VI. Modeling Inputs

E. Asset Lives and Depreciation

The models rely on assumptions regarding the economic lives of the assets used to provision UNEs, that is, the rate at which these assets depreciate. These depreciation assumptions are a critical input to the TELRIC cost model in today's competitive market and affect all of the UNE rates in the models,.

1. SBC-CA Proposal

SBC-CA proposes minor changes to the asset lives adopted by the Commission in the prior OANAD proceeding. Essentially, SBC-CA's analysis on this topic supports the continued use of depreciation lives that mirror those it uses for financial reporting purposes. SBC-CA contends FCC rules require economic depreciation lives. (SBC-CA/Vanston, 10/18/02, p. 8.) SBC-CA proposes asset lives consistent with its external financial reporting, and claims that "financial lives are conservatively long from the perspectives of both technology obsolescence and competitive risk." (Id., p. 9.) Moreover, in D.96-08-021, the Commission endorsed the use of economic lives that were the same as Pacific Bell's lives used for external financial reporting purposes. (Id.) While some state commissions rely on depreciation lives specified by the FCC, also known as "prescribed lives," the pace of competition and technology change in recent years makes a move to true economic lives imperative. (Id. p. 10.) FCC lives are inconsistent with economic reality because they were set at a time of minimal local competition and based heavily on the historical retirement pattern of assets. The FCC's reliance on retirement of assets as an indicator of asset life has resulted in prescribed lives that are too long and do not reflect the economic value of assets. (Id., p. 11.)

According to SBC-CA's witness Vanston, the network will transition from a primarily voice network to a full-service network based on fiber optics, advanced optical/electronic transmission equipment and packet switching and SBC-CA will have no choice but to make this transition because its competitors (e.g. cable television companies) are moving this same direction and making the existing network equipment obsolete. (Id., pp. 4-5.) Specifically, outside plant will transition from copper to fiber distribution cable, digital loop carrier equipment in the local loop will be replaced with fiber-based systems, and switching will transition from circuit to packet switching. (Id.) Vanston forecasts the rate of technology substitution and the impact of competition to compute an average life for the assets involved. Based on these forecasts, Vanston recommends a one-year increase in the projected lives for circuit equipment and metallic cable, and no change to the projected lives for switching equipment and non-metallic cable.

Table 3

SBC-CA's Proposed Asset Lives

Asset	Current Depreciation Life	Proposed Depreciation Life
Switching Equipment	10	10
Circuit Equipment	8	9
Metallic Cable (All)	14	15
Non-Metallic Cable	20	20

(Source: SBC-CA/Vanston Declaration 10/18/02, p. 7)

2. DOD/FEA and Joint Applicant's Proposal

DOD/FEA explains that in 1996, the Commission adopted depreciation lives used by Pacific in preparing its financial books and these lives were shorter than the ones previously adopted by the Commission and the FCC. (DOD/FEA/Lee, 10/18/02, p. 3.) According to DOD/FEA's witness Richard Lee, events since 1996 require a change in the depreciation lives used in setting UNE prices. Specifically, in 1997, Pacific curtailed its video and hybrid fiber-coaxial initiatives and implemented DSL technology which allows the provision of broadband services over existing copper loops. These decisions have extended the economic lives of SBC-CA's plant investments. (Id., p. 4.) Further, Lee contends that competition through resale and UNE's has no effect on SBC-CA's plant lives because its network continues to be used in these circumstances. (DOD/FEA/Lee, 3/12/03, p. 5.) Thus, SBC-CA's response to competition has resulted in economic plant lives that are longer than those previously adopted by the Commission. For this reason, DOD/FEA recommend that the FCC's projection of asset lives should be used when revising SBC-CA's UNE prices.

In 1999, the FCC reviewed the ranges for these asset lives, updated them, and stated:

These ranges can be relied upon by Federal and state regulatory commissions for determining the appropriate depreciation factors for use in establishing high cost support and interconnection and UNE prices.⁵⁵

Lee states that the asset lives prescribed by the FCC are the result of its analysis of depreciation studies filed by carriers and are forward-looking because they are based on statistical studies requiring detailed analysis of each carrier's asset retirement patterns, plans, and current technological developments and trends. (DOD/FEA/Lee, 10/18/02, pp. 5-6.) Lee maintains that recent trends in depreciation reserve⁵⁶ levels in the industry and for SBC-CA provide empirical evidence that the assets lives prescribed by the FCC are reliable and forward-looking. Lee shows that the FCC's current depreciation parameters have allowed SBC-CA to generate a surplus in its depreciation reserve. (DOD/FEA/Lee, 10/18/02, p. 10.) According to Lee, a surplus depreciation reserve indicates that the FCC's depreciation lives provide adequate compensation based on actual asset retirement rates. (Id.) Given this showing, DOD/FEA recommends that the Commission use the FCC's most recent depreciation parameters prescribed for SBC-CA in setting SBC-CA's UNE costs. According to Lee, the FCC depreciation lives assume a faster plant replacement than has actually occurred over the period of study. (DOD/FEA/Lee, 3/12/03, p. 4.) Moreover, Lee shows that several states have used FCC lives when setting UNE prices using TELRIC. From 1996 through 2002, 24 states have adopted the FCC lives in UNE proceedings.⁵⁷ (Id., pp. 14-19.)

Lee contends that economic lives for financial reporting purposes are not appropriate for TELRIC because financial reports are governed by the GAAP principle of "conservatism," which tends to understate net income in order to protect investors. (Id., pp. 12-13.) In a TELRIC cost model, conservatism would use shorter asset lives to overstate depreciation expense, thereby inflating UNE costs.

Joint Applicants support DOD/FEA's proposal to use depreciation lives prescribed by the FCC. Joint Applicants criticize the depreciation lives proposed by SBC-CA because they are lives used for financial reporting purposes, which are significantly shorter than the lives prescribed by the FCC. (JA, 2/7/03, p. 57.) According to Joint Applicants, financial accounting lives are biased low, or shorter, so accountants can conservatively err on the side of overstating costs for financial reporting. The FCC expressly rejected the use of financial accounting lives for its cost model in the Universal Service proceeding. (Id. pp. 58-9, n. 185; citing the FCC's 1999 Update.) Joint Applicants also point out that the analysis of SBC-CA's witness Vanston has been rejected by the FCC in its 1999 Update Order (JA, 2/7/03, p. 60, n. 188.)

ORA/TURN join in support for the proposals of DOD/FEA. ORA/TURN claim that SBC-CA's witness Vanston implies that the current and expected level of competition in California justifies shorter depreciation lives for UNE assets without any empirical evidence or California specific market share data to support his conclusions. ORA/TURN note that while CLCs now serve 10% of U.S. access lines, only 3.2% of those lines use CLC facilities, indicating that facilities-based competition has made very limited progress. (ORA/TURN, 3/12/03, p. 13.)

SBC-CA responds that the lives proposed by DOD/FEA are outdated because they were adopted prior to the revolutionary changes the 1996 Act brought to the local telephone market, and these outdated lives fail to consider the risks of competition and technological change intrinsic to the telecommunications industry. (SBC-CA, 3/12/03, p. 21.) According to SBC-CA, the FCC's prescribed asset lives do not keep up with the pace of competition and technology. (Id., p. 23.)

3. Discussion

TELRIC mandates that a forward-looking, empirical depreciation analysis reflect the values in a competitive environment. SBC has provided such an analysis, whereas DOD/FEA and ORA/TURN simply ignore the impact of technological change and intermodal competition in their analysis on the valuation of assets.

We note that the CPUC previously rejected the FCC's prescribed lives in 1996 because they are not appropriate in the "paradigm" of TELRIC.⁵⁸ In particular, we found instead that in the approach we adopted in D.96-08-021: "The schedules also appear realistic for a firm having to operate in a competitive environment, as Pacific will soon have to do."⁵⁹

We find no policy or economic grounds for reversing our previous approachIn particular, the widespread deployment of DSL technology does not cause us to adopt the FCC depreciation schedule. Currently, DSL speeds fall far below those of competing technologies. In addition, VoIP technology is poised to make serious inroads into the voice services provided over a twisted pair, thereby making a large amount of this investment obsolete. In I.04-07-002, we noted:

Penetration by VoIP providers into the voice telephony market is growing rapidly. Our Telecommunications Division (TD) has projected the penetration of VoIP over the next five years. Based on conservative estimates, by 2008 TD projects that VoIP will account for 40 percent to 43 percent of total intrastate telecommunications revenues in California. These projections assume no change in the number of residential and business access lines, and assume conversion rates from conventional voice service to VoIP service of 10 percent for cable/residential; 5 percent for ILEC/residential; and 10 percent for ILEC/business.⁶⁰

We therefore find Vanston's proposals that are based on technological challenges to the copper loop more compelling than predictions that copper plant will continue to have an extensive economic life. In addition, we note that Vanston proposes modest increases in the economic lives of some assets - switching equipment and non-metallic cable. These minor changes also comport with our understanding of the direction of technological change.

In the FCC's Triennial Review Order, the FCC clarified its views and specifically declined to endorse one method of calculating asset lives over another saying it could not conclude which more closely reflected the actual useful life of an asset in a competitive market. The FCC specifically invited States to use discretion in determining whether to use financial or regulatory lives in determining depreciation expense as it relates to TELRIC pricing. In response to requests to clarify the depreciation component of TELRIC analyses, the FCC stated:

We decline to adopt the incumbent LECs' suggestion that we mandate the use of financial lives in establishing depreciation expense under TELRIC. The incumbent LECs have not provided any empirical basis on which we could conclude that financial lives always will be more consistent with TELRIC than regulatory lives. Both financial lives and regulatory lives were developed for purposes other than, or in addition to, reflecting the actual useful life of an asset. [Footnote omitted] We cannot conclude on this record that one set of lives or the other more closely reflects the actual useful life of an asset that would be anticipated in a competitive market. Accordingly, state commissions continue to have discretion with respect to the asset lives they use in calculating depreciation expense. (TRO, para 688.)

Thus, our decision to continue use of the asset lives based on previous investigation as amended by the evidence proffered in this proceeding is clearly consistent with the TRO.

