V. Common Issues

A. Rate of Return

For estimating the earnings of the nuclear decommissioning trusts, PG&E estimates an 11.0% pre-tax return on equities and a 7.0% pre-tax return on its fixed income assets. SCE estimates a pre-tax return on equities of between 7.42% and 10.11%, and a pre-tax return on fixed income assets of between 4.21% and 6.03%. SDG&E estimates a pre-tax return on equities of 7.42%, and a pre-tax return on fixed income assets of 6.03%. ORA recommends a 12.5% pre-tax return on equities and a 7.4% pre-tax return on fixed income assets.

PG&E's equity return forecast is based on the annualized rate of return for the U.S. equity market over rolling 10-year periods covering 80 years, from 1920 through 2001. PG&E believes that forecasts of long-term market returns are traditionally based on historic market experience over very long time periods, and it is preferable to include more data points where available to decrease the variance in the results. In PG&E's last general rate case (D.00-02-046), the Commission adopted an 11.0% pre-tax return on equities. PG&E believes an 11.0% pre-tax return on equities remains a reasonable and conservative forecast. In D.00-02-046, the Commission also adopted a 7.0% pre-tax return on the fixed income portion of PG&E's trusts. PG&E recommends the same value in this proceeding.

SCE used two sets of return assumptions to establish a range of contributions to its decommissioning trust funds for San Onofre Nuclear Generating Station Units 2 and 3 (SONGS 2&3) and Palo Verde Nuclear Generating Station Units 1, 2, and 3 (Palo Verde). The first set of assumptions relies on DRI-WEFA (DRI)⁴ projections for: (1) the Standard & Poor's (S&P) 500 Stock Price Index, and (2) the dividend yield for the S&P 500 Stock Index, to calculate future equity returns. SCE maintains that when compared to estimates derived from historical data, DRI's Treasury bond yield projections are too high relative to their inflation projection, and DRI's estimate of future equity returns is too low. Therefore, it constructed an alternative set of return assumptions that adjust Treasury bond yield projections and future equity returns to reflect historical relationships. SCE believes that its two sets of return assumptions bound expected returns for the decommissioning trust funds.

SDG&E argues that it does not make sense to adopt identical rate of return assumptions for itself, SCE and PG&E because each company has its own separate and independent decommissioning trusts with portfolios of hundreds of different domestic and international stocks. Moreover, each company has different investment committees with different risk tolerances. As a result of these differences, the three utilities may choose different portfolio asset allocations, investment strategies, and investment advisors, all of which will impact the realized investment rates of return.

SDG&E used DRI projections as the basis for computing expected equity and fixed-income asset returns. SDG&E maintains that DRI forecasts should be consistently used in determining funding requirements. It believes that using DRI forecasts consistently over time provides the Commission with a consistent gauge to assess performance, and provides fewer opportunities for gaming that could occur if methodologies are changed every three years.

Specifically, DRI projects that the average annual pre-tax return for the S&P 500 and 10-year Treasury bonds will average 7.42% and 6.03%, respectively, from 2002 through 2026, which covers the period that contributions will be made (through 2013) to the decommissioning trusts.⁵ SDG&E says the DRI forecast is also consistent with equity projections from a variety of investment professionals.

ORA recommends a 12.5% pre-tax return on equities, and a 7.4% pre-tax return for fixed income investments. ORA's 12.5% pre-tax return on equities is derived from the 48-year (1954-2001) average annual return for the S&P 500 of 12.77%. ORA contends that evaluating historic performance beginning in 1954, after the Federal Reserve removed its cap on government debt rates, creates a more reliable historic record than using data beginning before the Great Depression, as PG&E has done. Furthermore, using 1954 as a starting date allows analysis of 10-year Treasury bond data.

ORA says the Commission should not adopt PG&E's rate of return assumptions when the historic results have been much higher. ORA points out that PG&E's estimates are lower than readily available investment options such as tax-free municipal bonds. ORA believes its 7.4% pre-tax return for fixed income investments is comparable to the DRI forecast, current municipal bond rates, and actual performance of the trust funds.

While ORA does not oppose SCE's methods, it does oppose SDG&E's methods. SDG&E relied exclusively on DRI long-term forecasts. In contrast, ORA says that SCE's rate of return estimate uses DRI and its own estimates to forecast its decommissioning fund performance. ORA says SCE's approach is preferable because it incorporates consideration of the historical premium for equity risk that it believes has virtually disappeared in the DRI projections.

