7. Discussion

7.1 Value of Service Study

Value of service provides a means to quantify the value customers place on reliable electric service. Value of service information allows the utility to make cost effective decisions that are consistent with customer's desires. PG&E did not prepare a value of service study as part of its 2003 GRC application and has not prepared a value of service study for at least a decade. PG&E developed an internal value of service guideline, Utility Operations Guideline G12003, in May 2001. The values in G12003 are not based on customer information or surveys, but are based on the value PG&E proposed to place on reliability as part of a performance-based ratemaking filing in September 2000. G12003 is used exclusively for the evaluation of discretionary reliability improvement projects. PG&E does not use G12003 to develop service restoration plans.

ORA initially recommended that PG&E change its method of valuing service but in joint testimony with PG&E supports a PG&E review of existing VOS studies to assess their usefulness. Aglet and TURN recommend a new VOS study. By virtue of its reliance on G12003 in development of the proposed performance incentive mechanism payments and rewards, CUE appears to support continued use of existing value of service assumptions.

ORA points out that the "values" in G12003 do not equate the value of reliability to customers with the value of system investment, but rather equate PG&E's proposed value of reliability with PG&E's estimated effect on reliability. ORA argues that these proposed values are not an adequate basis for operational decision making. Noting that PG&E has no plans to develop better value of service data, ORA recommended that PG&E base value of service on empirical analysis and present the results to the Commission by no later than the next rate case. ORA argued that operational decision-making should be linked to the value that PG&E's customers place on reliability and that these values would remain unknown until PG&E updates its value of service data.

ORA also pointed out that because G12003 uses three different measures of value of service (cents per customer minute of interruption, dollars per customer outage and dollars per line mile per outage) derived from different proposed metrics, which bear no direct relationship to each other, G12003 is internally inconsistent. ORA believes that using three different measures could result in inconsistent decision-making. ORA also recommended that the Commission require PG&E to consistently apply an empirically-based value of service metric, to service guarantees and storm response, in addition to investment planning.

PG&E argues that the value of service numbers in G12003 are consistent with other value of service literature, in particular, a 1992 survey paper by Woo and Pupp. PG&E states that: "Because the values are consistent with VOS literature, the numbers do in fact equate the value of reliability to customers with the value of system investment."

In their joint testimony, ORA and PG&E agree that a new VOS study is not necessary. Instead, PG&E and ORA agree that PG&E should perform an assessment of service values (a "VOS assessment") by December 31, 2004. The VOS assessment would analyze and appraise the differences within and among unspecified prior VOS studies, as well as the relative merit of "willingness to pay" versus "willingness to accept" and explain the derivation of proposed VOS values.

TURN recommends that before major investments and other decisions are made based on value of service, a new VOS study of PG&E's customers be done. TURN states that it does not understand ORA's suggestion for an empirical analysis short of undertaking a VOS survey. TURN believes that there are enough unknowns that any analysis based on past studies will not adequately answer the questions at hand. TURN also recommends that any such VOS study research the value of service from both the point of view of willingness to pay and willingness to accept.

TURN explains that values of service differ significantly by customer class. TURN believes that using existing VOS information to establish incentive mechanisms, like in CUE's proposal, is likely to lead to a situation where: 1) more reliability is paid for by the residential class than it wants, and 2) more reliability is provided to the residential class than it wants to pay for.

According to Aglet, the CUE/PG&E proposal is based on uncertain and out of date VOS information. The uncertainty stems from changed public perceptions about electric power. Aglet points out that one of CUE's calculations of VOS assumes that all costs of the Department of Water Resources contracts that the State of California executed in 2001 were directed toward improving SAIDI performance. Aglet argues, as does TURN, that this assumption is unrealistic and yields VOS figures that are too high.

Even ORA, who testified in the PG&E/ORA joint testimony that a "VOS assessment" is a sufficient mechanism to use in updating values of service, takes issue with the PG&E/CUE claim that the "adoption of additional reliability incentive expenditures are justified on a value-of-service basis." According to ORA, nowhere does the PG&E/CUE proposal indicate exactly which study, if any, the VOS data were obtained from. ORA believes that this is a very important omission because the PG&E/CUE proposal bases its calculation of rewards and penalties on VOS data. ORA argues that any performance mechanism should be based on the current values placed by affected ratepayers on that performance, not on an unspecified, outdated, value of service study that may have been conducted sometime in 1993, or perhaps 1990, for another utility. Since PG&E has not done a VOS study for at least a decade, the record does not contain any value of service information that would capture the tremendous changes that have transpired in the electric industry in the last seven to eight years.

Moreover, even PG&E testified that a new VOS study should be done prior to initiating any improvements based on customer value of service, suggesting that "any consideration of improvement targets [should] be held until PG&E can present the results of a customer value of service study... prior to PG&E's next GRC proceeding."²¹ PG&E's witness Bhattacharya testified that "a lot of things has [sic] happened since G12003 was put together...we had the energy crisis and customer's expectation on reliability has, I would believe, changed. Customers willingness to pay or expectation of reliability has changed."²²

Although the values contained in G12003 may be consistent with the Woo and Pupp study, we cannot find that they adequately represent PG&E's customers' current value of service. Additionally, as PG&E admits, the VOS data underlying G12003 does not indicate PG&E customers' willingness to pay for added reliability. It is time for PG&E to prepare a new value of service study. The most recent PG&E study was prepared in 1993, with updated estimates prepared for PG&E's September 2000 Performance-Based Ratemaking (PBR) filing. All parties agree that in the intervening years, electric restructuring and the electricity crisis may have significantly altered PG&E's customer's value of service. In addition, great changes have occurred in the California economy in general in the past few years that may also have affected PG&E's customer's value of service. A review of existing studies will likely result in parties making the same claims in the next GRC as they have here - that the studies are too dated.

PG&E's latest VOS survey was prepared in 1993. SCE prepared a 1999 update to its VOS, which quoted PG&E's 1993 study and translated those values to 1998 dollars, but a study of SCE's customers is not necessarily useful in determining the VOS of PG&E's customers. As TURN points out, for the one outage scenario that was common to both PG&E's 1993 study and SCE's 1999 update, PG&E's residential customers had a higher value of service than SCE's (PG&E's VOS/unserved kWh was $4.37 and SCE's was $2.52). Furthermore, TURN also points out that even SCE's 2000 study predates the energy crisis and that the record does not contain value of service information from any utility that shows whether or not the energy crisis or the 1995 or 2002 storms have changed PG&E's customers' perceptions of their value of service.

We agree with TURN that while an assessment of prior studies may offer some insight into how to better prepare a VOS study, such an assessment of prior studies will not shed any light on customers' current value of service, nor will it evaluate current customer's willingness to pay for improved reliability. For this reason, we decline to approve PG&E/ORA Agreement 6. Instead, we direct PG&E to prepare a new VOS study prior to its next GRC. PG&E should prepare a proposed value of service study approach and cost estimate for review and comment by ORA and other interested parties. PG&E should file its proposal by advice letter. At a minimum, the new VOS study should include a "willingness to pay" element.

7.2 Funding OIS Improvements

Agreement 7 of the PG&E/ORA proposal would allow PG&E to establish a memorandum account to record the costs associated with four specific upgrades to its OIS and emergency response systems designed to improve both its outage communication and storm response. Agreement 7 would establish a memorandum account to record the costs associated with the four upgrades and would allow PG&E to recover an amount up to $9 million in 2003, and $2.3 million for each of the years 2004, 2005 and 2006 (2003 nominal SAP dollars). The amount incurred in 2003 would be recoverable to the extent that PG&E's actual expenses in FERC Account 588 exceed 2003 GRC adopted FERC Account 588 revenue requirement by the amount that actual expenses exceed adopted revenue requirement up to the amounts in the memorandum account. For the expenses incurred in 2004, 2005, and 2006, the amounts would be recoverable up to the incremental amounts described above to the extent that PG&E's total electric O&M expenses exceed GRC adopted electric O&M revenue requirement. The four upgrades are 1) linking the OIS to mobile data terminals, 2) integrating three separate computer systems, 3) enhancing the mapping associations in OIS, and 4) programming to retain single customer outages in OIS. PG&E's request for funding of additional OIS upgrades warrants careful scrutiny given that PG&E was authorized $19.4 million in capital plus $3.6 million annually in expense for a new OIS as recently as its 1999 GRC.

