the NCO method finding that:

The NCO method is preferable to the rental method as it improves both the price signal sent to the customer and costing accuracy. Parties have not presented any new evidence in this proceeding that causes us to change the conclusion we reached in PG&E's last BCAP, D.95-12-058, or Edison's GRC, D.96-04-050. (D.97-04-082, Slip Opinion, p. 59.)

SoCalGas subsequently filed a petition to modify D.97-04-082 noting that application of the NCO method to the small, rapidly growing gas air conditioning and gas engine classes resulted in rate shock. In response to the petition, TURN made several proposals to ameliorate the rate shock. However, since these proposals were not a part of the record in the proceeding, we elected to retain the rental method for SoCalGas. At the same time, we continued to apply the NCO method to SDG&E and further indicated our continuing preference for the NCO method. (D.97-08-062, p. 4.)

SoCalGas and other proponents of the rental method continue to point to the impact of the NCO method on small rapidly growing classes as a reason for rejecting it. However, based on the evidence of record, that argument is now moot since the growth rates of the gas air conditioning and gas engine classes have now subsided. As indicated by SoCalGas' workpapers, those classes grew rapidly in the first few years after they were created. However, the growth rate has now flattened out and is expected to remain relatively flat through the BCAP period. The gas air conditioning class shows a zero growth rate, while the gas engine class shows 7.7% growth rate. For the air conditioning class, the NCO method produces marginal costs which are lower than those resulting from the rental method. For the gas engine class the NCO method produces marginal costs that are only 8% higher. In short, rate shock is no longer a viable basis for rejecting the NCO method. In any event, we have numerous tools at our disposal, such as rate caps, for preventing rate shock. We agree with ORA that the potential impact of the NCO method on small, and rapidly growing classes is an insufficient basis for rejecting the methodology given that the argument has been considered and rejected on several prior occasions.

C. Replacement Cost Adders

The Commission's initial LRMC policy decision adopted a methodology for estimating marginal capital costs for the demand related functions of transmission, distribution, and storage that focused solely on the incremental investments needed to satisfy demand growth over the planning horizon while maintaining the appropriate level of reliability. The methodology gave no consideration to the capital investments required over the planning period to replace equipment which was either worn out or which had to be upgraded to satisfy environmental requirements.

ORA identified a number of problems with this methodology in PG&E's 1995 BCAP. First, the practice of ignoring the replacement of worn out facilities for the demand functions was inconsistent with both the rental method and NCO methods of calculating marginal customer costs. Second, ignoring these "opportunity costs" could either prevent capital recovery for these long-life investments or shift the cost responsibility to captive customers. Third, the methodology artificially lowered the marginal costs. Since the utilities were authorized to discount down to LRMC to meet potential competition from bypass pipelines, an artificially low marginal cost for a function such as transmission had the potential to stifle competition. To remedy this problem, DRA recommended that the Commission adopt a "replacement cost adder" to account for capital additions needed to replace worn out facilities or to satisfy environmental requirements. Our decision adopted the DRA recommendation. (Re PG&E, 63 CPUC2d 414, 432.)

SoCalGas and others argue that there is no need for a replacement cost adder. In response to ORA and TURN arguments that the 1995 PG&E BCAP included a replacement cost adder, they cite the 1996 SoCalGas BCAP decision which states "we do not view that decision as precedential because it was based solely on the circumstances surrounding PG&E's resource plan in that case." (D.97-04-082, p.49.) They also argue that including a replacement cost adder in the existing LRMC methodology would result in a double counting of replacement costs, and that the replacement costs considered by ORA and TURN are not actually marginal costs.

In SoCalGas' last BCAP, ORA and TURN recommended that the replacement cost adder adopted for PG&E be applied to SoCalGas and SDG&E. However, we rejected the replacement cost adder because it was precluded by the Global Settlement:

While pure economic theory argues for inclusion of replacement costs in a true long run marginal cost methodology, the Global Settlement does not allow a methodology change of this magnitude which goes beyond a mere "refinement" and results in a significant cost shift not envisioned by the signatories to the Global Settlement. Even if the Global Settlement could be overlooked, which this decision finds if cannot, the Commission should more properly consider a change of this magnitude in a reexamination of our natural gas strategy and policies. (Id., p. 49.)

We said that while the replacement cost adder had been adopted for PG&E, we did not view that case as precedent because it was based solely on the circumstances surrounding PG&E's resource plan. With the Global Settlement no longer an issue, the parties opposing the replacement cost adder argue that the PG&E BCAP decision should not be considered a precedent and that the issue should be deferred to the Natural Gas Strategy proceeding. The record evidence in this case indicates that they are wrong on both counts. The PG&E BCAP decision essentially agreed with ORA's generic analysis of the problems with the existing methodology. Since the ORA analysis was generic, it is not surprising that each of the problems we identified with the PG&E resource plan is also present with the SoCalGas resource plan.