ORA also says that SDG&E did not back-test the DRI projections for accuracy. DRI's short-term equity performance forecast from the 1998 NDCTP did not forecast the current state of the equities market. ORA believes that using the DRI projections alone, without any adjustments for historical risk premium, is not a good methodology.

Discussion

As pointed out by SDG&E, each utility has its own separate and independent decommissioning trust portfolios. In addition, each utility has different investment committees with different risk tolerances. As a result of these differences, SCE, SDG&E, and PG&E's realized investment rates of return will be different. However, in this proceeding, none of the participants has indicated specifically how these factors are incorporated into its estimates. In addition, the three utilities' trusts will have access to the same markets. As a result, their trusts will have the same investment opportunities. Therefore, we will adopt a uniform set of rate of return projections for all three utilities.

For equity returns, there is merit in using long-term historical data as used by PG&E and ORA. However, their presentations demonstrate that selection of which data to use can give quite different results. In contrast to the historical data, the DRI forecasts, which SDG&E and SCE use in different ways, yield much lower returns. No participant has demonstrated that its estimate is substantially better than the rest. The midpoint of the range of values recommended by the participants is below the 11.0% pre tax return on equities we adopted for PG&E in D.00-02-046.⁶ This leads us to believe that some reduction is appropriate. Therefore, we will adopt a 10.5% pre-tax return on equities, which is slightly above the midpoint of the range of values estimated by the participants.

Regarding fixed assets, no participant has demonstrated that its estimate is substantially better than the rest. Since the midpoint of the range of values recommended by the participants is below the 7.0% pre tax return on fixed assets we adopted for PG&E in D.00-02-046, some reduction is appropriate. Therefore, we will adopt a 6.0% pre-tax return on fixed assets, which is slightly above the midpoint of the range of values estimated by the participants.

B. Escalation Rate

The escalation rate is used to bring the current estimate of decommissioning costs to the future years in which the costs will be incurred.

PG&E calculated the simple average of the escalation rates for labor, LLRW burial costs, contract labor, materials, and other costs to arrive at an annual escalation rate. It then added a 20% contingency factor to arrive at its recommended overall escalation rate.

PG&E's escalation rates, except for LLRW burial costs, are based on DRI forecasts. The DRI forecasts do not extend beyond 2023. Therefore, PG&E used a DRI forecast to calculate escalation rates until 2023, and used the 2023 rate for subsequent years. It represents that its labor, materials, contract labor and other escalation rates are comparable to the most recent DRI forecasts.

PG&E believes that using a weighted average rate simply adds false precision to a highly speculative estimate. PG&E says that its methodology is the same as was used to calculate the overall escalation rate used by PG&E, and adopted by the Commission in D.00-02-046.

PG&E added a 20% contingency factor to come up with its overall escalation rate.⁷ PG&E states that the contingency factor ensures against future ratepayer liabilities by recognizing uncertainties with regard to changes in the economy, and protects against uncertainties in how much decommissioning costs may increase in the future.

PG&E recommends a 7.5% LLRW burial cost escalation rate for use in this proceeding as it was in D.00-02-046. PG&E says it is uncertain where the LLRW will be buried, and how much it is going to cost. PG&E believes that since the uncertainty is even greater now, with the Ward Valley disposal site stalled, and other sites about to stop taking California LLRW, a 7.5% LLRW burial cost escalation rate is a conservative and reasonable assumption.

SCE and SDG&E (the utilities) calculated separate escalation rates for: (1) labor, (2) the combined category of material, equipment, and other, and (3) LLRW burial costs. They based the separate escalation rates for labor, and the combined category of material, equipment, and other upon DRI projections. The escalation rate for the combined category of material, equipment, and other was based on a weighted average of the escalation rates for each component.

The utilities used Nuclear Regulatory Commission (NRC) published data to estimate an escalation rate for LLRW burial costs. The NRC data shows rapidly increasing burial costs followed by large, discrete jumps. The utilities utilized two similar statistical models to produce ten estimates ranging from 6.8% to 19.9%. They then chose a 10% LLRW burial cost escalation rate because of the possibility of additional large jumps in LLRW burial costs.

The utilities did not include a separate contingency factor in their calculation of escalation rates.

ORA argues that an unweighted average escalation rate makes no statistical sense, and overestimates actual escalation. ORA maintains that PG&E's unweighted calculation gives a 20% weighting to each of the five categories. However, the equipment and materials category accounts for 29%, and the "other" category accounts for 6% of actual expenditures, rather than the 20% used by PG&E for these two categories. ORA contends that this proves the inaccuracy of using an unweighted average. As a result, ORA recommends that a weighted average, based on expenditures, be used.