F. Cost of Capital

Another critical input to a TELRIC cost model in today's telecommunications industry is the estimated cost of capital, which is the cost a firm will incur in raising funds in a competitive capital market. The cost of capital is usually expressed as a weighted average of the cost of equity and the cost of debt for the firm, or a proxy group of firms, with a similar risk profile and in the same line of business as the firm. Therefore, there are several key components used to calculate the weighted average cost of capital:

· Cost of equity -The Capital Asset Pricing Model (CAPM) and the Discounted Cash Flow (DCF) analysis technique are two quantitative financial models commonly used to estimate cost of equity, also called return on equity (ROE). These methods require assumptions regarding company growth rates, the premium that a stock of average risk commands over the risk free rate (market risk premium), the risk-free rate of return, and a measure of the risk of the company's stock (beta).

· Cost of debt - this involves estimates of the interest rates on long term, and perhaps short-term, debt instruments.

· Capital structure of the firm - this refers to the amount of debt and equity outstanding for the company, or proxy group.

· Proxy group - this key assumption involves the composition of the group of companies used as comparables to the ILEC's UNE business.

Federal regulations require that a "forward-looking cost of capital shall be used in calculating the [TELRIC] of an element." (47 C.F.R. 51.505(b)(2).)

In its recent Triennial Review Order, the FCC provides clarification on the cost of capital component of a TELRIC analysis. The FCC states that there are two types of risk that should be reflected in the cost of capital. First, a TELRIC-based cost of capital should reflect the risks of a competitive market. Specifically, the FCC says:

Because the objective of TELRIC pricing is to replicate pricing in a competitive market, [footnote omitted] and prices in a competitive market would reflect the competitive risks associated with participating in such a market, we now clarify that states should establish a cost of capital that reflects the competitive risks associated with participating in the type of market that TELRIC assumes. The Commission specifically recognized that increased competition would lead to increased risk, which would warrant an increased cost of capital. (TRO, para. 681.) (Footnote omitted.)

Second, the FCC states that a TELRIC-based cost of capital should reflect any unique risks (above and beyond the competitive risks discussed above) associated with new services that might be provided over certain types of facilities. The TRO specifies that states may establish UNE-specific costs of capital to reflect in UNE prices any risk associated with new facilities that employ new technology and offer new services. (TRO, para. 683.) Nonetheless, the FCC leaves states the option to adopt a single cost of capital for all UNEs. (TRO, para. 684.)

Table 4 summarizes the parties' proposals for the appropriate cost of capital to incorporate into SBC-CA's UNE prices and compares it to the cost of capital incorporated into current UNE rates.

Table 4

Current and Proposed Cost of Capital

Current Cost of Capital	SBC-CA Proposal	Joint Applicants' Proposal	XO Proposal	Z-Tel Proposal	ORA/TURN Proposal
10.00%	12.19%	7.63%61	7.7%	6.6%	7.7%

While these proposals differ by 559 basis points,⁶² the methods used by all parties are remarkably similar. SBC-CA and JA offered the most commentary concerning cost of capital. Both of these parties calculated a weighted average cost of capital based on their own unique assumptions regarding the cost of equity, cost of debt, and capital structure of the firm. We will discuss each of these components of the cost of capital calculation separately. But first we will give a brief overall description of each party's proposal.

1. SBC-CA Proposal

SBC-CA proposes a cost of capital of 12.19%, which is 219 basis points above the 10.0% cost of capital adopted by the Commission for use in the prior OANAD proceedings. SBC-CA's witness Avera used a group of seven LECs and estimated their average market-value capital structure, cost of equity, and cost of debt. The study incorporates a 13% rate of return on equity and a 7.18% cost of debt. His study incorporates a market value capitalization consisting of 86% common equity and 14% debt based on the average capital structure of the proxy group. Avera's analysis is based on data from year-end 1998. (SBC-CA/Avera, 10/18/02, Attachment WEA-1, Table 1.)

SBC-CA contends that investment risks associated with the telecommunications industry, and LECs specifically, have increased significantly since the Commission adopted a 10% cost of capital. Further, he contends changes in capital market conditions warrant the increased cost of capital. For example, changes in long-term bond rates have been modest and stock valuations for telecommunications firms have weakened. Avera concludes that investors are less willing to provide capital, which means higher borrowing costs. (SBC-CA/Avera 10/18/02, pp. 22-23.)

JA criticize SBC-CA's analysis for relying on financial data that is four years old, primarily from year end 1998 and first quarter 1999, which they contend is too stale to form a reasonable basis for estimating a forward-looking cost of capital. (JA/Murray, 2/7/03, p. 65.) JA's witness Murray notes that the stock market has declined sharply since SBC-CA's estimate of its market capitalization based on year-end 1998 financial data. There have been substantial changes in the proportions of debt and equity in market capital structures given declines in incumbent LEC stock prices. The average proportion of debt in the market capital structure has increased significantly since Avera's analysis. He fails to recognize that long-term debt costs have decreased significantly since his analysis. (Id., p. 66.) In addition, Murray contends Avera makes four serious methodological errors as well. We will discuss these in greater detail below, but essentially Murray criticizes Avera for averaging financial information across companies with disparate capital structures, using an inflated risk premium based on one academic study, relying on a purely market-based capital structure, and using unrealistic growth assumptions in the DCF model.

JA revises SBC-CA's estimates using more current information and shows that SBC-CA's analysis, when revised with current financial information, converges on JA's own proposed cost of capital of 7.7%. (Id., p. 53.)

2. Joint Applicants' Proposal

JA's witness Murray proposes a cost of capital of 7.63%, which is 237 basis points below the 10.00% cost of capital adopted by the Commission in prior OANAD decisions for SBC-CA.⁶³ Murray's financial modeling of the cost of capital is based on a proxy group of SBC-CA, Verizon and BellSouth, and uses holding company level data for these three companies. Murray's financial analyses incorporate a 50/50 weighting of market and book capital structure, a return on equity of 9.92%, and a cost of debt based on forecasts of short and long-term interest rates. Short-term debt cost is 3.18% and long-term debt cost is 5.51%.

JA note the sharp interest rate declines since the current 10% cost of capital was adopted in the first triennial review of the Commission's New Regulatory Framework proceeding in 1994. According to JA, interest rates are at 40-year lows and these low interest rates reduce SBC-CA's debt costs and the opportunity cost of investing in equity, which reduces SBC-CA's cost of equity. Low debt costs encourage SBC-CA to take advantage of low-cost debt in its capital structure, which lowers its weighted average cost of capital. Murray argues SBC-CA is vastly different today than when the 10% rate was set in 1994 because it is subsidiary of SBC, a far larger company. Also, the legal and regulatory environment has changed given the passage of the 1996 Telecommunications Act, which opened local market to competition.

SBC-CA criticizes JA's cost of capital proposal for resting too much on trends in current interest rates and ignoring the risks that investors perceive for the local exchange telephone and UNE businesses. SBC-CA contends that JA approach their cost of capital estimate using traditional rate-case methods for a traditional utility, while ignoring the fact that operations and rates of telephone companies are no longer regulated like traditional utilities.

3. Proposals of XO, Z-Tel and ORA/TURN

ORA/TURN and XO support the Joint Applicants' original proposal for a cost of capital of 7.70%, noting that SBC-CA's analysis uses financial data that is four years old. (ORA/TURN, 2/7/03, p. 18; XO/Montgomery Decl., 2/7/03, p. 24.) Commenting on the age of the financial data in SBC-CA's filings, ORA/TURN state, "Unlike fine wines, cost of capital studies do not improve with age." (ORA/TURN/Roycroft, 2/7/03, p. 79.) Similarly, Z-Tel criticizes SBC-CA's use of four-year-old data. When Z-Tel's witness Ford updates the SBC-CA analysis, he obtains an estimated cost of capital of 8%. Ford then makes revisions to SBC-CA's methodology and proposes a cost of capital below 7% based on SBC-CA's target capital structure. (Z-Tel/Ford Decl., 2/7/03, p. 32.)

4. Discussion

Both SBC-CA and JA agree it is time to revise the 10% cost of capital set in 1994. We wholeheartedly concur. Financial conditions are vastly different today than they were in 1994, not to mention the legal and regulatory landscape after the passage of the 1996 Act and subsequent litigation. As Murray notes, the numerous mergers in the industry have created entirely different companies than the ones that existed when we last set a cost of capital for Pacific. Pacific is now SBC-CA with a new parent company, SBC, which has in turn merged with another former regional bell operating company (RBOC), Ameritech, to form one of the four remaining RBOCs. SBC is no longer one of seven RBOCs that existed in 1994, but rather one of the four surviving RBOCs. This change alone calls for a new evaluation of SBC-CA's cost of capital for its UNE line of business.

Despite the many pages of rhetoric on this topic, all parties essentially used the same financial modeling techniques, but with differing inputs and assumptions. We analyze each of their positions on the various components of the financial models below in order to determine the most reasonable inputs for financial modeling of the cost of capital. A summary of the financial modeling with the inputs we select is found in Section VI.B.4.f.

It is important to note that while we will review the financial modeling presented by the parties, particularly where it estimates the cost of equity, we will use judgment as well as the models to render our decision. As we stated in our order in 2002 where we established a return on equity for the four major energy utilities:

In the final analysis, it is the application of informed judgment, not the precision of financial models, which is the key to selecting a specific ROE estimate. We affirmed this view in D.89-10-031, which established ROEs for GTE California, inc. and Pacific Bell, noting that we continue to view the financial models with considerable skepticism. (D.02-11-027, mimeo. at p. 19.)

Finally, although the FCC's recent Triennial Review Order discusses the option to set unique costs of capital for each UNE, we will establish one cost of capital for all UNEs because we have no record to do otherwise.

We now turn to an examination of the inputs to the financial models used by the parties.

a. Proxy Group

A starting point for the quantitative analysis of SBC-CA's cost of capital is a reference or "proxy group" of companies. The proxy group is needed because there is no company purely in the business of selling UNEs that we could look at to see its cost of equity, cost of debt, and capital structure. Instead, it is logical to look to a group of companies in a similar line of business and determine the average capital structure, cost of equity, and cost of debt faced by those companies. Both SBC-CA and JA used a proxy group, but differed in the composition of that group.

