IV. Financial Modeling Issues

Apart from the UDC risk adjustment, the financial modeling issues in this case are essentially the same as they have been in past cases. The major areas involved in financial modeling are:

1) What is the right composition of the group of companies used as comparables to the UDC?

2) Which cost of capital model should be used? The discounted cash flow model (DCF) single stage or multi-stage? The risk positioning models such as the capital asset pricing model (CAPM) or the empirical CAPM (ECAPM)?

3) What are the appropriate data inputs for the various models, to wit, the growth rate for the DCF model; the risk free rate and the market risk premium for CAPM; the beta of comparable groups; etc?

The parties' modeling presentations offered a wide range of responses to those questions, which, as PG&E says, is very confusing. Each modeling input was the subject of debate and all were tempered by judgment.

A. DCF

The DCF method attempts to measure the cost of equity by assuming that the market price of a stock is equal to the present value of all future dividends that shareholders expect to receive. To implement this method a major simplifying assumption is made: the future is divided into one or more periods (stages) of differing lengths; the dividend growth rate may differ for each of the periods. The industry needs to be sufficiently stable to make reasonably accurate forecasts for the period(s) involved. A single-stage DCF model considers the entire future as one period with an infinite number of years. A multi-stage DCF model breaks up the future into multiple periods, not necessarily of the same number of years.

The parties which utilized the DCF model and their resultant estimates of the cost of equity are shown in Table 3.

Table 3

Discounted Cash Flow - Cost of Equity

PARTY	SINGLE STAGE	MULTI-STAGE	REFERENCE
PG&E	8.7	9.3	Ex. 1 ALK-7 2T.24-26
SDG&E	11.50	--	O.B. p. 7
FEA	8.8 - 10.2	--	Ex. 16 p. 30, 50
C-K	--	10.32	Ex. 23, RJC-1, p. 2
ORA	--	8.95	Ex. 14, p. 4a

PG&E asserts that because of the current instability in the electric industry "DCF results are nonsense" (O.B. p. 18). SDG&E's witness finds that his DCF result supports his recommended range for ROE of 11.30% to 12.50% (O.B. p. 6). FEA, C and K, and ORA all support their DCF results as falling within their recommended overall ROE.

The problem all parties confronted was how to construct financial models generally acceptable to the Commission in the new unbundled environment. One answer was to use a proxy group or proxy groups of companies (most likely utilities) with risk corresponding to the electric distribution function. It was agreed that there are no pure play domestic electric distribution utilities upon which to construct a sample; the construction of sample groups provided much controversy. To cite all of the methods of each expert and the reasoning behind their choices, and each expert's criticism of the choices of the other experts would only add confusion to what is always a complex analysis. To show the variety of proxies used by the witnesses, we will cite only the method used by FEA's expert.

FEA's witness rejected the use of the U.K. utilities, and instead relied on the samples of electric utilities developed by PG&E's and Edison's experts. He eliminated companies that were subject to this proceeding, and companies considered by Edison's witness to be generation companies. He limited his group to companies followed by Value Line, and excluded companies which arguably could be considered to be in financial distress. He also used natural gas distributors in his analysis. Starting with the combined sample of all three utility witnesses, he eliminated companies which Edison's witness considered to be transmission companies, and eliminated companies which SDG&E's witness considered to be involved in acquisition activity or were not predominately gas distributors. As checks on his estimates based on the sample of electric distributors and natural gas distributors, he worked with samples of electric generators and transmission companies. The focus of his analysis was on those groups which approximate electric distribution.

The other experts were equally exacting in composing their proxy companies. It would serve no useful purpose, in our opinion, to chose one proxy group or another as a standard. What is most interesting in the DCF analysis is the closeness of result (except for SDG&E) when the "distribution adjustments" are eliminated. Rather than a spread from 8.64% to 12.1%, we have a more reasonable spread from 8.7% to 10.32%.

B. CAPM

The essential inputs to the CAPM are the risk-free interest rate, the premium that a security of average market risk commands over the risk free rate (market risk premium or MRP), and the risk of a particular company or business relative to the risk of the market (beta). Beta is multiplied by the MRP to obtain the business-specific risk premium. The debate between the parties over CAPM inputs primarily concerns different input estimates for beta and the MRP. The parties recommendations are:

Table 4

PARTY	CAPM	REFERENCE
PG&E	8.30 - 10.3	Ex. 1 ALK 9 (p. 2T 31-34) ALK 10 (p.2T-35)
EDISON	11	Ex. 10 p. 33-34
SDG&E	11.35	O.B. p. 7
FEA	9.52 - 11.0	Ex. 16 p. 48
C-K	10.69 - 10.82	Ex. 23, Sch. RJC-2
ORA	9.31	Ex. 14 p. 20