TURN opposes PG&E/ORA Agreement 7. TURN recommends that the Commission disallow recovery of the incremental costs of two of the upgrades. TURN does not dispute that the projects are necessary but contends that these two projects should have been incorporated into the new OIS that was capitalized in 1999. In particular, TURN argues that the expenditures to integrate the three separate computer systems ($0.8 million in hardware and $3.25 million in expense) should be disallowed on the grounds that this integration should have been designed into the new CIS system that was capitalized in 1999. TURN also recommends that the Commission direct PG&E to amortize the costs of linking the OIS to the mobile data terminals and enhancing the mapping associations in OIS over three and five years, respectively. TURN believes that PG&E's request for $1 million in capital and $2.45 million in expense to remedy the treatment of single customer outages that are dropped off the system after 30 minutes should be disallowed for the same reasons.

TURN does not oppose PG&E's proposed $3.05 million expense on software upgrades for the mobile data terminal units, but suggests that this expense should be averaged over 3 years because it is a one-time expense that should minimize inaccuracies and reduce restoration time. Similarly, TURN does not oppose PG&E's request for $7.38 million in expense to enhance the mapping associations within OIS, but recommends that this expense be averaged over 5 years consistent with the expected length of this effort and the amount of projected expenditures per year. Aglet opposes Agreement 7 for another reason. Aglet believes that approval of Agreement 7 would constitute retroactive ratemaking by allowing PG&E to record costs incurred in 2003 in a memorandum account that was not approved until after the costs were recorded.

In response to TURN's claim that the expense on software for the mobile data terminals should be recovered over a three-year period, PG&E admits that the upgrade in the mobile data terminals is a one-time activity, but argues that as operational requirements change or are added, PG&E must continue to maintain the mobile data terminals, and upgrade the software. PG&E does not describe the nature of any ongoing activities or provide an estimate of the expense associated with any ongoing activities. Since PG&E admits that the software upgrade to link the mobile data terminals to the OIS is a one-time activity, we agree with TURN that PG&E's request should be averaged over three years. We modify Agreement 7 to require that the revenue requirement associated with upgrading the software to link the mobile data terminals to the OIS be amortized over three years. We note that PG&E has already revised its testimony to reduce the cost of enhancing the mapping associations within OIS and to amortize the cost over four years, and we approve that request.

In response to TURN's criticism that the integration project should have been built into the new OIS approved in 1999, PG&E states that it was not possible to integrate its three separate computer systems at that time due to limits in hardware and software that made the mapping rectification impossible. PG&E also states that it did not have a common SCADA platform available until 2002.

PG&E states that the OIS software that addresses single customer outages was originally configured such that during normal conditions, single customer outage calls or "tags" would be sent to the FAS to be individually handled. During storms, the process would change and PG&E's Operation Emergency Centers would manage these single customer outages via the OIS through an aged off report and personal callbacks to customers. ORA reports that in November 2002, the OIS software was upgraded.

PG&E states that it turned off the aging script at some point during the December 2002 storms. According to PG&E, this resulted in the OIS being "flooded" with single customer outages, slowing down the system and making the dispatcher's job more difficult. PG&E states that it took this step because when the system "aged off" the single customer outage, the recorded history of that outage was deleted by the system. This resulted in some customers calling to get an update on service restoration and being informed that PG&E had no record of their earlier call, as well as a delay in restoration of service.

PG&E states that the software configuration that caused the deletion of the single customer outages has since been resolved. PG&E also states that it intends to eliminate the "aged off" feature and that software will be developed to assist in better managing single customer outages more efficiently. PG&E states that at the time the OIS and the FAS were configured, PG&E's outage restoration process for single customer outages was handled most efficiently through the FAS rather than through OIS. However, PG&E has now determined, based on its experience during the December 2002 storms, that modifying OIS will be more effective than continuing to use FAS during major events. PG&E indicates that software modifications will be necessary to accomplish this upgrade, and provides an estimate of the cost associated with this upgrade, but does not provide any detail regarding the assumptions or criteria used in developing the cost estimate.

We do not believe that ratepayers should be required to fund the same OIS functionality twice. When the new OIS was funded in the 1999 GRC, the Commission justifiably anticipated that these system improvements would improve PG&E's outage response, including its response to single customer outages. PG&E states that it did not design the OIS system to handle single customer outages during storm conditions, and that the FAS was intended to deal with single customer outages. In response to TURN's suggestion that PG&E not be granted funding for these upgrades because they should have been incorporated into the OIS system that was funded beginning in 1999, PG&E simply states that the technology available in 1999 was not conducive to incorporating single customer outages in a system such as PG&E's OIS.

PG&E does not provide sufficient detail to allow us to evaluate the veracity of its claim, but we need not address whether such technology was available or not to find that PG&E's handling of single customer outages was unacceptable. When the FAS and new OIS were approved and funded in the 1996 and 1999 GRCs, PG&E designed the systems such that single customer outages during storms were managed through the FAS. During the December storms, that method was not successful. Customer outage calls were aged off and, following an upgrade to the OIS, deleted from the system as a result of an oversight on the part of PG&E. Had the outage reports not been deleted, presumably there would have been an accurate list from which the dispatchers would work to identify the problems, and restore service. While the nature of the storms made the problem more severe, it did not cause the problem.

In D.96-09-045, we found that meeting minimum system reliability measurements is not an adequate defense for failure to communicate with customers effectively during an emergency. Despite having completely overhauled its OIS system and spending at least $34 million of ratepayer funds in the process, not to mention the funds spent on the FAS, PG&E has still not managed to meet customer's expectations with respect to outage communications. It is unreasonable for PG&E to have allowed the system to function in such a manner as to allow single customer outages to go unrecorded and unresolved. We agree with TURN that PG&E should not be authorized to recover the cost associated with revising its treatment of single customer outages in this OIS since that functionality should have been incorporated into either the original OIS system or the FAS, both of which have already been fully funded by ratepayers. Therefore, we modify Agreement 7 to remove funding for the single customer outage issue.

In response to Aglet's claim that Agreement 7 would result in retroactive ratemaking, PG&E notes that in order to avoid the prohibition against retroactive ratemaking when impractical or inequitable circumstances arise, the Commission has authorized utilities to establish memorandum accounts in advance of the final decision that would determine the rates in questions and has issued "interim relief" decisions making the rates to be established in a later, final decision, effective as of a date earlier than the final decision. PG&E points out that the Commission adopted a similar process in the proceeding and that Conclusion of Law 2 in D.02-12-073 states: "the Commission intends that any changes to PG&E's gas and electric revenue requirements adopted in PG&E's TY 2003 GRC will become effective January 1, 2003." Ordering Paragraph 2 states "To the extent that, upon further order in PG&E's TY 2003 GRC, the Commission authorizes revisions to PG&E's authorize revenue requirements for the 2003 TY, such authorization may be made effective January 1, 2003.

PG&E states that its March 17, 2003 supplemental testimony supplemented not only the description of PG&E's operations contained in the original GRC application, but also the revenue requirement associated with such operations in response to the February 13, 2003 ACR. PG&E believes that the Commission's finding in D.02-12-073 that final revenue requirements adopted in this case would become effective as of January 1, 2003, applies equally to the revenue requirement requested in PG&E's supplemental testimony.