SBC-CA used a proxy group of seven LECs and reviews financial data for this group of seven LECs from year-end 1998. (SBC-CA/Avera, 10/18/02, Attachment WEA-1, Tables 1 and 2.) JA used a proxy group of three companies, SBC, Verizon and BellSouth. JA's witness Murray notes that the group of seven LECs SBC-CA uses as a benchmark for capital structure has changed substantially since 1998 due to mergers and acquisitions involving SBC and Ameritech, GTE and Bell Atlantic, Qwest and US West, and Broadwing and Cincinnati Bell. (JA/Murray, 2/7/03, p. 56.) According to Murray, Qwest and Broadwing are no longer comparable to SBC and should be excluded from the group because they are much smaller, experiencing major financial difficulties, and investors perceive greater risk in these two companies.⁶⁴

SBC-CA does not deny that the proxy group of seven companies Avera uses has changed substantially since 1998.⁶⁵ We agree with Murray that because of the mergers and acquisitions of several of these companies and other changes affecting their financial position and risk, the group of seven companies used by SBC-CA is not appropriate. We will exclude Qwest and Broadwing because these companies are much smaller than SBC, they have both experienced major financial difficulties indicating higher risk levels, and they have negative earnings so they cannot be included in a DCF analysis. (JA/Murray, 2/7/03, p. 78.) We will, therefore, use the proxy group of 3 companies proposed by JA, namely SBC, Verizon and BellSouth.

b. Cost of Equity

Despite SBC-CA criticizing JA for using "traditional rate case methods," both SBC-CA and JA use fairly similar and standard methodologies for estimating SBC-CA's cost of equity. Namely, they both use the CAPM and DCF methods to estimate cost of equity. The DCF method estimates the return that investors require on equity investments by assuming the market price of a stock equals the present value of all future dividends investors expect to receive. The CAPM model estimates investors' required return on a particular stock over the return required by the market in general.

SBC-CA's Avera obtains two estimates of the cost of equity using the DCF model and two using the CAPM model, then he averages the four results to arrive at his proposed cost of equity of 13.0%. Murray does essentially the same thing, except she derives only one estimate from each model. She then averages her DCF and CAPM results to arrive at her proposed cost of equity of 9.92%. The difference between these proposals derives from differing assumptions that underlay the DCF and CAPM models. We will now discuss these.

i. DCF Analysis

DCF models attempt to replicate the market valuation process that investors use to determine the price they would be willing to pay for a share of a company's stock. (SBC-CA/Avera, 10/18/02, p. 10.) SBC-CA's witness Avera uses a simplified form of the DCF method known as the "constant growth" form, which depends on an assumption that long-term growth for the company will occur at a constant rate. Avera performs two separate DCF analyses, using 1999 constant growth assumptions of 9.6% and 11.6%, which results in a DCF cost of equity of 12.2% and 14.3%, respectively. (Id., Attachment WEA-1, p. 19.)

JA contend that SBC-CA's DCF analysis makes the unrealistic assumption that a company can continue to grow forever at a faster rate than the overall economy. Instead, JA witness Murray uses a different "three-stage" DCF model that assumes three stages of company growth, in which a company's growth rate regresses toward the same growth rate as the overall economy in the long-run. (JA/Murray, 2/7/03, p. 77.) Murray contends this is a more realistic approach because extraordinary growth in the near term typically slows to a more stable level. (Id.) Murray's growth rates for her proxy group of three companies range from 3.77% to 6.7%, based on recent growth forecasts from I/B/E/S (now Thomson First Call) and forecasts of overall economic growth by the Federal Reserve Bank of Philadelphia's Survey of Professional Forecasters. (JA/Murray, 10/18/02, p. 57-58; JA/Murray, 2/7/03, p. 59.) As a result, her average DCF cost of equity estimate is 9.97% (JA/Murray, 3/12/03, Exhibit TLM-REB 5.)

In reply to criticisms of his constant growth approach, Avera introduces a new methodology for calculating the growth rate used in the DCF formula, known as the "sustainable" or "b x r approach."⁶⁶ In other words, rather than updating his constant growth DCF method with more current data, he chooses a different approach to calculate the growth rate. Avera's new "b x r" approach results in growth rates ranging from 9.2% to 11.5%, and a cost of equity ranging from 13.1% to 14.6%. (SBC-CA/Avera, 2/7/03, p. 11 and WEA-1.)

SBC-CA criticizes Murray's three-stage DCF analysis for using growth estimates that SBC-CA believes are too low. (SBC-CA/Avera, 2/7/03, pp. 9-11.) Avera argues that current growth rates Murray uses are depressed and that "accelerating growth in excess of the economy as a whole is consistent with investors' long run view of telecommunications as one of the most dynamic segments of the economy." (Id., p. 10.)

Murray contends that running the constant growth DCF with updated numbers gives almost the same results as her own 3 stage DCF analysis, and in fact, the 3 stage produces a higher cost of equity.⁶⁷ She criticizes Avera's new b x r "sustainable growth" method, which is based on an r-value for expected rate of return ranging from 17.4-17.8% for SBC-CA, Verizon and BellSouth (JA/Murray, 3/12/03, p. 56.) Murray says these return estimates are not sustainable unless one believes these three companies will take over the majority of the US economy within the next 30 years. (Id., p. 57.)

Z-Tel revises Avera's growth numbers in the DCF method with current Value Line and I/B/E/S growth estimates that range from 4% to 6.15%. These growth estimates produce DCF cost of equity estimates ranging from 7.6% to 9.8%. (Z-Tel, 2/7, p. 22.)

In reviewing these various DCF analyses, we are immediately struck by the outdated growth estimates used by Avera. The financial outlook for telecommunications firms today is without question vastly different from the outlook four years ago in the first quarter of 1999, the time frame of Avera's data. Therefore, we find that Avera's original DCF analysis is outdated and we will not rely on it. Second, we prefer Murray's three-stage DCF analysis rather than the constant growth DCF used by Avera. Murray's explanation of the three stage model with growth rates that converge upon the growth rate of the economy is more reasonable than assuming telecommunications firms will continuously grow at a faster rate than the whole economy. Further, the growth rates Murray uses are more reasonably based on recent analyst growth estimates.

Third, we agree with Murray's criticism that Avera's updated 9.2 to 11.5% growth rates based on his "b x r approach" are excessive to assume in a constant growth formula. We find Murray's long-term growth in the 5% range is more reasonable, as is her three-stage DCF formula. We would have preferred to see Avera update his growth rates using the same source, rather than a new methodology. Indeed, we find it interesting that when Murray did in fact update Avera's own analysis, she achieved results using the constant growth DCF that are only a few basis points lower than her results with the three-stage formula. (JA/Murray 2/7/03, p. 64.)

Therefore, we find JA's DCF result of 9.97% more reasonable than SBC-CA's DCF results of 14.3% and 12.2%. We will consider this information along with the results of the CAPM analysis when determining the appropriate cost of equity.

ii. Market Risk Premium in CAPM

In the CAPM approach, the cost of equity is estimated based on three key inputs: (1) the risk-free interest rate, (2) the risk of a particular company or business relative to the risk of the market (beta),⁶⁸ and (3) estimates of the additional return investors require to forego the safety of no or low risk bonds and to bear the greater risk of common stock, also known as the "market risk premium."⁶⁹ SBC-CA provides two alternative versions of the CAPM based on two different estimates of the market risk premium. One CAPM study estimates an "expectational" cost of equity based on a forward-looking estimate of the market risk premium. The other CAPM analysis involves a historical view of the market risk premium. (SBC-CA/Avera 10/18/02, p. 16.)

For SBC-CA's "expectational" approach, SBC-CA witness Avera uses an estimated market risk premium of 6.47% over long term government bond yields.⁷⁰ Avera then adjusts this risk premium upwards to 9.1% because of declines in interest rates since the time of the 1992 study. Avera justifies this adjustment by claiming there is substantial evidence that equity risk premiums move inversely with interest rates, so that when interest rates are low, the premiums investors demand for equity rise. (SBC-CA/Avera 2/7/03, p. 5.) Avera then adjusts this required rate of return for the S&P 500 using the beta for his proxy group of companies. Avera averages the betas reported for the proxy group of seven LECs which is .83. (SBC-CA/Avera 10/18/02, Attachment WEA-1, Table 2, citing Value Line Investment Survey, April 1999.) This results in a cost of equity estimate of 13.35%.⁷¹ Next, Avera performs a historical CAPM analysis using a risk premium of 7.5%.⁷² Inserting the 7.5% risk premium into the CAPM formula results in a cost of equity of 12.03%.⁷³ On reply, Avera presents a new analysis of the market risk premium based on the S&P 500 that shows a 9.86% market risk premium. (SBC-CA/Avera 2/7/03, p. 13.)

JA criticize what they consider SBC-CA's inflated "expectational" estimate of the market risk premium. JA witness Murray maintains this market risk premium is out of line with other academic sources, and then inappropriately adjusted up another 2.53% based on changes in interest rates. (JA/Murray 2/7/03, p. 67.) Murray shows that Avera's source (Harris & Marston) has performed an updated analysis that decreases the prior estimate of the equity premium's sensitivity to interest rates. (Id., p. 70.) Murray contends that Avera's risk premium calculations are outlandishly high compared to other sources he used such as DRI-WEFA, which predicts S&P 500 equity returns of 6% over the next 25 years, and the Survey of Professional Forecasters, conducted by the Federal Reserve Bank of Philadelphia, which estimates average returns of 7.47% over the next 10 years. (JA/Murray 3/12/03, p. 59.) Murray states these two forecasts imply equity premiums of only 3 to 4% above the risk free rate, rather than the 9.1% premium proposed by SBC-CA. (JA/Murray, 10/18/02, p. 63.)

For her own analysis, Murray uses four academic studies that forecast equity premiums in the 3-4% range, which is lower than historical return levels. (JA/Murray, 10/18/02, p. 63.) She also cites the historical equity premium data of Ibbotson Associates, which measured stock market returns from the 1926 through 2002 time period, indicating a historical premium of approximately 7.4%. (JA/Murray 2/7/03, p. 60.) She then constructs an average estimate of the market risk premium based on these sources, giving equal weight to the historical and forecasted risk premiums, which results in an average risk premium of 5.8% (JA/Murray 3/12/03, Exh. TLM-REB-5).