ORA also says that PG&E's use of the 2023 value for years after 2023, when using DRI forecasts in calculating an average escalation rate, gives undue weight to the 2023 value. It points out that, while the escalation rates in the earlier years have some relation to historic costs, the years after 2023 are not based on any independent forecast.

ORA contends that PG&E relied on a DRI forecast from 2001 in generating the labor escalation rate, and that a more recent DRI forecast yields significantly lower numbers. Therefore, ORA recommends that the Commission adopt the most recent DRI data.

ORA also says that PG&E's request for an additional 20% contingency factor is redundant since an overall contingency factor is already built into its decommissioning cost estimate.

ORA recommends a 5% escalation rate for LLRW burial costs. This is because burial costs increased only 2.4% from 1996 to the present, and only 4.3% from 2000 to 2001. ORA says that PG&E's only rationale for using a 7.5% burial cost escalation rate is that the Commission has previously adopted it.

ORA also opposes the utilities' proposed 10% LLRW burial cost escalation rate. It says the utilities relied entirely on NRC disposal cost indexes from 1986 to 2000, but did not attempt to independently verify the data. It believes that a reasonable cost escalation projection should consider additional factors to help explain a data set, and should look beyond the numbers to determine causes for their variation, as well as possible future developments. ORA says the utilities performed no such evaluation, and did not inquire as to why certain years were missing from the NRC data, or why the costs jumped significantly in certain years.

ORA maintains that the utilities' choice of data is not representative of future costs. ORA says the data used by the utilities, from three disposal sites for the period 1986-2000, reflects non-competitive disposal pricing. It also says that more recent data under more competitive conditions for Barnwell in South Carolina, and Envirocare in Utah, including contracted San Onofre Nuclear Generating Station Unit 1 (SONGS 1) LLRW burial costs, were not considered in the utilities' estimate. ORA believes the utilities have projected the most expensive possible future scenario without consideration of the prospect of a more competitive market for burial of LLRW.

Discussion

While we agree with PG&E that we are dealing with a highly speculative estimate, that is no reason to deliberately introduce an error into the calculation. ORA has demonstrated that the actual expenditures do not support the equal weighting that results from a simple average. In addition, the utilities used a weighted average. Therefore, except for LLRW burial costs, we will require the use of a weighted average.

The participants agree that a DRI forecast should be used to forecast escalation rates, except for LLRW burial costs. The disagreement appears to be over which forecast to use. Here again, although forecasts of the future are speculative by nature, it makes sense to use the most recent available forecasts. Therefore, we will use the DRI forecasts used by ORA, which are the most recent DRI forecasts in the record.

We note that the DRI forecasts run only through 2023. When determining an average escalation rate for a forecast period, PG&E uses the 2023 rate for subsequent unforecasted years. However, as pointed out by ORA, this approach gives additional weight to the last forecasted year. There is no reason that the forecast for 2023 is any better than the forecast for other years. Therefore, the average rate for the forecast period shall be used for the subsequent unforecasted years. This means that the rate for 2024, and each year thereafter, would be the average of the rates for 2002-2023.

We adopt contingency factors for cost estimates when the work to be done may change substantially over time due to such things as changing NRC requirements. This is the case with the decommissioning cost estimate. However, the escalation rate is an estimate of the rate of change in the cost of specified work. The Commission routinely adopts forecasts of cost increases, in general rate cases for example, without applying contingency factors. Since the risk of substantial changes in the work to be done and the requirements that must be met to do the work is covered by the contingency factor applied to the decommissioning cost estimate, there is no reason to apply a separate contingency factor to the calculation of the escalation rate. We also note that the utilities are not requesting one. Therefore, we will not adopt a separate contingency factor for escalation rates.

Regarding the LLRW burial cost escalation rate, the utilities estimate a 10% rate based on economic modeling of NRC data, PG&E proposes a 7.5% escalation rate based on our previous adoption of it, and ORA proposes a 5% escalation rate based on burial cost increases from 1996 to the present. Since the NRC data shows significant jumps and has no data for some years, we believe that it demonstrates the uncertainty of the costs, but does not provide a good basis for estimation. Therefore, we will not adopt the utilities' 10% escalation rate. Likewise, ORA has not demonstrated that the recorded burial costs increases from 1996 to the present provide a better basis for estimation than the NRC data. Therefore, we will not adopt ORA's 5% escalation rate. As pointed out by PG&E, it is uncertain where the wastes will be buried, and at what cost. Burial costs are no less certain now than they were when the Commission adopted a 7.5% escalation rate for PG&E in D.00-02-046. Therefore, since no participant has demonstrated that its estimate is more accurate than the other estimates, it is reasonable to continue using the previously approved rate. This rate also happens to be the midpoint of the rates recommended by the participants.