We agree. We understand Aglet's concern that the Commission cannot approve costs recorded by the utilities prior to an order or the adoption of a memorandum account that authorizes such recovery, but that is not what is contemplated here. The memorandum account proposed in Agreement 7 would be for the purpose of tracking the costs associated with specific upgrades to PG&E's OIS in excess of PG&E's recorded costs to allow the Commission to segregate the reliability-related costs adopted as part of PG&E's base revenue requirement from the costs associated with the OIS upgrades. In this instance, we are not tracking PG&E's actual 2003 expenditures for purposes of approving them separately from the PG&E's base revenue requirement; instead we are tracking the difference between authorized spending and actual spending to ensure that the cost of the OIS upgrades is not recoverable unless PG&E's actual expenses in FERC Account 588 exceed 2003 adopted FERC Account 588 revenue requirement by the amount that actual OIS upgrade expenses exceed adopted revenue requirement up to the amount in the memorandum account. Given the findings in D.02-12-073, and the purpose of the memorandum account, we concur that the memorandum account does not constitute retroactive ratemaking.

We modify Agreement 7 to approve the requested memorandum account and adjust the amount approved for recovery. Specifically, we adjust the amount eligible for recovery to reflect our findings that: 1) the cost for the mobile data terminals should be amortized over three years to reflect that it is a one-time project, and 2) the cost associated with addressing the single customer outages problem should be denied.

PG&E's initial request for the four OIS upgrades included an additional $3.050 million in expense to link OIS to the mobile data terminals, an additional $3.250 million in expense to integrate three computer systems, an additional $2.45 million in expense to retain single customer outages in OIS, and an additional $460,000 in expense in 2003 and $2.3 million in each of the years 2004, 2005, and 2006 (for a total project cost of $7.360) million for enhancing OIS mapping associations (all expressed in nominal 2003 SAP dollars). PG&E also initially requested $1.8 million in common utility plant.

Agreement 7 would provide for recovery of $9 million in expense in 2003 and $2.3 million in expense in 2004, 2005, and 2006 (nominal 2003 SAP dollars). We approve the requested memorandum account as an acceptable mechanism under which PG&E would be allowed to recover the costs of the OIS upgrades, but we will reduce PG&E's request to eliminate duplicative funding to address single customer outages. We also adjust the total to reflect that the expense associated with linking OIS to the mobile data terminals should be amortized over three years because it is a one-time project.

7.3 Other ORA/PG&E Agreements

a. Agreement 1 - Division Level Benchmarks

Agreement 1 requires PG&E to record and report reliability performance at the division level in addition to the system level. We find that ORA/PG&E Agreement 1 will resolve the concern expressed by ORA and Aglet about system level indices masking poor performance at the division level. Although Agreement 1 provides that PG&E, in consultation with ORA, would develop the format for reporting division data, we direct that, at a minimum, the reporting requirements should be the same as those adopted in D.96-09-045 for system level reliability performance information. Therefore, the report filed annually pursuant to D.96-09-045 must include data detailing the division level average interruption frequency, division level average interruption duration, division level customer average interruption duration, and division level average momentary interruption frequency.

We find that Aglet's request that we adopt division level reliability measures as the primary measure of reliability performance is unnecessary at this time. Receiving division level data will provide us with additional information upon which to evaluate the PG&E's performance to determine whether such a designation is necessary. Furthermore, as we stated in
D.96-09-045, meeting system performance levels is not a shield that can stave off liability for unreasonable performance at other levels or in other areas, such as outage communications.

b. Agreement 2 - Five-Year Average Benchmarks

Agreement 2 would require PG&E to investigate and report to the Commission when the adopted reliability performance measures vary by ten percent or more in any division and/or five percent or more at the system level from the five-year rolling average of reliability performance.

Agreement 2 is reasonable and should be adopted. We emphasize that approval of this Agreement does not in any way limit our ability to conduct other investigations or direct additional reports as we deem necessary.

Although Aglet argues that using rolling averages to measure performance tends to reduce the variability of performance metrics, we believe this Agreement will assist in improving PG&E's performance by requiring automatic investigations and reporting if division-level performance decreases significantly.

c. Agreement 3 - Definition of Major Outage

In Agreement 3, PG&E and ORA recommend that the Commission initiate statewide workshops to address the definitions of Excludable Major Event, Major Outage, and Measured Event, as well as the restoration performance standard included in Standard 12 of G.O. 166. Agreement 3 also provides a list of topics to be considered in the workshops.

Agreement 3 stems from ORA's concern that PG&E and other utilities may not understand some of the requirements and definitions within D.96-09-045 and GO 166, or are interpreting them such that the Commission may not be receiving the uniform reliability data it desires. ORA recommended that the Commission establish a process to eliminate the various conflicts within D.96-09-045 and GO 166 and make the definitions uniform. ORA points out that while GO 166 clearly defines a Major Outage as a situation where 10% of PG&E's customer simultaneously experience an outage, D.96-09-045 defines an Excludable Major Event as an event that affects 10% of its customers (with no mention of simultaneous or cumulative) or 15% of its facilities. ORA also notes that while GO166 clearly defines how start and end times for a Measured Event are to be determined, D.96-09-045 does not define how start and end times should be determined for Excludable Major Events.

As we stated in D.96-09-045, uniform and measurable system standards are an important first step in defining reliability. The December 2002 storms were a significant event for PG&E, and from the perspective of the Commission, and consumers, clearly constituted a major event on PG&E's system even if they did not meet the technical definition of a major event. We agree with ORA that common sense dictates that either an event is major, subject to the standards set for performance during major events and excludable from any performance indices as an abnormal and infrequent event, or it is minor, in which case it should be considered part of a normal weather pattern, included in the performance indices, and subject to any performance standards applicable to normal weather. Currently, neither the Emergency Response Standards, nor the standards applicable to performance during normal conditions apply to the December 2002 storm event. In addition, we believe it is necessary to ensure a consistent understanding of the terms between or among utilities. We find that the workshop process proposed in Agreement 3 will assist in ensuring that the utilities consistently interpret and apply the requirements in D.96-09-045 and GO 166. Although it is currently extremely difficult to compare reliability performance between two or more utilities, a consistent interpretation and application of our standards and rules is a necessary first step in this direction. For these reasons, we will adopt Agreement 3.

d. Agreement 4 - Tap Fuse Installation Program

ORA's initial testimony supported PG&E's proposal to accelerate the installation of overhead lines fuses but opposed PG&E's request for additional funding. ORA argued that sufficient funding for this program is provided in PG&E's base TY 2003 GRC revenue requirement request.

Agreement 4 provides that PG&E will install as many additional sets of overhead fuses in 2003 to fully utilize the GRC requested amount of $5.4 million in MWC 49. Under Agreement 4, this is expected to result in the installation of no fewer than 2,000 overhead fuses in 2003.

Agreement 4 would result in the installation of additional fuses on an accelerated time frame, thereby improving system reliability. For these reasons, Agreement 4 should be approved.

e. Agreement 5 - Balancing Length of Outages

Agreement 5 states that PG&E will modify its current restoration practices to balance the length of outages with the number of customers affected and will keep ORA actively involved and informed in the process of developing this policy. PG&E's stated that, at the time of the December 2002 storms, its policy was to restore the largest number of customers first, irrespective of how long some customers were out. Due to the succession of storms that occurred in December 2002, customers in areas with small numbers of customers out of service as a result of the storm on Saturday who were due to the have their service restored on Monday or Tuesday were pushed to the back of the prioritization queue after another storm resulted in outages to a higher number of customers elsewhere on PG&E's system.