Table 5

PARTY	MARKET RISK PREMIUM OVER LONG TERM TREASURY	REFERENCE
PG&E	8.5 (Over T-bills)	Ex. 1 ALK 9 (p. 2T 31-34)
EDISON	7.5	Ex. 10, A-2
SDG&E	7.8	Ex. 7, JVW-26
FEA	7.8	Ex. 16 p. 45-50
C-K	7.8	Ex. 23 p. 13
ORA	5.5	Ex. 14 p. 19

One of the main sources of MRP is the Ibbotson Associates data, which shows 9.2 percent average premium of stock returns over Treasury bills from 1926 to 1997. Treasury bills are short-term securities, typically 3 months. Treasury bonds are typically 20- to 30-year securities. The risk premium for the post WWII period, 1947-1997 also is 9.2 percent over Treasury bills. Without the impact of the recent bull market, the Ibbotson MRP over T-bills has been close to 8.5 percent. PG&E's witness used the 8.5 percent MRP over T-bills.

SDG&E's witness and FEA's witness have utilized the Ibbotson MRP over Treasury bonds of 7.8 percent in their CAPM modeling. Edison's witness developed a MRP of 7.5 percent over T-bonds.

PG&E, SDG&E, and FEA are consistent in how their MRPs are developed. All three used well-established data series, to wit, 1926 to 1997 and/or 1947 to 1997 from Ibbotson. All three also rely heavily on the MRP from Ibbotson's arithmetic averages to estimate the expected return on the market.

ORA's witness discussed a number of topics that can bear upon MRP development. He provided an array of different data periods: 1802-1997, 1926-1997, 1951-1997, and 1971-1997. He introduced the concept of geometric averaging of market returns as opposed to arithmetic averaging. He referenced literature which indicates that the future MRP should be lower than historical MRP results. He believes the large number of data smooth out historical aberrations and capture much historical and financial information. After considering these data, he adopts a MRP over Treasury bonds of 5.5 percent and a MRP over Treasury bills of 7.5 percent. ORA's choice of MRP accounts for a substantial portion of the difference between ORA's CAPM results and the CAPM results of the other parties.

The evidence is persuasive that MRP should be based on an historical record that reaches back no further than 1926. Data from the 19^th century is too remote in time and relevance. We need not determine whether geometric averaging is superior or inferior to arithmetic averaging when seeking a market risk premium. ORA's witness considered both, but relied on neither. He looked at all the data to draw his conclusion. On the evidence presented, we are most comfortable with an MRP of 7.8%.

Another critical input for the CAPM and related models is the relative risk measure for electric utility companies. That risk measure is the "beta" of the stocks in question, or the measure of the systematic risk of the stock. Beta measures the extent to which a stock's value fluctuates more or less than the average fluctuation of the market. The theory behind beta is that risks that cannot be diversified away in large portfolios are more important for rate of return than risks which can be eliminated by diversification. A stock with a beta of .5 will tend to move 5 percent when the market moves 10 percent. A stock with a beta of 2 will tend to move 20 percent when the market moves 10 percent.

Betas are estimated from actual stock returns using standard statistical techniques. Although betas can and do vary depending on factors such as the time period and the choice of monthly or weekly returns, the main beta-related controversy in this case is whether adjusted or unadjusted betas should be used.⁵ Adjustment of beta essentially is designed to correct for a perceived (by some experts) tendency for low beta estimates to be smaller than the true beta values, and for high beta estimates to be higher than the true values.

The parties to this case used a variety of different sources for their electric utility betas. Edison's and SDG&E's witnesses used the Value Line reported betas for each selected proxy group, which are adjusted. PG&E's witness used betas adjusted with the Merrill Lynch method. FEA's witness used both the Value Line adjusted betas and the S&P unadjusted betas. ORA's witness used the Dow Jones unadjusted betas.

None of the experts in this case contest the fact that estimates of beta are subject to error. In his book, ORA's expert states that beta is always estimated with error. Similarly, Edison's expert testified that true beta cannot be captured and analyzed scientifically. The controversy concerning adjusted versus unadjusted betas has generated a huge academic literature on how to improve the estimates of beta. We are not going to attempt to ascertain the true method for determining beta as applied to ROE. We accept that each expert used his best judgment regarding beta to arrive at his recommendation. We prefer to apply our judgment to the results achieved by the expert witnesses rather than make a futile attempt to reconcile the positions. Considering the evidence regarding CAPM we are satisfied that a range of 9.52% to 11.35% is reasonable.

C. ECAPM

PG&E's expert testified that the empirical capital asset pricing model (ECAPM) is based on a body of research that can be used to improve the accuracy of the CAPM to estimate the cost of capital and should be accepted by the Commission in this case. He said the CAPM is one of the most common risk positioning models based on beta. Research, however, has established that CAPM does not perfectly capture the relationship between risk and stock returns. He said that empirical research has shown that the CAPM tends to overstate the actual sensitivity of the cost of capital to beta: low-beta stocks tend to have higher risk premia than predicted by the CAPM and high beta stocks tend to have lower risk premia than predicted. Thus, there was developed a more robust versions of the models. One such model, based on empirically determined adjustment factors, is an enhancement to the CAPM, the empirical CAPM.