C. LLRW Burial Costs

LLRW burial costs are the costs of burying the LLRW generated by the decommissioning of a nuclear power plant. PG&E estimates LLRW burial costs of $404 per cubic foot.⁸

PG&E points out that, in D.00-02-046, the Commission adopted LLRW burial costs at the Ward Valley site of $509 per cubic foot in 1997 dollars. Because there is no indication that Ward Valley will ever be available during the times it will be needed, PG&E based its estimate on the costs of the only facility in America to which it can send more-contaminated LLRW, at Barnwell, South Carolina. Even though Barnwell is going to stop accepting wastes from non-Atlantic Compact generators such as PG&E, SCE, and SDG&E, PG&E believes Barnwell's costs are appropriate because they include all of the costs a future disposal facility (such as Ward Valley is intended to be) would likely bill a generator. Given the complete uncertainty over where these wastes will eventually go, and how much it will cost once that place is identified and operational, PG&E believes its $404 per cubic foot estimate is optimistic.

The utilities' burial cost estimate is $72.60 per cubic foot for SONGS 2&3. This estimate is based on the assumed availability of a licensed disposal facility with rates comparable to the Envirocare facility, and located within 1,500 miles of the SONGS site.

SCE's LLRW burial cost estimate for Palo Verde is $87 per cubic foot. SCE says its estimate is consistent with APS'⁹ assumptions about the burial sites that APS will use for Palo Verde LLRW.

ORA recommends that the Commission adopt the utilities' current LLRW burial cost estimate of $72.60 per cubic foot. ORA claims that PG&E derives its $404 estimate from recent cost increases at Barnwell and other facilities. ORA believes that PG&E's methodology is faulty because it ignores the likely availability of alternative facilities. ORA argues that the utilities' $72.60 per cubic foot estimate reflects their current burial cost for all classes of LLRW. ORA does not oppose the utilities' estimated LLRW burial costs for Palo Verde.

Discussion

In D.00-02-046, we adopted burial costs of $509 per cubic foot (in 1997 dollars). In this proceeding, the participants have recommended costs ranging from $76.20 to $404 per cubic foot. Therefore, it appears that the participants agree that the costs should be lower. However, they disagree on how much lower they should be.

Only PG&E and SCE actually prepared LLRW burial cost estimates. SDG&E and ORA recommend use of SCE's estimate. In addition, we have no reason to believe that there will be sufficient alternative burial sites available to lower costs due to competition, as recommended by ORA. Therefore, we are left with PG&E and SCE's estimates.

Although both PG&E and SCE's estimates are based on actual costs, neither estimate has been demonstrated to be substantially better than the other. This circumstance argues for using a cost of $240 per cubic foot, the midpoint of the range of the proposed values. However, since SCE has done a more comprehensive analysis of decommissioning costs, especially for SONGS 2&3, we will give slightly more weight to its estimates. As a result, we will adopt a LLRW burial cost of $200 per cubic foot. This amount is a bit more than twice SCE's estimates, slightly less than half of PG&E's $406 estimate for Diablo Canyon, more than PG&E's original estimate for Humboldt early decommissioning, and substantially less than the cost adopted in D.00-02-046.

D. Contingency Factors

The contingency factor is used to increase the estimated decommissioning costs to allow for uncertainties in the required decommissioning work and, therefore, the costs. PG&E recommends an overall contingency factor of 40% for Diablo Canyon. It also proposes an overall contingency factor of 40% for Humboldt for 2015 decommissioning, and 30% for early decommissioning. In contrast, ORA recommends that the Commission adopt the engineering contingency factors estimated for PG&E by TLG Services, Inc. (TLG) for Diablo Canyon and Humboldt, as the overall contingency factors.

The decommissioning cost studies, performed by TLG for PG&E, identified engineering contingency factors of 18.83% for Diablo Canyon Unit 1, 17.95% for Diablo Canyon Unit 2, and 18.54% for early decommissioning, and 21% for 2015 decommissioning of Humboldt. PG&E proposes an overall contingency factor of 40% for Diablo Canyon to take into account the engineering contingencies addressed by TLG, as well as other non-engineering contingencies such as costs associated with delays in approval of decommissioning plans, changes in the project work scope, regulatory changes and policy decisions at the state or federal level which change the scope, timeframe or level of work required for decommissioning.¹⁰ Similarly, it also proposes an overall contingency factor of 40% for Humboldt for 2015 decommissioning, and 30% for early decommissioning.