ORA argues that PG&E's restoration performance during the 2002 storms would have been more effective if PG&E had a clear policy to ensure that no customers are left without service for an inordinate amount of time. ORA's Consultants, Stone and Webster, note that there is no indication as to the longest interruption duration experienced by a PG&E customer; therefore, the magnitude of the problem is not clear. ORA's Consultants recommend that the Commission direct PG&E to track and report any customers that experienced outage durations longer than 24 hours in one-hour increments and request specific explanations for lengthy outages. ORA Consultants also recommend that the Commission consider an additional measure for restoration performance such as the number of customer that had an individual outage duration of more than 200% of the overall storm CAIDI or in excess of a particular time limit such as 24 hours.

Under Agreement 5, PG&E would modify its restoration policy to attempt to balance the outage duration with the number of customers. PG&E states that it will have employees dedicated to small numbers of customers to restore them if their service has been out for an inordinate amount of time. Agreement 5 does not address the latter two recommendations by ORA's Consultants, or define the amount of time considered "inordinate," but it does provide for active involvement by ORA in the development of the policy.

We will approve Agreement 5 because it will reduce the potential for customers experiencing unusually long outages in the future by balancing the length of outages with the numbers of customers affected. In addition, although the latter two recommendations offered by ORA's Consultants are not included in Agreement 5, we find that these recommendations have merit and should be considered as well. We direct PG&E to track and report any customers that experienced outage durations longer than 24 hours in one-hour increments and provide specific explanations for outages longer than 48 hours. We also direct that, during the workshops to be conducted pursuant to PG&E/ORA Agreement 3, parties should develop an additional measure for restoration performance, such as the number of customers that had an individual outage duration of more than 200% of the overall storm CAIDI or in excess of a particular time limit, such as 24 hours.

f. Agreements 8 and 9

According to Agreement 8, PG&E will monitor and report to ORA on its implementation of the existing measures in its action plans (the improvement initiatives) and well as its investigations into additional technical measures to improve the accuracy of its VRU systems and potential methods to prevent its Safety Net line from being overburdened during high-call volume emergencies. Under Agreement 9 PG&E and ORA agree on a mutual approach to monitoring and reporting to ORA on any needed adjustments to its OIS, Customer Information System, FAS, VRU, and all customer interface and response systems that would aid PG&E in making resource deployments to address outages.

Agreements 8 and 9 respond to recommendations made by ORA. We recognize that ORA's testimony did not request specific action from PG&E or the Commission on these issues, and only requested that PG&E take its concerns into consideration. No other party commented on Agreements 8 and 9. Agreements 8 and 9 will allow PG&E and ORA to work cooperatively to identify potential improvements to PG&E's OIS, Customer Information System, FAS, and VRU and should be approved.

7.4 Performance Incentive Mechanisms and Metrics

a. Reliability Mechanisms

Various reliability incentive mechanisms were proposed for PG&E that would establish specific performance targets, provide for performance incentive payments and penalties, and include an incremental annual revenue requirement. Subsequent to the evidentiary hearings in this phase of the proceeding, CUE and PG&E filed joint testimony recommending approval of a modified version of the original CUE proposal as described in Section 6.1 above. To summarize, the PG&E/CUE proposal include the following elements:

· The term of the proposal is six years, 2004 through 2009.

· The performance metrics are the SAIDI and the SAIFI;

· The SAIDI target in terms of minutes per customer, by year, is: 171, 164, 158, 151,151,151;

· The SAIFI target in terms of interruptions per customer, by year is 1.42, 1.33, 1.24, 1.16, 1.16, 1.16;

· The maximum annual reward and penalty is $31.6 million for SAIDI and $10 million for SAIFI;

· The Deadband is 4.2 minutes per year for SAIDI and 0.05 outages per year for SAIFI on either side of the target;

· The Liveband is 15.8 minutes per year for SAIDI and 0.10 outages per year for SAIFI, for livebands on each side of the targets; and

· An Additional Revenue Requirement of $27 million annually for six years to be used for reliability improvement expenditures, subject to balancing account treatment;

· Any revenues not spent by PG&E in 2004, 2005, and 2006 would be carried over to the following years, up to 2007. Unspent revenues at the end of each year 2007, 2008, 2009 would be credited back to ratepayers at the end of each year.

CUE believes that PG&E's reliability needs improvement and argues that by combining incremental revenue with an incentive mechanism, the PG&E/CUE proposal would encourage improvements in PG&E's reliability, and ensure that PG&E's customers do not pay for improved reliability unless they get it. CUE compares PG&E's performance measures to SCE's and SDG&E's performance measures and suggests that PG&E's performance is much worse than the other two utilities. CUE also asserts that reliability should be improved because PG&E's average restoration time, as measured by the CAIDI, is increasing. CUE compares PG&E's current field staffing levels with prior PG&E field staffing levels and argues that decreases in customers per electric field personnel correlate to increases in outage restoration time. CUE suggests that PG&E be given financial incentives to maintain and improve reliability structured like the incentives already in place for SCE and SDG&E.

We will approve some aspects of the incentive mechanism as proposed in the joint proposal by CUE/PG&E but decline other aspects of the join proposal. We make some modifications as described below. We believe that this approach best balances our desire to encourage improvements in system reliability in addition to those the company intents to pursue through its adopted revenue requirement while providing the financial incentive to do so. Further, with the modifications as described below, we establish our intention that PG&E have a set of conventional statements about our expectations to continually improve reliability. We are interested in encouraging further improvements in the safety and quality of PG&E's system reliability as adopted in PG&E's revenue requirement in D.04-05-055. The incentive mechanism we establish in today's order provides for an opportunity to improve on the reliability metrics. We reiterate our expectation that PG&E's managers will give the proper attention to this area of performance. We note that the commission has various regulatory mechanisms for addressing deviations from expected levels of service. Most include after-the-fact performance reviews such as penalties, disallowances, and ratemaking enhancements. A properly structured incentive program that is layered on top of cost-of-service ratemaking as we intend to practice it can be a targeted regulatory mechanism that promotes improvements and discourages retrogressions in system reliability and performance. However, we share parties' concerns that the Joint Proposal may overly burden ratepayers with rewards to the company for improvements it expects to achieve absent the Joint Proposal. As such, we make further modifications to refine the incentive mechanism proposed by PG&E/CUE to properly overlay it to a cost of service regime.

CUE places much emphasis on the fact that PG&E's performance measures appear worse than the other utilities. However, all of the parties in this case acknowledge that it is difficult to compare indices between and among utilities due to significant differences in system design, geography, weather patterns, and measurement methods. As PG&E points out, even CUE has previously testified that an individual utility's "numbers can't be meaningfully compared to other utilities because other utilities do not use the same method of calculating SAIDI (and also because of weather and geographical differences."²³ PG&E also notes that CUE previously characterized SCE's attempt to compare its SAIDI performance to 43 other utilities as "not meaningful."²⁴ PG&E states that as measured by the absolute values of SAIDI and SAIFI, without further analysis or understanding of the differences in outage reporting methods, CUE's assertion that PG&E's performance is worse is technically accurate, but not meaningful. PG&E states that it calculated PG&E SAIDI and SAIFI values using its understanding of the methodology used by SCE to measure outage duration and frequency and found that using SCE's methodology, PG&E's SAIDI value for the last five years would improve (decrease) by 21 percent, and PG&E's average SAIFI value for the same time period would improve (decrease) by 12 percent. Although PG&E is not certain that it exactly mimicked SCE's methodology, it is certain the measurement method used affects the results.²⁵ In addition, CUE and PG&E admit that SCE's PBR mechanism has a different definition of an excluded event than PG&E uses.