SBC-CA opposes Murray's CAPM analysis, and particularly her market risk premium of 5.8%, as "predicated solely on historical results," whereas forward-looking estimates of investor's required rates of return are higher.

XO criticizes Avera for using new DCF and CAPM methods in his reply declaration, rather than updating the four year old financial information inserted into the methods he originally used. (XO/Montgomery, 3/12/03, p. 5.)

Z-Tel criticizes the risk premium Avera uses based on Harris & Marston study. Z-Tel alleges that Harris & Marston study is flawed for several reasons, chiefly that it is limited to the 1982 through 1998 time frame when the market exhibited exceedingly high returns. (Z-Tel/Ford, 2/7/03, p. 24.) Instead, Z-tel proposes a risk premium of 5% based on historical market returns from 1970 through 2002.

SBC-CA criticizes the 1970 through 2002 time frame chosen by Z-Tel as biasing the risk premium down. SBC-CA says the most exhaustive and widely accepted study is published by Ibbotson Associates. Their 2002 Yearbook indicates an average equity risk premium of 7% over long-term government bonds for 1926 through 2001. (SBC-CA/Avera 3/12/03, p. 16.)

Table 5

Market Risk Premium Proposals

SBC-CA Historical	SBC-CA Expectational	SBC-CA S&P 500	JA	XO	Z-Tel
7.5%	9.1%	9.86%	5.8%	5.8%	5%

After reviewing the various proposals, we find it most reasonable to use the historical measure of risk premium documented by Ibbotson Associates and cited by both SBC-CA and JA. We will use the latest update to the Ibbotson Study provided by JA, which indicates a risk premium of 7.4%. We find this updated 7.4% estimate is a generous estimate of the market risk premium given the variety of studies with much lower findings in the 3 to 4% range, such as those cited by JA's Murray from reputable sources such as DRI-WEFA and the Survey of Professional Forecasters. We decline to use Z-Tel's figure because it is derived from a short study period. Similarly, SBC-CA's proposal of a 9.86% market risk premium is based on a time period of 1982 to 1998 that is very short and therefore not reasonable.

We will not average the Ibbotson historical 7.4% with the forecasted equity premiums provided by Murray for JA. We prefer to base our cost of equity analysis on documented historical returns rather than disputed expectations of future returns. Likewise, we decline to use SBC-CA's expectational analysis because SBC-CA makes a controversial "interest rate" adjustment to achieve a market risk premium of 9.1%. We do not agree with this interest rate adjustment. Harris & Marston have updated their assumptions regarding interest rate effects and it is not entirely clear that making this kind of adjustment to the equity risk premium is appropriate. Certainly, the results of Avera's "adjustment" are out of line with other forecasts of the equity risk premium cited by Murray. We are not convinced that it would be wise to add short-term interest rate volatility into the equity risk premium portion of the CAPM formula. Interest rate changes are accounted for in the cost of capital calculation through revisions to the risk-free rate, and through the cost of debt estimates used to weight debt given the firm's overall capital structure. In prior cost of capital reviews, the Commission has occasionally adjusted estimates of cost of equity for changes in interest rates after the analysis is complete, not by making an adjustment to the market risk premium and inputting this into the analysis. (See D.99-06-057, mimeo., p. 49.)

iii. Risk Free Rate

SBC-CA proposes a risk free rate of 5.8% based on long-term government bond yields from March 1999. (SBC-CA/Avera, 10/18/02, Attachment WEA-1, p. 17.) Avera updates this risk-free rate to 4.9% based on 30 year Treasury bond yields reported by Moody's in January 2003. (SBC-CA/Avera, 2/7/03, Attachment WEA-2.)

JA propose a risk free rate of 4.73%, which is the yield to maturity on 10-year U.S. Treasury Notes. (JA/Murray 10/18/02, p. 64; JA/Murray 2/7/03, p. 64, n. 104.)

SBC-CA's original risk-free rate of 5.8% based on 30-year Treasury bond yields is quite outdated. Murray proposes we use the yield on 10-year Treasury notes. While this is a more current figure, we would prefer to use a risk-free rate based on a longer investment horizon to match the long-term risk premium analysis that is incorporated into this cost of capital analysis. Therefore, we will use Avera's updated 30-year Treasury bond yield figure of 4.9% for our calculations.

iv. Beta

SBC-CA's Avera used an average beta of 0.83 for his proxy group of seven ILECS, based on a Value Line Investment Survey from April 1999. (SBC-CA/Avera, 10/18/02, Attachment WEA-1, p. 18.) Avera performs an updated analysis using Value Line Investment Survey results from January 2003 for the three companies in Murray's proxy group. The average beta for this proxy group is .93 (SBC-CA/Avera 2/7/03, Attachment WEA-2).

JA's witness Murray alleges that Avera improperly uses a simple averaging of betas across companies with disparate capital structures. Companies face financial risks based on the amount of debt, or "leverage," in their capital structure and they face business risk from earnings fluctuations. The purpose of averaging betas for comparable companies is to measure their business risk, not the risk inherent in their capital structures. To remove the financial risk associated with the company's chosen leverage, the betas should be "unlevered" before they are averaged. Murray maintains it is preferable to determine SBC-CA's "unlevered beta," and then average the unlevered beta with other companies with comparable business risk. (JA/Murray 2/7/03, p. 65-6.)

For her own analysis, JA's Murray used a beta coefficient of .929, which is the "average relevered beta" of SBC-CA's stock, based on betas for her proxy group of companies.⁷⁴ (JA/Murray 10/18/02, pp. 65-66; and JA/Murray 3/12/03, Ex. 5.)

Z-Tel's witness Ford uses a beta coefficient of .59 in his calculations, which he obtained from the marketguide.com website.

We will adopt Avera's updated beta coefficient of .93 because it is based on the same proxy group of three that we have used for other calculations and it is based on recent data for these companies. Murray's description of unlevering
and relevering betas makes intuitive sense and results in almost the identical number. Therefore, we will opt to use Avera's updated estimate of .93. We note that using a beta of .93 is actually higher than SBC-CA's original proposed beta of .83. A higher beta means a relatively riskier investment where investors require a higher return on equity. So in this case, by using Avera's update, we are actually increasing the estimated cost of equity because we assume investors want a return closer to the market risk premium when they invest in the proxy group of companies.

v. CAPM Summary and Resulting Cost of Equity

SBC-CA proposed a 13% cost of equity, while JA had proposed 9.92%. We decline to use SBC-CA's inflated DCF results based on outdated 1999 growth estimates, and we find JA's DCF results more reliable than those proposed by SBC-CA. We also prefer SBC-CA's historical CAPM analysis using a 7.4% market risk premium rather than forecasts of future market risk premiums provided by SBC-CA and JA. Thus, we decline to adopt SBC-CA's controversial expectational market risk premium of 9.1% and JA's market risk premium of 5.8%. Similarly, we revised other inputs to the CAPM model, namely the risk free rate and beta.

Using CAPM, we apply the 7.4% historical market risk premium, multiply it by a beta of .93, and add this to our chosen risk free rate of 4.9% to obtain our cost of equity for SBC-CA. The calculation is:

(7.4% x .93) + 4.9 % = 11.78%.

When setting the cost of equity, we prefer to use CAPM model results, rather than the DCF model. The DCF model relies heavily on growth forecasts for telecommunications firms, which vary greatly depending on the source. This leads to a large disparity in DCF results depending on the time period and forecasters selected. It appears that the DCF model is too dependent on this one forecasted input, and we prefer instead to use CAPM, which is based on betas, the risk-free rate and the market risk premium rather than highly disputed growth forecasts for one industry. Therefore, we will base our adopted cost of equity on the conservatively higher CAPM results.

c. Qualitative Risk Issues

In conjunction with Avera's quantitative analysis of SBC-CA's weighted average cost of capital, Avera provides a qualitative discussion of the risks SBC-CA faces in providing UNEs. Avera contends that SBC-CA faces competition from an ever expanding array of alternative carriers and technologies such as specialized fiber, wireless, and cable companies that offer a full array of broadband services. (SBC-CA/Avera, 10/18/02, pp. 17-18.) He contends that SBC-CA must invest in network architecture while it simultaneously faces the threat of high operating leverage, exposure to loss of profitable customers, and risk of rapid technological change. Avera says the rapid pace of technological change increases investors' risk perceptions for UNEs. (Id., p. 31, citing a UBS Warburg study from 2002.)

He states that risks are magnified for UNEs because continued regulation of UNEs hampers SBC-CA's ability to respond in an increasingly volatile market. ILECs face a combination of competitive and regulatory uncertainty that exceeds the risk of other ILEC business segments. (Id., pp. 4-6.) SBC-CA is obligated to install and maintain sufficient capacity to meet competitors' demand for interconnection and UNEs, but CLCs are free to drop off the network anytime. Thus, while there is volatility in demand for UNEs, SBC-CA is constrained by regulation from altering the price of UNEs. (Id., p. 26.)

Plus, there is a risk under the current regulatory structure, that UNE prices will be set incorrectly, hurting SBC-CA's cash flow, and its ability to attract capital and its ability to develop alternative networks and new technologies. According to Avera, this combination of competitive and regulatory risks makes SBC-CA's cost of capital to provide UNEs higher than the cost of capital in SBC-CA's other business segments, particularly since SBC-CA, as a stand-alone provider of UNEs, would not have the advantages of diversification.

Avera compares his proposed return on equity of 13% to authorized rates of return for energy companies that are between 10.9% and 11.6%.⁷⁵ Avera maintains that because this proceeding is not about determining the rate of return for a traditional utility but a forward-looking rate of return for a competitive telecommunications network, it is reasonable that SBC-CA proposes a return on equity higher than the one approved for the energy utilities. It is not reasonable that JA propose a return on equity of only 9.92%, lower than the one authorized for the energy companies. Moreover, Avera defends his proposal as consistent with the FCC's 11.25% cost of capital, which has been in place for several years.