PG&E's witness used the ECAPM to estimate the cost of capital. He started by taking the results based on the empirical finding that risk premia are related to beta, but are not as sensitive to beta as the CAPM predicts. He then adjusted the CAPM upward to reflect the empirical findings. Edison's witness also employed the ECAPM to develop his estimates of return differences. Both contend that since ECAPM more correctly captures the sensitivity of the cost of capital to beta, the Commission should consider the ECAPM results and should not rely solely on the CAPM's less accurate ability to estimate the cost of capital.

We are not persuaded that ECAPM produces a result that should be considered. Electric utilities in general have low betas. Adjusting betas upward guarantees a higher ROE. As Edison's witness says "Investor return requirements are largely a function of long-term expectations and perceptions of long-term risks." (Edison, Ex 10, p. 38). If betas make sense, then to claim that low-beta stocks tend to have higher risk premia contradicts the efficient market theory. What is certain is that in every example offered by PG&E's expert the ECAPM results produced higher overall cost of capital estimates than the CAPM results. (PG&E, Ex. 1, p. 2E-14.)

D. ATWACC

PG&E's expert states that the after-tax weighted average cost of capital (ATWACC) is the theoretically correct measure of the cost of capital and is used by academic and business finance professionals; its use provides a mechanism to control for changes in a firm's capital structure. He says that modern financial theory indicates that the cost of capital is constant over a broad range of capital structures. In the past, utility regulation has considered debt to provide ratepayer benefits since interest is tax deductible. However, the cost of equity increases as debt is added, keeping the overall after-tax cost of capital constant unless the company endures financial distress. In essence, this increase in equity cost occurs because the use of debt loads the entire variability in operating earnings on the smaller equity asset base, magnifying the risks to the earnings on the equity subset of assets and increasing the cost of equity. The overall cost of capital is expressed as the after-tax weighted average cost of capital. This cost of capital is independent of a company's actual debt/equity capital structure as long as its structure is within the broad range where cost of capital remains constant.

PG&E argues that the ATWACC provides a sound, accepted way to handle the capital structure issue for the UDC without needlessly debating debt/equity ratios. The ATWACC procedures set forth in PG&E's testimony are in its expert's words "absolutely in accord with textbook principles for making investment decisions as well as the way well-managed companies actually behave." (Ex. 2, p. 2-48.) Nonetheless, PG&E recognizes that the ATWACC is a new concept to the world of utility regulation with which the parties are not yet comfortable. PG&E notes that even SDG&E and Edison need more time to consider the concept. For these reasons, PG&E accepts that the Commission may not want to adopt the ATWACC in this case. PG&E requests that ATWACC may be addressed in future cost of capital cases, or their successor proceedings.

PG&E's expert testified that his analysis put the ATWACC for the bundled electric utility in the 7.25% - 7.75% range. He believes that unbundling and partial deregulation will increase the overall cost of capital by 1.0%. This yields a range of 8.25% to 8.75%, with a point estimate of 8.5%, resulting, at PG&E's book capital structure, in a cost of equity of 13.1%. PG&E's policy witness said that PG&E does not wish to increase its revenue requirement by the recommendation of its expert; rather, PG&E requests an ATWACC of 8.0% and an ROE of 12.1%.

ORA argues that PG&E has not met its burden to show that the ATWACC is a model useful to this Commission. ORA observes that whatever simplicity the model represents is immediately dissipated by its adjustments and recomputations to account for perceived omissions, not captured by the formula. Further, ORA asserts that PG&E has not met the burden of proof associated with a methodology that is new and untried in any regulatory jurisdiction. ORA asks: Is it robust? Does it yield reasonable results over time? How does utility ATWACC compare to the broader market? There is no historical data for electric or gas utilities for the Commission to assess how ATWACC would perform under a range of economic conditions, nor is there comparative information to gauge how utilities compare to the broader market.

We will not reject a proposal merely because it is new, nor need we wait for other Commissions to pronounce upon it. But the evidence presented does not give us confidence that it is more accurate or useful than other methods with which we are comfortable. As we consider the ATWACC, as presented in this proceeding, its proponent adds one full percentage point for subjective competitive risks which we cannot find, and it produces an ROE that its sponsor, PG&E, prudently reduces. If we eliminate the one percent competitive risk adjustment the ROE becomes 11.1%, a much more reasonable estimate.