Based simply on the annual SAIDI, SAIFI and MAIFI performance reported by PG&E, we cannot find that the mere presence of a difference between PG&E's performance results and SCE's or SDG&E's performance results justifies a change in performance standards. PG&E's service may indeed be less reliable that the other two utilities, or it may simply appear less reliable due to system differences or different methods used to calculate SAIDI and SAIFI.

CUE also states that it compared PG&E's SAIDI, SAIFI, and CAIDI data to the SAIDI, SAIFI and CAIDI data for six New York State utilities, the City of Redding, Pacificorp, and Portland General Electric and, in each case, found that PG&E's results reflected lower reliability. Again, however, we cannot effectively compare the summary reliability data provided for these other utilities without also comparing any differences in measurement techniques, weather, geography, and system size and design. CUE did not provide the information necessary to make such a comparison. As PG&E notes, although individual utilities may have certain similarities, other differences may exist that would explain the differing results. We agree with PG&E that inter-utility comparisons are of limited value because different utilities measure different things, serve different customers mixes, and experience different weather. Based on the record in this proceeding, using the performance indices alone, we cannot find, as CUE claims, that PG&E's service is significantly worse than SDG&E's and SCE's, due to the different methods of calculating the inputs that the indices are derived from.

As we have found in previous decisions, it is not particularly useful to compare utilities with different customer counts, different geography and weather patterns, different system configurations, not to mention different methods of calculating SAIDI, SAIFI, and MAIFI. Given these factors, it is extremely unlikely that any two utilities would ever achieve similar performance results; therefore, we are reluctant to place much faith in such comparisons. We believe that the more appropriate comparison to make is a comparison between PG&E's historical performance and its current performance.

In D.00-02-046 we stated that PG&E's performance during the 1996-1998 period was reasonable, consistent with historically accepted levels of service. In this phase of the proceeding, we are conducting a high level review of PG&E's performance during the period from 1999 through 2002, including, but not limited to, the December 2002 storms. We find that PG&E's overall reliability performance during normal conditions, as measured by SAIDI, SAIFI, and MAIFI²⁶ performance reported in PG&E's annual reliability reports, has either improved relative to 1996-1998 levels or has remained consistent with 1996-1998 levels.

The data provided in Table 2-4 and Table 2-5 of PG&E's Exhibit 13 on the number of customer interruptions demonstrates an increase in reliability compared to the 1996-1998 period referred to in D.00-02-046. PG&E's three-year SAIFI average (excluding major events) decreased 13 percent from 1.66 during 1996 through 1998, down to 1.44 from 1999 through 2001. PG&E's four-year SAIFI average from 1999-2002 was 1.36. PG&E's 2002 SAIFI average was even lower, at 1.11.

Including major events, the trend is the same. PG&E's three-year SAIFI average from 1996 through 1998 was 2.10, decreasing to 1.48 in 1999 through 2001. PG&E's four-year SAIFI average from 1999-2002 was 1.53.

The SAIDI averages have fluctuated from year to year and do not demonstrate a trend in either direction. PG&E's three-year SAIDI average (excluding major events) was 173.3 from 1996 through 1998, while PG&E's three-year SAIDI average from 1999 through 2001 was 178.8. PG&E's four-year SAIDI average from 1999-2002 was 168.9. Including major events, PG&E's three-year SAIDI average from 1996-1998 is 278.4, while the three year average from 1999-2001 is 191.6. PG&E's four-year SAIDI average including major events is 239.2.

PG&E's CAIDI increased from 103.67 during 1996-1998 to 124.25 from 1999-2002.

A key assumption underlying the PG&E/CUE reliability proposal is that the Commission and PG&E's customers desire additional reliability above and beyond the levels already established in previous Commission decisions. PG&E and CUE state that the public response to the December 2002 storms is evidence that customers desire additional reliability. PG&E and CUE also view the issuance of the ACR as an indication that the Commission desires an improved level of service to customers. Other than that, the record is very limited on PG&E's customer's desires.

Although PG&E and CUE argue that the values of service in G12003 show that PG&E customers are willing to pay for additional reliability, we reject this argument for the same reasons we reject ORA/PG&E Agreement 6; we do not base our analysis of whether a performance incentive mechanism is needed on the outdated VOS information. We also reject CUE's claim that the billions of dollars of Department of Water Resources contracts executed by the State of California in 2001 were directed toward improving SAIDI and can be used as a proxy for current customer value of service data. Therefore, we must evaluate the reliability incentive mechanisms proposed on a different basis.

Another major premise of the PG&E/CUE proposal (as well as the original CUE proposal) is that substantial increases in funding are necessary for PG&E to improve reliability performance. PG&E argues that in order to significantly improve reliability, it requires additional revenues over its forecast in the base GRC case. PG&E states it has requested funding in the GRC for only one program intended to improve reliability: the overhead tap fuse installation program contained within MWC 49, which is the subject of Agreement 4 with ORA. PG&E asserts that the forecasts associated with the other MWCs are intended to maintain the adequate service level identified in D.00-02-046 and that in order to have an opportunity to meet the targets in the PG&E/CUE incentive mechanism, PG&E requires additional revenues.

ORA, TURN and Aglet argue that there is no evidence that increasing reliability requires additional funding beyond the levels requested by PG&E in its base GRC. In its Opening Brief, TURN compares the PG&E/CUE performance targets with PG&E's expected reliability performance absent any new funding. TURN's forecast assumes the base level of future SAIDI performance is the 1998-2002 SAIDI average of 171.1. TURN then calculates the expected impact of PG&E's overhead tap fuse installation program on SAIDI and SAIFI levels. According to PG&E, the overhead tap fuse program is expected to result in a total savings equal to 10% of the 1997-2000 SAIDI average or 16.7 minutes of improvement in SAIDI. (10% of 166.6 minutes equals 16.7).²⁷ The program began in 2002, and according to PG&E, should yield benefits of 2 % improvement per year, or 3.33 minutes in SAIDI per year over the course of the five-year program.²⁸ TURN then forecasts SAIFI using PG&E witness Blastic's estimate that the tap program would result in a reduction of 0.1 outages by 2007.²⁹ The resulting TURN forecast is shown in the table below.

TABLE 5

ORA offers a similar comparison starting with the SAIDI average prior to the implementation of the tap fuse program. Using 166.6 minutes as the starting SAIDI value (the average SAIDI value from 1997-2000, excluding major events), and assuming PG&E reaches its goal of a two percent reduction annually in SAIDI to reach a 10 percent total reduction, ORA calculates that PG&E would expect to achieve a SAIDI value of approximately 150 minutes in 2006. Therefore, without any incremental revenue or an incentive mechanism, PG&E would surpass the proposal's 2006 performance target of 158 minutes by approximately 8 minutes.

According to ORA's calculation, under the PG&E/CUE proposal, PG&E would receive $81 million in incremental revenue from 2004 through 2006 ($27 million X 3 years) and collect a total reward of $65.2 million ($28.8 million in 2004, $20.2 million in 2005 and $16.2 million in 2006) by meeting a SAIDI target that the company intends to meet absent the proposal.

Taking even more data into consideration by using the five-year average of annual SAIDI values from 1997-2001, ORA calculates that PG&E would expect to achieve a SAIDI of 158.1 minutes in 2006, equal to the 2006 target in the CUE/PG&E agreement.³¹ In this case, the PG&E/CUE proposal would provide PG&E $81 million in incremental revenue from 2004 to 2006 for achieving a target that it should achieve under its current plans without additional revenue.

As several parties have pointed out, with respect to the expected levels of improvement associated with the tap fuse installation program, PG&E witness Blastic's testimony in support of the joint PG&E/CUE proposal is inconsistent with the prior testimony of PG&E witness Camara. As noted above, while witness Camara's prepared testimony, written response to ORA's data request, and oral testimony in response to cross-examination questions describe the tap fuse installation program as a five-year program designed to result in a 2% improvement in SAIDI per year and a 10% reduction in SAIDI over the five year program, PG&E's Opening Brief states that the length of the program was modified by Witness Blastic on the stand and it is now expected to be a seven year program. PG&E further states that the expected reliability benefit of this program is now approximately a two minutes reduction in SAIDI per year, instead of a 2% reduction in SAIDI per year.³² PG&E does not provide a rationale or any supporting evidence for this change.