In response, Murray has several criticisms of Avera's qualitative assessment of business risk. First, she says it is merely conjecture and ignores important context about SBC-CA's financial strength.⁷⁶ Murray contends that Avera's assumption of higher risks to provide UNEs cannot be substantiated with quantifiable analysis of actual capital costs faced by SBC-CA as a whole. (JA/Murray 2/7/03, p. 79.) Murray maintains that it is more prudent to rely on a quantitative analyses of capital costs, using current market data that incorporates the capital markets' assessment of all the qualitative considerations. Financial market participants have already incorporated qualitative considerations into the share price of the companies. (Id., p. 78.)

Second, Murray says that the kinds of competitive and regulatory risks Avera describes are "company specific risks" which are diversifiable. Murray explains that the Commission has concluded in the past that when setting the cost of capital in a regulatory proceeding, the Commission "should give little weight to risks that are diversifiable." (Id., p. 84, citing D.94-11-076, p. 31.) Third, Murray says Avera ignores provisions for universal service support and pricing flexibility, which mitigate SBC-CA's risks. (JA/Murray 2/7/03, p. 86) Fourth, Murray says Avera improperly focuses on retail rather than wholesale risks. Cost of capital should be based on risk associated with leasing UNEs at wholesale, not competition for end-users of telephone service. (JA/Murray 3/12/03, p. 53.)

Fifth, Murray says that the cost of capital of 11.25% set by the FCC in 1990 is extremely stale. In 1996, the FCC found that an 11.25% cost of capital is much higher than the rate required to attract capital and earn a reasonable profit, and it determined it should begin a new proceeding to review the 11.25%. (Id., p. 61.)⁷⁷ Finally, in contrast to Avera's position, Murray contends that the provision of UNEs is less, not more, risky than other operations of the SBC-CA holding company, such as DSL and long-distance concerns, which cause investors to demand a higher return for the company as a whole. Because her analysis focuses on holding company level financial data, which includes the capital costs for SBC-CA's unregulated business segments, she believes her analysis overstates the cost of capital for UNEs alone. (JA/Murray 10/18/02, p. 44, n. 44.)

XO agrees with Joint Applicants' position that UNEs are subject to less competitive threat than SBC-CA's other product lines because XO and other competitive carriers have no viable alternative to several of SBC-CA's UNEs. XO explains that it has strong incentives to obtain facilities from sources other than SBC-CA, but they are simply unavailable.

It is interesting that although SBC-CA argues that UNEs are a high-risk venture, it does not propose using a cost of capital greater than the one it calculates for the firm as a whole. We find that despite SBC-CA's lengthy qualitative discussion of the risks facing SBC-CA, Avera has not persuaded us that UNEs are more risky than SBC-CA's other unregulated ventures, which are subject to competitive markets, such as the long distance and DSL markets.

SBC-CA's commentary on the relative risk of UNEs is not convincing. Avera admits that "investors... establish the forward-looking rate of return in the capital markets." (SBC-CA/Avera 3/12/03, p. 12.) This statement affirms that a valid approach to setting the cost of capital is to look at market returns and apply them using the traditional cost of capital financial modeling exercises that all parties have used. We have in fact done this, although we use our judgment in applying these models and relying on their results. We find that Avera's statements support the concept that the risk of UNEs is the same as that of the company at large, but not greater. We prefer to adhere to the quantitative financial modeling that the parties have offered to determine the cost of capital, tempered with a measure of judgment. It is reasonable to assume that markets have already figured the relative risk of all of SBC's operations, including UNEs, into the returns they require.

Avera maintains that because UNEs are regulated and competitive, they face regulatory risk greater than SBC-CA's other ventures. He argues that UNEs face a "double-whammy" of regulatory constraints and encroaching competitive pressures. SBC-CA argues for a higher cost of capital for UNEs because regulators might get it wrong and not apply the right price to UNEs. In that case, SBC-CA faces competitive risks, although SBC-CA has not shown that competitive risk in the UNE business is greater than the risks in SBC's other competitive businesses.

Further, we agree that SBC-CA faces regulatory risks regarding the accuracy of its TELRIC pricing and business risks due to the potential for rapid technological change and competition for its retail customers. Yet, SBC-CA has not proven that these risks are greater than risk levels in SBC-CA's other business lines which face myriad competitive risks. While the FCC's TRO decision states that increased competition could lead to increased risk and warrant an increased cost of capital,⁷⁸ we are not convinced that UNEs are riskier than SBC-CA's other ventures, and we find that SBC-CA's cost of capital should equate to the cost of capital for SBC as a whole, not that it should be greater than the cost of capital for the entire firm. On balance, we think that the quantitative models capture investors' views of regulatory risks facing SBC-CA's UNE business and there is no need to increase our adopted cost of capital based on this qualitative information.

Regarding comparisons to the returns on equity set for energy utilities, we agree with Murray that these are not relevant due to the energy utilities' differing capital structures, financial conditions, and regulatory policies. As Murray points out, SBC-CA is in excellent financial health and enjoys an AA bond rating, unlike PG&E, which was in bankruptcy during the study period, and Edison, which has teetered on the brink of bankruptcy with bonds rated well below investment grade. (JA/Murray 3/12/03, p. 60.) Nevertheless, we note that the 11.78% return on equity incorporated into our analysis in this order is in fact slightly higher than the 10.9% to 11.6% returns on equity the Commission has set for California's energy companies. Thus, the ROE we use in our analysis for SBC-CA is higher than ROEs for the energy utilities, even though they have faced such great uncertainty and risk with the energy crisis and bankruptcy.

d. Cost of Debt

SBC-CA's Avera estimates the company's cost of debt at 7.18%, based on March 1999 yields on single and double-A corporate bonds as reported by Moody's. (SBC-CA/Avera, 10/18/02, Attachment WEA-1, p.20.) Avera contends that declining trends in short-term borrowing are not indicative of trends in utility capital costs, whereas long-term debt costs have remained largely constant. Avera cites Fall 2002 DRI (now Global Insight) long-term forecasts for double-A utility bonds anticipating an average yield of 7.2% for 2003 and 7.2 to 7.8% over the next 10 years. (SBC-CA/Avera 3/12/03, p. 4.)

Murray maintains that long-term debt costs have decreased since 1999, and forward-looking interest rates are even lower. Therefore, Murray updates Avera's 7.18% debt rate using the current interest rate on 30-year utility bonds. This adjusts Avera's analysis downward by 84 basis points to 6.34%. (JA/Murray 2/7/03, p. 61-2.) For her own analysis, Murray calculates forward-looking debt costs based on historical and forecasted interest rates for 3-month Treasury notes and 10-year Treasury bonds and assumes SBC-CA rolls over short and long-term debt. (JA/Murray 10/18/02, p. 66.) As a result, she estimates a short-term debt cost of 3.18% and a long-term debt cost of 5.51%. (JA/Murray 3/12/03, exh. 5.)

SBC-CA contends Murray's proposed long-term debt cost is too low because Standard & Poor's bond guide reports 7.12% as the yield to maturity on SBC-CA California's 40-year debt, rather than the lower figures cited by Murray. (SBC-CA/Avera 2/7/03, p. 18, n. 30.) Moreover, SBC-CA maintains Murray should not use short-term debt because UNEs are long-lived assets. (Id., p. 24.)

Z-Tel criticizes Avera for using March 1999 debt rates. Z-Tel suggests using December 2002 yields of 10-year AA and A corporate bonds, which average 5.53%. (Z-Tel/Ford, 2/7/03, p. 21.) Avera criticizes Ford for understating the current corporate bond rates. His own citations show double A bond yields at 6.59% for January 2003, and long term forecasts for double A utility bonds at 7.2%. (SBC-CA/Avera 3/12/03, p. 14.)

We find it most reasonable to use in our analysis the current rate applicable for SBC-CA's long-term debt, which is the 6.34% 30-year utility bond rate cited by JA's witness Murray as the update to SBC-CA's original debt figure. We prefer the 30-year utility bond rate over the 40-year debt rate provided by SBC-CA in its reply because the 30-year rate is a closer match to the asset life assumptions incorporated into our model runs. In addition, Murray is merely updating Avera's original number rather than providing an updated number based on longer debt instruments.

We decline to use Murray's analysis, which includes short-term debt costs and rollover of short-term debt to long-term debt at forecasted long-term interest rates. We are not convinced that short-term debt costs have a place in a TELRIC-based cost of capital analysis. The Commission has typically excluded short-term debt when setting the cost of capital and return on equity for utilities. (See D.02-11-027, mimeo. at p. 4.) Furthermore, because we have assumed longer-asset lives for UNEs, we will assume long-term financing to match the asset lives. SBC-CA has argued that this is a reasonable approach and we agree. Short-term rates are more volatile, as Murray herself has noted, and we prefer to base our analysis on the more stable long-term debt costs. Besides, we find Murray's rollover and forecast method not well documented or explained.

Similarly, we will not use Z-Tel's suggested debt costs based on 10-year debt instruments, because we think a 10-year horizon for debt is too short.

e. Capital Structure

SBC-CA's Avera calculates a capital structure for his proxy group of seven ILECs using the market values of common equity and debt outstanding for the group based on year-end 1998 data. This results in a proposed capital structure that is 86% equity and 14% long-term debt. Avera opposes use of short-term debt in the capital structure because he says UNEs are long-lived assets that are not properly matched with capital sources having a maturity of less than one year. (SBC-CA/Avera 2/7/03, p. 24.)

Avera contends that his capital structure based on market value is more forward-looking than a capital structure calculated using book values of debt and equity. Specifically, Avera says:

"A market value capital structure is necessary because telephone companies are operating primarily in a competitive world, where investors focus on market value capital structures.... To be able to raise capital, telephone companies, like other competitive firms, must pay returns that are competitive at the current market prices of their securities, not the embedded book value of the mix of stock and bonds." (Id., p. 20.)

Avera contends that a book value approach has been used for traditional utilities operating within the historical rate-of-return regulatory compact, but it is not appropriate for a competitive firm. Avera cites reports that telephone firms are increasing the equity in their capital structures in the face of mounting business risks. (Id., pp. 22-23.) He cites a Value Line projection that the market value capital structure for SBC-CA will be 15.7% long-term debt and 84.3 % equity. (Id., p. 25.)