We first found that PG&E's resource plan did not measure the entire quantity of service being provided nor did it measure all changes in output. This is also true of the SoCalGas plan since it fails to include investments needed to replace worn out facilities thereby maintaining the level of reliability (i.e., service provided). As noted by TURN, in both the PG&E BCAP and this proceeding, an investment in replacement facilities is a change in cost to prevent a negative change in output. By failing to consider replacement costs, the SoCalGas resource plan fails to measure this change in output. The second problem with the PG&E resource plan was that it measures a shorter time horizon than the long term. The SoCalGas and SDG&E resource plans have the same problem as the PG&E plan since all three are based on a 15-year planning horizon. (Re Pacific Gas and Electric Co., D.95-12-053, 63 CPUC2d 414, 430.) The third problem with the PG&E plan was that it reflected only a small portion of the forward looking capital costs it would spend in providing service. The same is true of the SoCalGas resource plan. SoCalGas proposes spending $18 million over a 15-year period, or less than $1 million per year on growth related investments. At the same time, its expenditures on replacements over the 1994-1998 time frame averaged over $12 million per year.

We concluded our discussion of the replacement cost issue in the PG&E BCAP decision by noting a number of negative consequences associated with the understating of marginal costs.

· it would send an improper price signal to customers

· it would permit PG&E to subsidize potentially competitive sectors of its business

· it would provide less incentive for economic efficiencies

· it would cause revenue responsibility to unfairly shift to captive customers, and perhaps most importantly

· it would allow PG&E to collect revenues in a manner not available to firms subject to competitive market forces. Re Pacific Gas and Electric Co., supra, p. 433.)

None of the problems associated with understating marginal costs by excluding replacement investments are unique to PG&E. These negative consequences flow equally to the ratepayers of SoCalGas and SDG&E. In summary, all of the marginal cost related problems we identified with PG&E's plan, as well as the negative consequences that flow from that plan, are present in this case.

The parties opposing the replacement cost adder continue to argue that this issue is more properly addressed in the GIR proceeding, R.98-01-011. While we may have viewed the GIR as an appropriate forum for this issue at the time we issued the BCAP decision in 1997, the issue was never actively considered and the rulemaking has now been closed. (D.99-07-015, p. 146.) Furthermore, it was not one of the issues identified for consideration in the upcoming cost/benefit phase of the proceeding. In short, this issue never found a home in the GIR. Since this is the type of cost allocation issue which has been routinely considered in past BCAPs, including the PG&E BCAP which adopted the replacement cost adder, it is appropriately addressed here.

D. Customer Costs

We must adopt an estimate of the costs associated with installing SRM regardless of which methodology, rental or NCO, is adopted. ORA used SoCalGas' SRM data and a five year historical average of customer growth to develop its marginal customer cost estimates.

The NCO methodology adopted for PG&E includes a replacement cost component to reflect the replacement of existing customer services as they wear out. ORA used five years of historical data to develop a replacement rate. For meters and service lines, the replacement rate recommended by ORA is approximately 2% and 0.5% respectively. The ORA estimate also considers that 50% of the meters are refurbished and the cost of replacing service lines is twice as expensive as new installations.

E. Wholesale Rates

Long Beach continues to take issue with the use of the marginal cost methodology for purposes of allocating costs to wholesale customers such as itself. It continues to request that costs be allocated to wholesale customers on the basis of embedded costs, yet it presents no cost studies showing the results of an embedded cost allocation. As a fallback, it requests that the EPMC scaler not be applied to wholesale customers because it reflects costs not directly attributable to wholesale customers. Long Beach has raised these concerns on two prior occasions and lost both times. (D.94-12-052 58 CPUC2d 306, 337; D.97-04-082, Slip Opinion, p. 63.) The recommendation is rejected again.

F. Distribution Marginal Costs

The Commission has adopted a linear regression methodology for calculating distribution marginal costs which relies on 10 years of historical data and five years of forecasted data. In the model, 15 years of cumulative investment is regressed against cumulative incremental load. SoCalGas has used this approach in calculating marginal costs for both its medium and high pressure distribution systems.

Both ORA and TURN are of the view that the forecast of distribution investments for the period 1998-2002 is unreasonable and should not be used in estimating marginal costs. ORA proposes taking a five year historical average and applying a 3.75% annual growth rate to derive a forecast of investments for the period 1998-2002. TURN proposes two alternatives: (1) a regression using the entire 15-year period but assuming a constant medium pressure distribution cost per customer for 1998-2002 equal to the 1993-1997 costs; or (2) a regression based on just 10 years of historical data. Of the two alternatives, TURN prefers the first. A comparison of SoCalGas' distribution marginal costs and those of ORA and TURN are set forth in the following table. To place the estimates on an equal footing, the replacement cost adder has been removed from the ORA and TURN estimates.