In evaluating the conflicting testimony, we give more weight to the estimate provided in PG&E's prepared testimony, primarily because the original estimate was provided by witness Camara in two separate documents, PG&E's prepared testimony and in a response to a data request from ORA. PG&E had an opportunity to review and modify its response through its rebuttal testimony and did not do so. Finally, we note that since the PG&E/ORA Agreement 5 results in a lower tap fuse cost estimate and a corresponding increase in the number of annual fuse installations, the reliability benefits should be expected to exceed the original estimate, not decrease it.

ORA also notes, as does TURN, that PG&E fails to account for any potential reliability improvements associated with MWC 08 and the improvement initiatives identified by PG&E in response to the outage communications problems encountered during the December 2002 storms. Despite PG&E's argument that its forecasts in its other MWCs are not adequate to provide measurable improvements in reliability, a close review of PG&E's testimony shows that it expects several activities to affect reliability.

PG&E states that the main activities it relies upon to directly improve reliability are contained in the "Dependability" program. The Dependability program includes Capital MWCs 08, 09, and 49, and Expense MWC HX. Although MWC 49 is the only category for which PG&E calculates specific reliability improvements, PG&E acknowledges that there are several other areas that directly affect reliability as well.

As TURN points out, PG&E's TY 2003 request in MWCs 08, 09, 49 and HX is $19.56 million (capital and expense combined). This represents a 43% increase over 2001 expenditures in these MWCs. In addition, PG&E's witness Bhattacharya admits that, assuming equivalent weather, granting PG&E's total request in the Dependability MWCs would lead to performance that is consistent with 2002 performance, and most likely better.³³

Witness Bhattacharya also states that "the vast majority of PG&E's expenditures are linked to reliability, even though reliability is not the main driver for the work."³⁴ As an example, witness Bhattacharya explains that distribution capacity work can improve reliability by providing additional capability for use during emergency switching, or by providing additional circuit ties for improved restoration switching. Witness Bhattacharya also reports that customer connection work can improve reliability as crews perform minor corrective maintenance in conjunction with installing facilities to serve new customers. We find that PG&E's expenditures in those areas should be expected to maintain and potentially improve reliability, whether measured by SAIFI, as in the case of pole replacements, or SAIDI, as in the case of PG&E's OIS improvements.

We also find that the improvements that PG&E intends to make to its OIS and its call center performance are likely to improve PG&E's reliability performance as well. For example, enhancing the mapping associations within the OIS so that smaller portions of PG&E's circuitry can be pinpointed may reduce outage duration by providing PG&E with more accurate information regarding damaged facilities and the repair personnel and equipment necessary. More accurate information regarding the number of customers affected by outages will allow PG&E to focus its repair efforts. Similarly, utilizing mobile data terminals to accelerate the input of outage causes and damage assessment information to the Operations Emergency Centers should also have a positive effect on outage duration. Integrating three of the company's internal information control systems may also have a positive effect on reliability by reducing the number of entries required by a system operator.

Comparing TURN's estimated PG&E performance without any additional funding beyond that requested in PG&E's GRC application to the PG&E/CUE target performance levels, we find that, if it pursues the programs for which it sought funding in its application, PG&E will be able to meet and exceed the target levels without incremental funding or incentives. Furthermore, since the joint PG&E/CUE proposal contains deadbands of 4.2 minutes per year for SAIDI and .20 outages per year for SAIFI, it is highly likely that PG&E will receive incentive payments each year for its performance even without any additional effort on the part of PG&E.

Although we recognize that reliability improvements cannot be predicted with 100% precision, on balance we are persuaded that TURN and ORA's estimates of PG&E's future performance are more reasonable than PG&E's forecast. The fact that TURN and ORA's estimates are based on PG&E's own testimony adds weight to our conviction. In our judgment, while neither the TURN nor ORA estimates are likely to match precisely the year-to-year performance levels, they reflect a reasonable expectation of future performance.

ORA, TURN and Aglet also express concern that PG&E and CUE have offered no indication of the costs of reliability improvement apart from the tap fuse installation program. When asked to justify the need for an additional $27 million annual revenue requirement to achieve the target levels, CUE admitted that the number was not based on the estimated cost to PG&E to achieve the levels, because CUE "does not know what the cost might be." CUE also states that "there is no analysis of the precise amount. It could be more, it could be less."³⁵

Given the absence of any evidence supporting the need for this increment of funds, ORA, TURN and Aglet argue that the Commission should reject the proposed incentive funding mechanism. We agree. Given the discussion above, we find insufficient analytical support for the requested funding of $27 million per year for six years to achieve the reliability targets. Neither CUE nor PG&E has provided any analysis to support this amount of incremental funding other than claiming that the $27 million somehow equates to the value of the benefits to ratepayers if PG&E meets the new performance targets. We find this unpersuasive, particularly in light of the detailed analysis provided by PG&E in support of its tap fuse installation program. As such, we decline to adopt the portion of the CUE/PG&E Joint Proposal which authorizes additional annual revenue requirement for improvements in reliability metrics.

While PG&E forecasts $5.4 million per year for five years to fund a tap fuse installation program designed to achieve a 16.7 minute reduction in SAIDI over five years, the PG&E/CUE proposal would have us approve an incremental revenue requirement of $27 million annually for six years to allow PG&E to meet reliability performance targets that reflect, at best, a 20 minute reduction in SAIDI when compared to the 1997-2000 performance. Compared to PG&E's SAIDI performance in 2002, the PG&E/CUE proposal would have ratepayers funding an incremental revenue requirement of $27 million per year to meet performance targets that reflect worse performance than that achieved in 2002.

CUE asks us to find that the rate impact of the PG&E/CUE proposal is minimal. We cannot make such a finding. As TURN points out, the PG&E/CUE assertion that the maximum rate impact of its reliability incentive mechanism will be less than 0.08 cents per kWh is based on the assumption that costs will be allocated on an equal cents per kilowatt hour basis across all customer classes. TURN points out that this assumption is not consistent with the Commission's past cost allocation decisions. TURN calculates that a $68 million increase represents a 15% increase compared to PG&E's electronic distribution revenue requirement request of $447 million and could result in a total average distribution rate increase of almost 23%.

Furthermore, as Aglet points out, the PG&E/CUE proposal's claim that there is a direct relationship between additional revenues and reliability improvements is disputed by PG&E's prepared testimony that asserted no linkage between spending and specific reliability improvements.

Another critical assumption of the CUE/PG&E proposal is that the $27.5 million allocated annually to the Reliability Improvement Memorandum Account (RIMA) would facilitate initiatives that are: 1) incremental and 2) spent on reliability related improvements. According to the PG&E/CUE joint testimony, the RIMA would be a one-way balancing account designed to separately identify amounts adopted in PG&E's TY 2003 GRC and the incremental expenditures incurred in planning and implementing distribution system reliability improvement activities funded by the $27 million annual incremental revenue. Although PG&E claims that the proposal does not grant PG&E an "additional revenue requirement," because recovery is not guaranteed, the authorized $27 million in additional revenue will be credited to the RIMA each year and any potential refunds would not occur until after the year 2007 at the earliest.