Murray criticizes Avera for advocating a "forward-looking" market value capital structure based on outdated 1998 market values. Murray updates Avera's approach of a market value capital structure with more recent financial information for her proxy group and arrives at a capital structure that is 74% equity, 26% debt. (JA/Murray 2/7/03, p. 64.)

In addition, Murray criticizes Avera for relying on a purely market-based value of debt and equity that differs from SBC-CA's internal target capital structure. According to Murray, a market-valued capital structure can become obsolete due to dramatic swings in stock prices, which can make a company's market capitalization volatile. She notes that several of the companies in Avera's proxy group have substantially increased their debt levels in recent years, and indeed SBC, Verizon and BellSouth have increased debt in their market capital structures to an average of 23%. (JA/Murray 2/7/03, p. 57.) She contends it is better to use a capital structure based 50% on market values and 50% on book values, which is less sensitive to changes in market conditions (Id., p. 72). Moreover, Murray maintains that Avera improperly excludes short-term debt from the capital structure, although SBC's book capital structure shows much of its debt is short-term. (Id., p. 75.) Compounding this problem, Avera uses long-term bonds (with maturities longer than 25 years) that have a higher interest rate. Murray contends that this long-term financing is inconsistent with the shorter economic lives proposed by SBC-CA for its assets.

For her own analysis, Murray calculates a capital structure for SBC-CA that is based on an average of book and market values of debt and equity. Murray's updated figures based on averaging book and market values are 57.45% equity, 24.87% long-term debt, and 17.68% short-term debt. (JA/Murray 3/12/03, Exh. 5.) She notes that the Commission has traditionally used a capital structure derived from book value. Other analysts use market capitalization, or a blend. Ibbotson Associates suggest that "[i]deally, a firm's target or optimal capital structure should be used in weighting the cost of equity and the cost of debt." (JA/Murray 10/18/02, p. 68, citing Ibbotson Associates, "Valuation Edition: 2002 Yearbook," at 14.) Murray cites studies that the market value of equity converges toward its book value. (Id., p. 69.) Therefore, she uses what she describes as a conservative approach that favors the higher market value of equity by averaging it with book value. She explains that the results of her approach comport with SBC-CA's own internal target capital structure used in its capital budgeting process. (JA/Murray 3/12/03, p. 68.)

Z-Tel proposes use of SBC-CA's target capital structure, which gives a greater weight to debt levels and includes short-term debt. (Z-Tel/Ford, 2/7/03, p. 21.) Ford cites two sources in support of the use of target capital structure over the firm's current capital structure for valuation purposes. (Id., p. 29.)

We will not adopt the 84% equity and 16% debt capital structure proposed by SBC-CA because we do not find a capital structure to be forward-looking if it is based on market values from 1998.

We will adopt the capital structure that Joint Applicants' witness Murray stated as her estimate of a "forward-looking" market value. This was a capital structure of 74% equity and 26% debt. We decline to follow her recommendation of averaging this "market-based" capital structure with the book value capital structure for the proxy group. The essence of TELRIC is to develop forward looking costs based on the concept of a competitive market. In that situation, relying on capital structures that bear the vestiges of rate of return regulation is fundamentally inconsistent with TELRIC.

We do not agree with Murray that we should use any short-term debt in the capital structure. In our forward-looking analysis of a hypothetical competitive network, we will assume that all debt is long term consistent with our assumptions regarding asset lives.

f. Summary of Weighted Average
Cost of Capital

The results of our analysis are summarized in the table below. In short, we derive the capital structure for our analysis based on Murray's proposed 74/26 forecast of a "forward-looking" capital structure. The 11.78% cost of equity that we use is based on our revisions to the parties' CAPM analysis. We give no weight to the parties' DCF analyses. The 6.34% cost of debt that we use is based on an update to SBC-CA's 30-year debt rate. Altogether, these inputs result in a weighted average cost of capital for SBC-CA of 10.37%.

Table 6

Weighted Average Cost of Capital

Component	Percent of Total	Cost	Weighted Cost
Equity	74.00%	11.78%	8.72%
Debt	26.00%	6.34%	1.65%
	100%		10.37%

G. IDLC/UDLC

A key modeling input involves the technology choice for digital loop carrier electronics. Digital loop carriers (DLCs) are the electronics that connect fiber feeder cable to copper distribution cable, and which allow telecommunications services to pass from copper to fiber and back, and between the fiber feeder and the switch. (JA/Donovan-Pitkin-Turner, 2/7/03, para. 334.)

1. Modeling Proposals

Joint Applicants propose that DLC systems should be modeled as "integrated" or IDLC systems. In an IDLC system, voice signals remain digital all the way from the remote terminal to the switch. JA contend that IDLC is the more recent and forward-looking technology that requires less investment in multiplexing equipment, requires less space, and permits traffic engineering efficiencies. (JA, 8/1/03, p. 2.) According to JA's witness Donovan, an IDLC system can be used to provision a stand-alone unbundled loop at the DS-1 level using an interface known as GR-303. (Id., p. 3.) Further, JA claim that SBC-CA's own engineering guidelines call for greater deployment of IDLC systems. (Id., p. 2.)

In contrast, SBC-CA has modeled DLC systems that are known as "universal," or UDLC. In a UDLC system, voice signals are converted from analog to digital at the remote terminal, then converted back to analog at the central office. SBC-CA takes the position that a forward-looking network must allow a carrier to provide unbundled loops to its competitors and it is not technically feasible in a multi-carrier environment to provision a single, or "stand-alone" unbundled loop using an IDLC system.⁷⁹ SBC-CA does not dispute that it is technically feasible to unbundle IDLC loops at the DS-1 level (Hearing Tr., 4/14/03, p. 430-31), but it argues that this is not the same as providing a stand-alone loop because individual loops cannot be separately identified when embedded in the DS-1 signal. (SBC-CA 8/1/03, p. 5.) SBC-CA contends that various problems prevent provisioning of stand-alone loops over IDLC systems, which include operational, security, and administrative concerns (SBC-CA/Bash Decl., 3/12/03, p. 28). Essentially, SBC-CA says it is unclear how different switches owned and operated by competing carriers can connect to one
DLC system, and it provides a letter from its DLC vendor, Alcatel, in support of these assertions. (SBC-CA 8/1/03, p. 5, citing PHE-103.) Moreover, SBC-CA contends it would be costly to add IDLC capability to its existing switches and that IDLC is not cost effective in many situations. (JA/McNeill 2/7/03, p. 27-28.) Finally, SBC-CA says that some amount of UDLC is needed for circuits that cannot be provisioned over an IDLC system, namely ISDN, high capacity services, and burglar alarms. (Hearing Tr., 4/14/03, p. 448.)

In response, JA maintains that SBC-CA and Telcordia agree it is technically feasible to provide stand-alone loops over IDLC. SBC-CA raises some "operational issues" that JA argue are merely a "smoke-screen" and can be resolved. (JA, 8/1/03, p. 5-6.) ORA/TURN and JA respond that the Commission should ignore SBC-CA's claims that provisioning loops over IDLC is more costly because these arguments ignore the proper interpretation of TELRIC and are "based on the work involved with replacing UDLC with IDLC in the existing - embedded - network, not on costs of IDLC versus UDLC in the context of the forward looking network that is appropriate in a TELRIC study." (ORA/TURN 8/1/03, p. 7.) ORA/TURN support JA's proposed assumption of IDLC technology because they believe that JA's proposals during the evidentiary hearing could resolve the operational issues cited by SBC-CA. (Id., p. 4.)

XO states that SBC-CA's DS-1 loop study does not use any IDLC technology, even though SBC-CA admits that it provisions DS-1 loops using IDLC. XO proposes that an assumption of 20% IDLC usage should be used as a modeling input when calculating DS-1 loop costs. (XO 2/7/03, pp. 40-41.)

2. Discussion

First, we find that UDLC is the forward-looking technology choice to include in our model runs. We find dispositive the statement of the DLC vendor, Alcatel, in support of SBC's assertion that it is unclear how different switches owned and operated by competing carriers can connect to one DLC vendor. Moreover, we note that JA's witness Donovan admitted that he does not know of a stand-alone loop provisioned over IDLC by any carrier in the entire country. (SBC-CA, 8/1/03, p. 5; Hearing, Tr., 4/14/03, p. 439.) It is not reasonable to find that ULDC is the forward looking technology for providing a single unbundled loop. Moreover, it is not reasonable to develop costs based on JA's and ORA/TURN's optimism that the operational issues will be resolved. In a TELRIC analysis such as this one, however, we must adhere to what is currently available and technically feasible.

As a result, we adopt the a mix of IDLC and UDLC systems proposed by SBC-CA - 95% UDLC and 5% IDLC.

H. DLC Costs

Both models assume a forward-looking loop design that incorporates digital loop carrier electronics into the loop plant. The models differ in the engineering, furnishing and installation (EF&I) costs for above-ground cabinets, or "remote terminals" (RTs) and underground "controlled environmental vaults" (CEVs) that house these DLC systems.

1. JA Approach

The Joint Applicants maintain that the Commission should use their proposed costs of $51,425 for the installation of DLC equipment in a 6x16 CEV, and $5,740 for installation in a 1,016-line capacity above-ground RT. (JA 8/1/03, p. 32, and Table C-1, p. 37.) According to JA, DLC equipment is pre-assembled at the factory and there are only a "handful of tasks necessary to place and connect the largely pre-assembled DLC systems that SBC-CA California purchases from Alcatel." (Id., p. 32.) JA argue that SBC-CA's contract for DLC equipment with Alcatel contains numerous references to extensive installation requirements placed on Alcatel. Therefore, JA argue that given these contract terms SBC-CA incurs virtually no DLC installation costs and it would be improper to model any. Nevertheless, JA have included the costs described above in the event that the contract does not cover all installation activities.

In response to this proposal, SBC-CA maintains that its contract with Alcatel does not include installation and that SBC-CA incurs significant costs for the installation of the equipment on site, either by its own personnel or other contractors. (SBC-CA 8/1/03, p. 20.) SBC-CA maintains that JA have omitted key field installation activities from their analysis such as costs for testing the system, copper and fiber splicing, power connection costs, engineering, construction management, transportation, and right-of-way acquisition. (SBC-CA 8/22/03, pp. 26-27.) JA's witnesses acknowledged that "in the way that it appears that SBC-CA uses [the Alcatel contract], there is a separate installation cost." (Hearing Tr., 4/15/03, p. 618.)