E. Interest Rates

In this case, the parties agree that the Treasury rates for bonds and bills represent the normal risk-free interest rate benchmarks. The DRI April 1998 forecast of 30-year T-bonds for 1999 was 5.63%. In October 1998, the DRI forecast for 30-year T-bonds dropped from approximately 5.63 to 4.71 percent. This drop occurred during tremendous turmoil in foreign markets, when investors were fleeing to the safety of U.S. government backed securities and Treasury rates fell to unusually low rates. As a consequence, Treasury rates were not at equilibrium with other securities. Double A rated utility and municipal bonds had not experienced as steep a decline as Treasuries, especially the 30-year bond. Forecasts of AA utility bonds moved from 6.59% in April 1998 to 5.87% in October 1998.

PG&E and SDG&E are in agreement that we should consider the current estimate of interest rates when making our final decision, but should also consider the anomalous behavior of interest rates due to recent turmoil in the global financial markets. PG&E maintains that we should not implement an interest rate adjustment that exceeds 50 percent of the change in the benchmark Treasury between the time of the utilities' filings and the most recent benchmark Treasury in the record. PG&E is concerned that we would make a mechanical adjustment to reflect the Treasury interest rate change in the modeling results. They recognize that the current benchmark interest rate information is very important to our cost of capital determination, but they say that constant updating of all the model assumptions simply is not possible. They recommend that the most even-handed way to reflect the post-modeling interest rate change without updating other assumptions may be to make only a partial interest rate adjustment. In PG&E's 1997 cost of capital D.97-12-089, we stated "Our consistent practice has been to moderate changes in ROE relative to changes in interest rates in order to increase the stability of ROE over time" (mimeo, p. 12). Consistent with this statement, the Commission has had a practice of only adjusting rate of return by one half to two thirds of the change in the benchmark interest rate (D.94-11-076, 57 CPUC2d 533).

SDG&E recognizes that we must take changed interest rates into account, but does not recommend a numerical adjustment; it advises caution because of the volatility of interest rates. Edison make no recommendation.

ORA asserts that we should make a 1- to -1 adjustment between a change in interest rates and the change in investor expectations. C and K adopted a 60 basis point reduction. Their previous range of ROE of 10.7 to 10.9% drops to 10.1 to 10.3% with a recommended value of 10.2%. TURN recommends a .7 adjustment. FEA also recommends an adjustment but has not quantified it.

In prior decisions we have factored into our calculations changes in interest rates occurring after the parties have presented testimony. (D.94-11-076, 57 CPUC2d 533, 550-51; D.95-11-062, 62 CPUC2d 480, 494; D.97-12-089, p. 12). We will do the same here. We agree that interest rate changes and investor expectations do not move in lockstep. ORA makes a compelling argument in its comments that the proposed and alternate decisions moderated the change in interest rates in a manner that departs from historical practice. We have reviewed the adjustments in past cost of capital proceedings and find that in circumstances where the interest rate spread was much more significant than in this case (both upwards and downwards), adjustments ranged from 50 to 70%, with the largest adjustment applying when interest rates increased. As we view current conditions of low inflation and a stable economy, contrasted with the recent drop in the rate of the 30-year T-bond and AA utility bonds, we believe an adjustment of .6 of the decline in interest rates is warranted. We are convinced that we should not depart from past practice of adjusting model results based on changes in AA utility bond rates. As SDG&E points out in that comments, utility bonds were trading at more stable levels and the Commission has not previously relied on 30-year T-bond rates. The drop was 72 basis point for AA utility bonds; we will use 43 in our calculations.

F. Commission Recorded Risk Premium

A useful benchmark to assure that we do not act inconsistently in determining ROE is the recorded risk premium between the Commission's authorized equity returns for the integrated utilities and the interest rates forecast at the time of authorization. This benchmark was put into the record by PG&E. (Exh. 5.) As reflected in ORA's summarization, the average spread between authorized equity returns for the utilities and the 30-year T-bond rate between 1990 and 1998 was 4.76%. It is summarized in the following chart:

Table 6

CPUC Authorized Return on Equity

Spread Over Forecasted 30-Year Treasury Rate

1990-1998

This chart shows the correlation between interest rates and return on equity. We tend to increase ROE when interest rates are forecast to rise, and decrease ROE when interest rates are forecast to fall. This is no more than a reflection of the opinion of every expert testifying in this proceeding: investors always consider the opportunity cost of investments. All parties (except, perhaps Edison) agreed that their recommendations for ROE would be adjusted for a later change in interest rates forecast. The 30-year Treasury bond interest rate used by the experts in their original testimony averaged 5.94%.⁶ The April 1998 DRI forecast of 30-year Treasury bonds and AA utility bonds were 5.63% and 6.59% for 1999 respectively. The corresponding October 1998 DRI forecasts are now 4.71% and 5.87%. We will consider this reduction when adopting our ROE. We caution the parties - our use is not a computation, but a judgment.