The RIMA would record incremental distribution expense expenditures associated with MWCs HX, NEW 1 and NEW 2. Although PG&E would be required to demonstrate expenditures equal to GRC-approved base levels in these three accounts, the proposal does not guarantee that RIMA projects would not supplant the funding represented by base GRC levels, because MWC NEW 1 and MWC NEW 2 do not currently exist and there would be no GRC approved base level of expenditures for these two MWCs. PG&E admits that since these two MWCs do not exist, there would be no requirement that these two MWCs be fully utilized in order to identify expenses as "incremental" to approved base levels. Furthermore, CUE admits that the burden would be on other parties to show that expenditures were not "incremental" and that if PG&E "managed to fool you ... and convinced the Commission that it was an incremental expense, then it would be allowed," i.e. PG&E would be allowed to use incentive funds for non-incremental expenses.³⁶

While the RIMA would not be triggered until PG&E exceeds its GRC-approved base level spending forecast for MWC HX activities, it does not limit RIMA eligibility to activities that would otherwise be classified in MWC HX. This is also true for capital expenditures, if PG&E exceeds the GRC-approved base level spending forecasts for MWCs 08, 09, and 49, any other "reliability" expenditure could be classified as incremental regardless of whether the approved funding in the MWC normally applicable to that activity has been fully utilized. We agree with TURN and Aglet that this could result in PG&E redirecting existing resources towards reliability improvements in order to achieve incentives or simply reclassifying existing activities as reliability programs in order to access the additional $27 million.

Moreover, PG&E and CUE admit that any challenge to expenditures claimed to be reliability related would require affirmative review and action by other parties, like ORA and TURN. CUE admits that "if [the other parties] just sit there and do nothing and everybody sits there and does nothing then presumably it would get approved."³⁷ Since there is neither a definition of reliability related activities nor a definition of expenditures that are specifically ineligible for inclusion in this account, this burden would be extremely hard to meet. PG&E's witnesses acknowledge that many costs can be classified as "reliability related" and that it is hard to determine the difference between a reliability program and a non-reliability program.

Unlike the specific programs included in the vegetation management program adopted in PG&E's TY 1999 GRC, the definition of reliability programs to be included in the RIMA is vague, as a result, to the extent that PG&E manages to reclassify existing efforts from other MWCs as reliability programs, there remains the potential for capital expenditures to be shifted from other MWCs into the RIMA, rather than resulting in new activities designed to improve reliability. We agree with TURN that the RIMA would allow for the reallocation of program budgets to create the appearance of additional, reliability-related activities while permitting PG&E to substitute RIMA funds for base GRC funding levels.

Aglet is also concerned that the incentive mechanism would motivate PG&E to shift resources away from activities whose costs the Commission has allowed in rates in order to achieve incentive rewards. ORA shares Aglet's concern, and notes that through distribution rates, ratepayers already pay PG&E a rate of return on its electric distribution plant investment, which presumably provides customer with highly reliable service consistent with statutory requirements. Since PG&E admits that a general feature of incentive mechanisms is that they can result in reallocation or redirection of resources from one area of a firm's operations toward the area that is the subject of the incentive, we agree with ORA and Aglet that the CUE/PG&E proposal has the potential to force customers to pay for the same level of service three times, first through the base revenue requirement, a second time through the incremental revenue requirement and a third time through incentive payments.

ORA and Aglet also expresses concern that if the Commission approves that PG&E/CUE proposal, it will be difficult to accurately measure the reliability improvements that would have occurred in the absence of the program, causing uncertainty in how to evaluate any reliability-related revenue requirement requests in PG&E's next GRC. The PG&E/CUE proposal would lock the Commission into a long-term commitment beyond the term of the current GRC cycle. With no requirement that PG&E identify particular projects or programs with associated reliability improvements, it will be difficult to assess the reasonableness of PG&E's revenue requirement request in the next GRC.

We appreciate the concerns raised by Aglet, ORA and TURN with respect to the Joint Proposal. However, we find there is value in adopting some aspects of the Joint Proposal as a means to encourage further improvements in system reliability on top of those the company expects to achieve through the projects identified in its revenue requirement request. Performance incentives such as those at issue here are a relatively recent regulatory phenomenon for California electric utilities. They were first adopted in connection with the Commission's 1990's experimentation with PBR as an alternative to conventional cost-of-service ratemaking. For example, in 1994, in D.94-08-023, the Commission adopted a PBR proposal for SDG&E that included performance incentives to "reward or penalize the utility's ability to control employee safety, system reliability, and customer satisfaction." (55 CPUC 2d 592, 633.) The Commission adopted performance incentives in connection with PBR mechanisms for SCE in 1996 (D.96-09-092; 68 CPUC 2d 275) and for SoCalGas in 1997 (D.97-07-054; 73 CPUC 2d 469). In D.99-09-030 the Commission adopted a second-generation PBR for SDG&E that included performance incentives. In adopting SCE's PBR, the Commission stated the following:

Thus, we see PBR as emulating the competitive process to encourage utility management to make decisions which resemble an efficient or competitive outcome. An efficient utility will control rates which benefits ratepayers. However, we want to ensure fairness to ratepayers, employees and shareholders in the PBR process. This requires balancing potentially conflicting interests. The utility can increase short run profits through reducing variable costs, but without revenue sharing such cost reductions will not lower rates. Moreover, such reductions not only can affect staff immediately but the service quality impact may only appear much later. [¶] In this PBR for Edison, we balance these interests by requiring a progressive sharing of net revenue between shareholders and ratepayers and by having both the productivity and service quality measures increase over the duration of the PBR. (68 CPUC 2d 275, 290.)

In adopting the SoCalGas PBR, the Commission observed that the utility had proposed a service quality mechanism "[i]n order to ensure that SoCal's focus on increased productivity through cost reductions does not have a deleterious effect upon the quality of service . . . during the period when the PBR rates are in effect." (73 CPUC 2d 469, 490.)

We have previously rejected performance-based ratemaking for energy utilities in favor a transparent regime of cost-based ratemaking. Consequently it may appear anomalous to borrow targeted performance incentives from the obsolete PBR regime. Performance incentives mechanisms may add complexity to regulation and ratemaking. They can be controversial, as evidenced by the extent of litigation and lack of agreement among the parties over how to construct them, even after two of the parties reached a settlement agreement. Moreover, even if they are reasonably calculated to promote a specific regulatory objective, performance incentives could work at cross-purposes with other regulatory objectives or have other unintended consequences. For example, incentives that are intended to improve reliability by reducing the number of circuit interruptions might be inappropriate, even if effective, if they added unduly to ratepayer costs. While some argue that improperly constructed incentives are worse than having none, others point out that such incentives can lead managers to make non-optimal resource allocation decisions.

Nevertheless, we will approve a set of targeted performance incentives that provide both rewards and penalties because they provide a more responsive approach to deviations in service adequacy and quality than our other ratemaking mechanisms. As noted above, they were adopted to mitigate the potentially deleterious effects that the efficiency gains sought through PBR could have on service quality, including reliability, and on safety. They can be carefully adapted to the cost-of-service regime and enhance our ability to regulate in the public interest, providing both financial incentives to guide utility activities and an early warning of longer-term trends that we can use to guide more intrusive regulatory interventions such as complaints and investigations. They represent a calibration, not a contradiction, of our cost-of-service principles.

In the cost-of-service regime we attempt to determine the reasonable costs of adequate service and establish rates that cover those costs, including a reasonable profit. This entails describing the public's expectations about service quality, including performance benchmarks and standards, and providing revenues sufficient to pay for those service levels. In the cost of service regime we aspire to transparent accountability for revenues and costs, so that we can determine both the causes of failure to meet expectations and the resources needed to provide incremental improvements, if necessary.