2. SBC-CA Approach

In LoopCAT, SBC-CA proposes a factor-based approach to estimate DLC installation costs, based on the ratio of installation costs to material costs. SBC-CA used recent actual data reflecting a mix of DLC installation projects to calculate the relationship between installation and material costs. Material costs are multiplied by this factor in SBC-CA's model to estimate EF&I costs. (SBC-CA/Smallwood 3/12/03, p. 26.) The factor and the costs it produces are proprietary information, but the factor results in DLC installation costs ranging from $300,000 to $500,000, for RTs and CEVs, respectively. (JA 8/1/03, p. 37.) This range is orders of magnitude greater than the $5,740 to $51,425 estimated by JA.

JA criticize SBC-CA's EF&I factor for DLC installation as unsupported, grossly inflated, and out of sync with several other sources of data that allegedly show SBC-CA's actual DLC installation costs. (Id., p. 58; JA 8/22/03, p. 53.) First, JA contend that SBC-CA did not provide the accounting data underlying its factors. Second, SBC-CA's proposed installation costs equate to over 6000 hours of work, or two technicians working full time on the project for one and three-quarter years. According to JA, this contradicts Alcatel's own equipment installation instruction touting the "ease of installation" of DLC remote terminal equipment and the views of JA's expert witnesses that DLC systems can be installed in only a few weeks. (See PHE-17 p. PBRL-011829; JA 8/1/03, p. 33.) Third, JA compare SBC-CA's DLC installation factor to actual cost data provided by SBC-CA, noting that this data shows actual DLC installation costs are significantly lower than LoopCAT factors and the ratio calculated from SBC-CA's actual jobs is only 18% to 26% of the factor used in LoopCat.⁸⁰ (JA 8/1/03, p. 55.) SBC-CA admits that its actual DLC installation costs from a sample of jobs are lower than the factors developed for LoopCat. (SBC-CA 8/1/03, p. 21.) Finally, JA attack the credibility of SBC-CA's witnesses on this topic based on their initial inability to estimate actual DLC installation costs, and their repeated statements that they have no knowledge of the DLC factors used in LoopCAT or what is supposed to be reflected in a TELRIC study. (JA 8/1/03, p. 55.)

3. Discussion

We find that SBC-CA does incur some installation costs above and beyond those included in the contract with Alcatel. SBC-CA's witness Palmer explained that while the contract provides that Alcatel may perform some installation, the contract does not contain prices for this because SBC-CA has not chosen this option. (SBC-CA/Palmer Declaration, 3/12/03, p. 5.) In our model runs, we will include costs for the on-site DLC installation work that SBC-CA must incur.

It is at this point that we face dueling views of DLC installation costs. JA have assumed a least-cost scenario for all installations that is below the actual DLC installation costs provided by SBC-CA. SBC-CA has proposed an EF&I factor that is above the actual costs it provided, and it cannot adequately explain the difference between its factor and its actual costs.

We find SBC-CA has not met its burden of proof that the factor in its model is accurate, particularly given its actual cost information showing a much lower ratio of EF&I costs to material costs. SBC-CA's witnesses could not satisfactorily explain how the LoopCAT EF&I factor for DLC installation was derived. Although SBC-CA supplied three witnesses on the topic of DLC costs, none of these witnesses could reasonably explain how the LoopCAT factor was derived or how it relates to actual DLC installation costs. While SBC-CA devoted many pages to criticizing HM 5.3's DLC assumptions as too low, and defending its own EF&I factors, SBC-CA failed to provide a reasonable explanation of how SBC-CA's DLC EF&I factor was created. JA charge that SBC-CA's factor includes costs for items such as poles, conduit, and copper and fiber placement that are not appropriate to include here. SBC-CA has not been able to show that these costs are not double counted.

Further, SBC-CA's witnesses on this topic lacked credibility and appeared to operate in silos rather than as a team, deferring questions to another witness and professing little knowledge on the specific question at hand. JA contended that SBC-CA's expert on LoopCAT was unfamiliar with DLC actual costs, while the experts on actual costs could not explain what was assumed in LoopCAT. (JA 8/1/03, p. 35.) We find this to be an accurate criticism. SBC-CA witness Palmer contended that HM 5.3 DLC estimates were too low and did not match actual DLC costs, but he admitted that he did not know actual DLC installation costs. (Hearing Tr., 4/15/03, pp. 572-573.) SBC-CA's witness Bash stated that she thought LoopCAT DLC factors were based on averages across SBC-CA, but she admits she did not have direct input, and only provided guidance. (Id., p. 573.) Later, when asked about specific costs that SBC-CA may have included in its DLC EF&I, Bash stated she did not know what was in the EF&I loading. (Id., p. 586.) Ultimately, we cannot accept SBC-CA's EF&I factor because SBC-CA's cost witness Smallwood says he relied on actual information from recent DLC installations, but witnesses Bash and Palmer from whom he obtained this information cannot explain why actual DLC installation costs do not match the factor.

Therefore, given the record before us, we will not rely on SBC-CA's DLC factor and we will use actual cost information provided by SBC-CA rather than the bottom-up approach advocated by JA. We find this approach is more conservative and representative of SBC-CA's forward-looking DLC installation costs than JA's approach, which assumed an RT could be connected in two weeks for less than $6000. While JA advocate adjustments to SBC-CA's actual costs, we prefer to use the simple averages of costs resulting from SBC-CA's sample of 50 DLC installations, noting that there is a wide range of results here.

When we ran the SBC-CA models, we replaced SBC-CA's EF&I factor with the factor derived from the average of SBC-CA's 50 installations. This factor is about one-quarter of SBC-CA's original factor. (See JA 8/1/03, p. 43, and Exhibit C-4 and C-5.) SBC-CA supports use of this number for RT and CEV installations, and we use it along with SBC-CA's proposed new factor for installation of central office terminals. (SBC-CA, 8/22/03, p. 32.) When we run our version of HM 5.3, we assume EF&I costs for RTs of $22,814 and for CEVs of $49,569 based on SBC-CA's sample data. (JA 8/1/03, p. 43, and Exhibit C-4 and C-5.)

In comments on the Proposed Decision, AT&T claims the Commission erred in applying these actual RT and CEV costs to all sizes of RT and CEV installations. AT&T claims these actual costs should only apply to larger RT and CEV installations, and the Commission should scale these numbers down for smaller installations. (AT&T, 6/1/04, p. 18.) We make no change in our use of these actual RT and CEV installation costs and we will not scale them down for smaller RT and CEV projects as AT&T suggests. Just as we have not agreed that more costly equipment always costs more to install, we do not agree that less costly equipment always costs less to install. We will use these actual costs from average size RT and CEV projects as proxies of installation costs for all sizes of RTs and CEVs because we find the costs modeled by both SBC-CA and HM 5.3 in this area are not reliable.

I. Fill Factors in Loop Model

The parties have varying proposals for the amount of spare capacity that should be designed in a forward-looking local exchange network. In TELRIC cost models, designing a network with spare capacity entails use of a "fill factor," or utilization level, as a modeling input. For example, a fill factor of 40% means that 40% of the physical plant is in use, while 60% is available for spare capacity, or growth. (See D.96-08-021, mimeo. at 23.)

As the FCC stated in its First Report and Order,

We conclude that, under a TELRIC methodology, incumbent LECs' prices for interconnection and unbundled network elements shall recover the forward-looking costs directly attributable to the specified element, as well as a reasonable allocation of forward-looking common costs. Per-unit costs shall be derived from total costs using reasonably accurate "fill factors" estimates of the proportion of a facility that will be "filled with network usage); that is, the per unit costs associated with a particular element must be derived by dividing the total cost associated with the element by a reasonable projection of the actual total usage of the element. (First Report and Order, para. 682.)

Key fill factors used in the loop model determine the appropriate investment for copper distribution cable, fiber feeder facilities, copper feeder facilities, DLC equipment, serving area interfaces (SAIs), and premise termination equipment. In the prior OANAD proceeding, the Commission adopted fill factors of 38% for distribution cable and 76% for copper feeder cable. (D.96-08-021, pp. 29-32.)

According to JA, fill factors are a major concern in modeling forward-looking costs, especially distribution cable costs. The lower the fill factor, the more spare, or excess, capacity will be included in the cost study. Therefore, distribution plant costs are inflated at lower fill percentages. (JA/Donovan 10/18/02, p. 53.) JA also express concern that fill factors incorporate an accurate match of investment and demand projections. JA's witness Klick describes how TELRIC models should derive per unit costs by dividing total costs by a reasonable projection of actual total usage. If the numerator of the calculation reflects investment large enough to accommodate line growth, but the denominator ignores line growth by only looking at customers served currently, there is a significant mismatch. As a result, costs per line may be too high because they include more investment than necessary to serve today's customers. (JA/Klick 10/18/02, pp. 15-17.)

SBC-CA has used current fill levels derived from its actual network operations in its modeling, based on the assertion that "current fill levels represent forward-looking fill because SBC-CA's network is efficiently designed." (SBC-CA/Bash 10/18/02, p. 19.) SBC-CA states that "the current fills based upon historical evolution of the network infrastructure are the most reasonable projection of efficient future usage for each of the loop plant components in question, and therefore, comprise the appropriate fill." (Id., p. 19.) SBC-CA's witness Bash says that volatility and uncertainty in demand from "churn" in customers requires fill factors at lower levels than those proposed by JA, in order to serve "ultimate demand."

ORA/TURN criticize SBC-CA's fill factors as too low, particularly when compared to the fill factors SBC-CA assumes in its TSLRIC cost studies for pricing flexibility purposes. According to ORA/TURN, it is improper to incorporate a fill factor based on efficient network utilization into a TSLRIC study, but use embedded or historical network utilization for a TELRIC study. ORA/TURN see no reason to use two different fill factor methodologies, other than to achieve lower cost in a TSLRIC study for maximum pricing flexibility and to achieve a higher cost in a TELRIC study for higher UNE rates. ORA/TURN contend that SBC-CA should not be allowed to strategically select fill factors based on the nature of the end purpose of the cost study. (ORA/TURN/Roycroft Declaration, 2/7/03, pp. 37-38.) SBC-CA responds that it is appropriate for TELRIC to use current measures of average fill, and for a TSLRIC study to use design fill, or "fill at relief." The difference between the two fill measures is a fixed cost which should be shared, and shared costs are not included in TSLRIC studies. (SBC-CA/Tardiff 3/12/03, pp. 34-35.)