PG&E notes that for Palo Verde, a plant more like Diablo Canyon and whose decommissioning cost study was prepared by the same consultant that prepared PG&E's decommissioning cost studies, SCE is recommending a contingency factor of 40%. PG&E argues that SCE was able to reduce its contingency factor to 21% for SONGS 2&3 by making specific new estimates of costs that were previously undefined and assumed to be within the 40% contingency. Therefore, PG&E believes SCE did not eliminate its contingencies, but made individual estimates for many elements previously considered under contingency.

PG&E argues that the 40% contingency factor should be reduced only as it gets closer to the time that the actual work will be performed and costs become more certain, or as the components of potential contingencies are identified and separately estimated, as appears to be the case with SONGS 2&3. This is the reason it is recommending a 30% contingency factor for Humboldt early decommissioning.

ORA recommends that the Commission adopt TLG's estimated contingency factors. ORA says that TLG applied individual activity contingency factors of 10% to 75% to arrive at its estimates, as opposed to PG&E's 30% or 40% overall contingency factor. ORA argues that Diablo Canyon and SONGS 2&3 are roughly of the same vintage, and the utilities have been able to reduce the SONGS 2&3 contingency factor from 40% to 21%. ORA asserts that a reduction in the contingency factor is appropriate because of developments in industry-wide experience.

Discussion

PG&E's overall contingency factors for Diablo Canyon, and Humboldt, accommodate engineering, financial, regulatory, and industry uncertainties in the initial cost estimate. The contingency factors estimated by TLG for PG&E address only engineering contingencies. Because the TLG contingency factors do not address all of these contingencies, we will not adopt ORA's recommendation of 17.95% to 21% overall contingency factors for these units.

We note that SCE has utilized its decommissioning experience and knowledge to refine its estimate for SONGS 2&3.¹¹ These refinements led to a reduction in uncertainty, and therefore, a reduction in the overall contingency factor to 21%. PG&E has access to much of the same industry information as SCE. We expect that PG&E availed itself of this information and experience to produce its decommissioning cost estimates. This suggests that a contingency factor lower than 40% is appropriate. SONGS 2&3 and Diablo Canyon are estimated to begin decommissioning at about the same time. This too suggests a lower contingency factor. PG&E's estimate has not been refined to the same level as the Utilities' estimate for SONGS 2&3. As a result, the uncertainty in PG&E's estimate has not been reduced to the same degree for Diablo Canyon. Therefore, the 21% contingency factor used for SONGS 2&3 would be inappropriate for Diablo Canyon.

For the above reasons, we believe that PG&E's proposed overall contingency factor of 40% for Diablo Canyon is too high. The range of proposed contingency factors is 17.95% to 40%. Were we to give equal weight to PG&E and ORA's recommendations, we would adopt the midpoint of the range of 29%. However, ORA's proposed use of the engineering contingency factors estimated by TLG as the overall contingency factor does not address all of the contingencies the contingency factor is intended to cover. Therefore, we will give greater weight to PG&E's estimate. As a result, we will adopt a 35% overall contingency factor for Diablo Canyon.

As to Humboldt, we note that PG&E's request is based on early decommissioning for which it has reduced its contingency factor to 30%. Here too, PG&E has access to much of the same industry information as SCE. We expect that PG&E availed itself of this information and experience to produce its decommissioning cost estimates. In addition, with early decommissioning of Humboldt scheduled to start in 2006, we expect there to be substantially less uncertainty than for Diablo Canyon or Palo Verde, since they will all begin decommissioning much later. Therefore, a lesser contingency factor is appropriate. As to SONGS 2&3, they also will begin decommissioning much later than Humboldt. This tends to support a contingency factor closer to the 21% contingency factor used for SONGS 2&3. PG&E's estimate has not been refined to the same level as the utilities' estimate for SONGS 2&3. As a result, the uncertainty in PG&E's estimate has not been reduced to the same degree for Humboldt. Therefore, the 21% contingency factor used for SONGS 2&3 would be inappropriate for Humboldt.

For the above reasons, we believe that PG&E's proposed overall contingency factor of 30% for Humboldt early decommissioning is too high. The range of proposed contingency factors is 18.54% to 30%. Were we to give equal weight to PG&E and ORA's recommendations, we would adopt the midpoint of the range of 24%. However, ORA's proposed use of the engineering contingency factors estimated by TLG as the overall contingency factor does not address all of the contingencies the contingency factor is intended to cover. Therefore, we will give slightly greater weight to PG&E's estimate. As a result, we will adopt a 25% contingency factor for Humboldt.