We will consider whether the proposed performance incentives are necessary for achieving one or more of our regulatory objectives and are likely to be cost-effective; we do not believe that performance incentives should be adopted solely on the basis of their mere consistency with a particular objective. Since rates set through our conventional approach to ratemaking are intended to provide the funding required to meet the regulatory objectives of safe and reliable service, we must ask why the utility needs the possibility of additional ratepayer funding, or threat of reduced funding, to get the utility to do what it is already funded and expected to do. The burden is on the proponents of performance incentives to prove they are necessary, cost-effective, and otherwise reasonable.

Performance incentives can be structured to carefully layer on top of a rigorous cost-of-service approach to T&D operation, maintenance and capital spending . The performance measures we adopt today accurately reflect current levels of utility performance in service reliability as measured by the frequency and duration of outage events (SAIDI and SAIFI). However, we find that further refinements to the proposed metrics are warranted. Deviations from these levels of performance are rewarded or punished, financially, as the deviations occur. We would expect that this focus on results measured more frequently than the General Rate Case permits will help PG&E sustain its efforts.

In proposing performance incentives for system reliability the utility exposes itself to the inference that it may not perform reasonably in those areas even with the funding to do so. However, we find that the evidence supports the proposition that PG&E has sufficient incentive to provide "adequate" service through the cost-based rates we have authorized in D.04-05-055.

The performance incentives proposed by PG&E/CUE are, at most, incremental to existing incentives to perform well. Further, a number of negative business consequences could ensue including threats of lawsuits; higher insurance premiums; bad publicity; loss of goodwill and other forms of corporate prestige; and, in the most extreme cases, loss of franchise or municipal takeover.

Even in the absence of additional incentives such as those under consideration, we can expect that PG&E's managers will pay a great deal of attention to the company's performance in each of the areas for which incentives have been proposed.

Ratemaking consequences attached to deviations from the established benchmarks should reinforce those incentives and lead to improved performance, accurately measured.

We are not pursuaded by PG&E and CUE that the proposed RIMA adequately prevents PG&E from shifting funding from other areas to dependability in order to access the $27 million, nor does it provide protection against PG&E spending the $27 million on projects that do not result in increased reliability. The risk is simply too high that customers will end up paying twice for the same reliability improvements as a result of the proposed incentive mechanism.

The original CUE proposal is not justified because it would burden ratepayers with an additional $72 million annual revenue requirement and up to $145 million in incentive payments without demonstrating that the reliability target is appropriate, that customers are willing to pay for that level of reliability, or that $72.5 million is the correct price to pay to achieve that level of reliability.

PG&E's alternative proposal also fails because it would result in PG&E receiving incentive payments for maintaining the average level of service it has achieved over the last five years, as measured by SAIDI and SAIFI. PG&E provides no justification as to why customers should pay more to fund incentive payments for service that the Commission has already deemed to be the reasonably expected level of service covered by base rates.

As we stated in D.00-02-046, a complete analysis of the reasonableness of expenditures to improve the integrated utility system would include an assessment of how the cost of improvement alternatives is weighed against the improvements in utility system infrastructure. Value of service analysis is the recognized tool by which such weighing takes place. As discussed above, we find that the existing value of service data is too dated to rely on. While customers were frustrated by PG&E's performance during the December 2002 storms, and may prefer improved performance, they may not be willing to bear the additional rate increases associated with an increase in reliability performance. Given the absence of relevant value of service studies or sufficient evidence demonstrating that ratepayers prefer, and are willing to pay for, increased reliability generally, we are reluctant to approve the increases in funding associated with the PG&E/CUE proposal. Furthermore, since the incentive mechanism is based on system-wide metrics, there is no guarantee that those customers facing the worst performance would see increased reliability.

Instead, we adopt a reliability incentive mechanism as shown in the following table. We decline to adopt a mechanism whose length exceeds the TY2003 cycle and we modify the target metrics as follows:

TABLE 6

b. Employee Safety Mechanisms

CUE proposes that the Commission adopt an employee safety mechanism to prevent PG&E from cutting corners on safety in order to save money. CUE argues that an employee safety mechanism is particularly important in the context of reliability performance incentives, where there is a direct financial incentive to restore service as quickly as possible. CUE recommends that the mechanism be based on the OSHA recordables frequency rate, with the benchmark set at 5.42, PG&E's most recently attained safety level.

PG&E opposes CUE's recommendation, arguing that PG&E already has a comprehensive program to promote safety and health and manage the incidence of injury and illness in the workplace. PG&E states that its program has yielded continued and sustained improvement in safety and health in recent years. PG&E also argues that changes in the OSHA regulations for reporting injury or illness that took effect on January 1, 2002, make it impossible to develop an OSHA recordables metric that accurately compares recent statistics with historical records in any meaningful way. PG&E states it now uses Lost Workdays as the basis for monitoring and evaluating its safety performance as opposed to the OSHA recordables rate, because it believes that lost work time is the most accurate measure of the severity of injuries or illnesses. PG&E argues that the Commission should not adopt CUE's proposed employee safety incentive mechanism because it relies on unreliable data to measure PG&E's safety performance.

CUE admits that PG&E's OSHA recordables rate has continually improved. PG&E's OSHA recordables rate has gone from a high in 1993 of 11.16 to 5.43 incidents in 2002. Given the fact that PG&E's employee safety performance has been consistently improving and we do not adopt a reliability performance incentive mechanism, we find that there is no need to adopt an employee safety incentive mechanism at this time.

c. Call Center Metrics

Under D.95-09-073, PG&E is required to maintain a monthly ASA of 20 seconds, with monthly busy signals a maximum of 1 percent during normal times and 3 percent during outages. PG&E met this standard from January 1998 to December 2002 with only two exceptions that occurred during the energy crisis. After the installation of its new CIS, however, PG&E has been unable to meet the ASA standard. PG&E requests that the Commission adopt a TSL standard instead of the ASA standard. PG&E notes that it is the only utility that is subject to an ASA standard and that its request to switch to a TSL standard is uncontested.

TURN does not oppose PG&E's request, as long as the new standard reflects the same level of service as the existing ASA standard. TURN states that an ASA standard of 20 seconds is equivalent to a TSL standard of 80/20, or 80 percent of the calls answered in 20 seconds. TURN notes that PG&E's current call center situation requires longer wait times for live customer service representatives and a longer time on the phone to handle business in comparison to the previous CIS system. TURN points out that PG&E's 2003 revenue requirement request includes funding to maintain the Commission's 20 second ASA standard. PG&E's call center request incorporates an ongoing increase of $4.63 million (nominal dollars) for additional labor required to compensate for the expected lower efficiency in the new CIS and some increased call volume.

TURN also notes that the ASA standard includes calls answered by PG&E's VRU. During normal conditions, PG&E's customer service representatives answer 60-70% of the calls. Under storm conditions, the VRU handles a larger proportion of calls, such that it is possible that the ASA remains at 20 seconds or less, but the wait time for a representative is much longer. For example, on December 16, 2002, the day of peak call volume of the December storms, the ASA was 12 seconds, with 96% of calls answered in 20 seconds, yet the maximum wait time for a service representative was 76 minutes, and 12% of calls to the customer service representatives were abandoned. Nevertheless, TURN does not recommend that the Commission apply a stringent call center standard on a daily basis for storm conditions at this time, because it would be too expensive.

We agree with TURN that any new standard adopted should reflect the same level of service as the old standard. Since the TSL standard of 80/20 or 80% of calls answered in 20 second is equivalent to an ASA standard of 20 seconds, we will approve PG&E's request and adopt a TSL standard of 80% in 20 seconds. However, since neither the ASA standard nor the TSL standard currently differentiates between the response time associated with calls answered by a service representative and the response time associated with calls answered by the VRU, we find that the statewide workshops to be instituted under PG&E/ORA Agreement 3, above, should address whether or not call center standards should be revised to better reflect the use of VRU. In particular, the workshop participants should recommend a standard of reasonableness of Average Handle Time in addition to an ASA or TSL standard.