TABLE 3

DISTRIBUTION MARGINAL COSTS WITHOUT

REPLACEMENT COST ADDER

($ 1999)

SoCalGas ORA TURN

MP $/Mcfd	97.6561	86.1939	82.7713
HP $/Mcfd	0.75907	0.6923	0.6876

Any of the three alternatives presented by ORA and TURN is preferable to the SoCalGas estimate since, as noted below, its forecast of investments for the period 1998-2002 is simply unreasonable.

The historical distribution investments for the period 1993-1997 were $28, $18, $23, and $17 million, respectively, or an average of $21.5 million per year. For 1998 SoCalGas forecasted investments of $36 million and expected this estimate to escalate at a rate of 4% per year through 2002. However, the actual investments experienced in 1998 were only 50% of the forecast, or approximately $18 million. Put another way, the actual investment for 1998 was less than the historical average on which ORA relies, indicating that the ORA recommendation of using the 1993-1997 average and escalating it at a rate of 3.75% is reasonable.

Another indication that the SoCalGas forecast of investments is too high, is the fact that the significantly higher investment forecast is not matched by a significant increase in load growth. Indeed, the company's projected 1998 peak month and peak day demand were either lower, or about the same level, as that experienced over the last five years.

SoCalGas downplays the lack of growth in peak demand by claiming that the number of new customers, rather than peak demand, is the main driver of new investments. Over the forecast period the average number of new customers per year is 51,768 or 38% higher than the 1993-1997 average. Even if this is the case, it doesn't justify the high level of forecasted investments. In the last BCAP, SoCalGas forecasted an average customer growth rate of 54,000 customers per year. This previous forecast, which was higher than the current one, was accompanied by an investment forecast of only $23 million per year. In other words, in the last BCAP, the company was forecasting even greater customer growth, yet the investment forecast was more in line with the historical average on which ORA is relying. Furthermore, even that forecast proved to be too high, with the actual investments for the period 1993-1997 averaging only $21.5 million. In sum, there is absolutely nothing supporting a virtual doubling of the distribution investments for the 1998-2002 period. TURN's recommended adjustments to this forecast are reasonable.

G. Impact of the Joint Recommendation

The JR would resolve each of the issues addressed above except for the issue of whether Long Beach should continue to be subject to LRMC ratemaking. The parties agree to the use of the NCO method for calculating marginal customer costs. The parties also agree on the precise manner in which the methodology should be implemented.

First, the NCO method should be implemented without a replacement cost adder. This is consistent with the parties' agreement to exclude the replacement cost adder in calculating demand related marginal costs. As ORA noted in its testimony, the replacement cost adder should either be included for all functions or excluded for all functions in order to achieve methodological consistency. Second, the parties agree to use TURN's RECC factor and A&G loading factor in developing the customer marginal costs. SoCalGas had already agreed to TURN's adjustment to the A&G loader in its rebuttal testimony. Third, the parties agree to SoCalGas' treatment of developer contributions. Finally, the parties agree that the gas engine transportation rate will be set at SoCalGas' proposed rate of $0.20384 per therm. This agreement resolves the issue over the impact of the NCO method on new customer classes that experience significant growth in the early years. In effect, the JR would cap the rate to avoid rate shock. This is consistent with TURN's recommendation on this issue as well as past Commission practice. The shortfall of approximately $1 million would be allocated to other core customers on an EPMC basis.

For marginal demand related costs, the parties agree to exclude the replacement cost adder. Adoption of the replacement cost adder in a manner consistent with the PG&E BCAP decision would shift approximately $7 million to the noncore. The JR would also adopt TURN's recommendation regarding medium-pressure distribution marginal costs. While this would shift $1.6 million to the noncore, this amount is considerably less than what would occur if ORA's estimate was adopted.

The two major components of the marginal cost package described above are the adoption of the NCO method for calculating customer marginal costs and the exclusion of the replacement cost adder for each functional category. This compromise is more than fair to noncore interests considering that we have already adopted the NCO method for every utility except SoCalGas and have also adopted the replacement cost adder for PG&E. Furthermore, in SoCalGas' last BCAP we acknowledged that the replacement cost adder was conceptually sound even though it wasn't adopted because of the limitations contained in the Global Settlement. In short, if this issue is fully litigated there is a strong possibility that we would adopt both the NCO method and the replacement cost adder, consistent with the policy adopted for PG&E.