In general, we do not agree with SBC-CA's assertion that fill levels derived purely from current network operations are automatically forward-looking. We agree with JA that network enhancements such as Project Pronto or other one-time occurrences that are captured in the SBC-CA's actual operational data could skew fill levels at a point in time to less than optimal levels. Further, we agree with ORA/TURN that there appears to be a wide disparity between the fill levels SBC-CA proposes here versus those used in its TSLRIC studies. While we will not render an opinion on SBC-CA's assertion that different fill levels are appropriate for TELRIC and TSLRIC studies, we find that the large difference calls into question whether SBC-CA's current fill levels are actually forward-looking.

Furthermore, we agree with JA that it is important that fill factors reflect accurate projections of both investment to accommodate growth and a reasonable estimate of demand. We also agree with SBC-CA's witness Tardiff that an efficient carrier always carries spare capacity, and the cost of this spare is appropriately reflected in TELRIC costs. (SBC-CA/Tardiff, 2/7/03, pp. 9-10.) We will keep these principles in mind as we evaluate the fill factors proposed by the parties. We prefer to look at each fill level individually to determine whether SBC-CA or other parties have the better argument for a reasonable forward-looking fill level. We discuss each of the significant fill factors in the loop portion of the model separately below.

1. Copper Distribution Fill

This fill factor relates to the amount of copper facilities, or line pairs, that are modeled in the distribution network. JA propose cable sizing inputs that result in an "achieved fill"⁸¹ rate for copper distribution of 51.6%, while SBC-CA proposes an achieved fill level of 41.7%.⁸²

a. JA Proposal

JA contend that many significant changes warrant a reexamination of the 38% copper cable distribution fill factor adopted in the prior OANAD proceeding, such as the accelerated deployment and availability of DSL services which reduce the need for second lines, SBC-CA's updated engineering guidelines stressing increased fill levels, and guidance from the FCC on reasonable fill factors. (JA/Donovan 10/18/02, paras. 101-123.) According to JA, HM 5.3 uses engineering guidelines, based on a realistic assessment of outside plant design and available cable sizes, to model a level of fill sufficient to serve SBC-CA's current demand plus reasonably foreseeable demand growth.

Donovan explains that JA's approach to calculating fill is markedly different from the SBC-CA approach. HM 5.3 does not start with an achieved fill factor as an input in the way that LoopCAT does. Instead, HM 5.3 uses "sizing factors" which are inputs in the model to determine the minimum number of cables or fibers necessary to meet current demand plus a cushion for spare. (Id., p. 49.) According to JA, "The use of sizing factors as an input instead of achieved fill is appropriate for a bottoms-up model like HM 5.3 because the model constructs the network first based on sound engineering guidelines rather than building to achieve a certain level of fill." (JA 3/12/03, p. 64.) JA use SBC-CA's current engineering guidelines to design cable sizes to build 1.5 to 2 lines per living unit for residential customers. (JA/Donovan 10/18/02, p. 53.) Donovan contends that the HM 5.3 proposed fill factor of 52% can accommodate an assumed demand growth rate of 3% in order to serve all demand over an assumed 22 year economic life of distribution cable. (Id., p. 54.) In other words, when HM 5.3 models a network with a 52% fill level, the network has 48% spare capacity, or almost double the number of copper loops that are required to serve current demand.

SBC-CA criticizes JA fill rates as too high, with not enough spare capacity. (SBC-CA 2/7/03, p. 55.) SBC-CA says JA's cable sizing factors ignore standard network design of 2.25 lines per living unit, and therefore, would not allow SBC-CA to serve "ultimate demand" or meet service quality standards set by the Commission. Bash says that JA incorrectly use temporary guidelines for rural areas to justify 1.5 to 2 lines per housing unit, when for California, SBC-CA's current guideline is 2.25 lines per lot. (SBC-CA/Bash 3/12/03, pp. 20-21.) SBC-CA witness Murphy says that HM 5.3's cable-sizing factors are artificial and ignore sizing an actual network to accommodate churn, growth, maintenance, and administrative needs. (SBC-CA/Murphy 2/7/03, pp. 51-53.) McNeill says outside plant should be sized to meet potential demand of a given area because it is more cost effective to place additional capacity at the time of initial placement rather than at a later date. Further, he contends that higher fill levels are correlated to increased maintenance-related activities and longer service intervals. (SBC-CA/McNeill 2/7/03, pp. 18-20.)

All in all, SBC-CA maintains that JA have not proven the wisdom of raising the copper distribution fill factor above the 38% level adopted in the prior OANAD proceeding, particularly when HM 5.3 models far fewer pairs than exist in SBC-CA's network today. (SBC-CA 2/7/03, p. 20.)

b. SBC-CA Proposal

SBC-CA contends that current utilization rates, which are calculated by dividing working pairs by available pairs, are optimal because they were developed under the incentives of price cap regulation and are the best predictor of future utilization levels. (SBC-CA/Bash 10/18/02, p. 22.) According to SBC-CA, higher fill levels cause delays in service, service quality degradation, and higher installation costs. (SBC-CA 3/12/03, p. 43.) Instead, the proper level of spare remains constant over time because as SBC-CA states, "although some spare is used over time, additional spare is always being added so that on average the fill rates proposed by SBC-CA California are achieved. (Id., p. 44.) SBC-CA contends that the FCC supports use of actual network utilization in TELRIC models when it states that fill factors must be based on a "reasonable projection of the actual total usage of the element." (SBC-CA/Aron, 3/12/03, p. 46, citing First Report and Order, para. 682.)

JA and ORA/TURN respond with several criticisms. First, SBC-CA's actual fill factors are not forward-looking because they capture all of the cable facilities currently in SBC-CA's database, rather than reflecting the efficient amount necessary in a forward-looking environment. (JA/Donovan-Pitkin-Turner, 2/7/03, p. 69.) JA say that SBC-CA's newest guidelines plan for less than 2.25 lines per living unit, and SBC-CA's actual fill levels do not incorporate the new, lower guidelines. (Id., p. 133.) Further, SBC-CA has ignored the fact that growth is less today than when the capacity was installed, in part because of DSL line-sharing and broadband reducing the need for additional lines. Similarly, SBC-CA's current fill levels reflect the duplicative installation of copper and fiber facilities during the transition to a fiber-based network. (Id., p. 79.)

Second, JA and ORA/TURN claim that SBC-CA's approach to calculating fill violates FCC pronouncements that fill factors should reflect current demand and not the industry practice of building distribution plant to meet "ultimate demand" because it is too speculative. (ORA/TURN/Roycroft 3/12/03, para. 58; JA/Donovan 3/12/03, p. 84-87.) According to the FCC:

We find unpersuasive GTE's assertion that the input values for distribution fill factors should reflect ultimate demand. In concluding that fill factors should reflect current demand, we recognized that correctly forecasting ultimate demand is a speculative exercise, especially because of rapid technological advances in telecommunications... Given this uncertainty, we find that basing the fill factors on current demand rather than ultimate demand is more reasonable because it is less likely to result in excess capacity, which would increase the model's cost estimates to levels higher than an efficient firm's costs... (Inputs Order, para. 200.)

JA cite to several recent FCC orders where fill factors of 40% or lower were criticized, and where adopted fill rates were in the 50 to 75% range. (JA/Murray, 2/7/03, para. 35.) ORA/TURN further cite to the FCC's statement that "We find that a fill factor that assumes that more than two-thirds of capacity is idle for an indefinite time is unreasonably low." (Kansas 271, para. 80; cited by ORA/TURN/Roycroft Decl., 2/7/03, p. 43.)

Third, JA and ORA/TURN maintain that SBC-CA's varying fill levels for urban, suburban and rural areas are illogical. JA and ORA/TURN show that LoopCAT's results are lowest in urban zones, and higher in rural and suburban zones for DLC and copper cable. (JA/Donovan-Pitkin-Turner, 2/7/03, p. 80.) This is illogical because equipment in urban areas can be changed out more quickly, so fill in urban areas should be higher than rural fill. In contrast, rural equipment is placed before it is actually needed so it has a lower fill. (Id., pp. 79-80.)

c. Discussion

Both SBC-CA and JA suggest changes to the 38% copper distribution fill factor adopted in our prior OANAD decision. We will adopt the fill factor of 41.7% proposed by SBC-CA for our model runs. We find this fill factor is reasonable because it provides an adequate level of spare capacity to accommodate a reasonable projection of future demand and reflects current engineering standards. Further, we find that it is reasonable to use a higher fill factor than the prior OANAD given FCC decisions providing guidance in this area since the Commission's 1996 decision, and given trends in network usage, such as the availability of DSL technology, cable modems, and wireless technologies, that have reduced line growth projections.

There are several reasons why we find that SBC-CA has met its burden of proving that its embedded fill level is a reasonable proxy for forward-looking utilization. The fact that SBC-CA has maintained the same fill level over time suggests that level is either reasonably efficient or all that current technology makes possible.

Second, copper is a technology that phone companies have now used for approximately 100 years, and under competitive pressures for approximately a decade. Thus, we have more confidence in SBC-CA's projection of fill levels for copper than for other loop materials.

Third, we are somewhat persuaded that a fill level of 51.6% will cause dramatic service delays or installation cost increases, as suggested by SBC-CA. . Moreover, a fill level of 51.6%, is a full 25% above the fill level proposed by SBC-CA and is premised on the installation of only 1.5 to 2 lines per household. It is not reasonable to conclude that this level of spare can accommodate customer churn, maintenance, and growth without the need for service interruptions or the installation of additional lines.

Finally, we do not agree with JA and ORA/TURN that SBC-CA's fill factors that are lower for urban zones and higher for rural areas are illogical. Indeed, this outcome comports with SBC's actual experience in California.

2. Copper Feeder

HM 5.3 inputs provide an achieved fill rate of 77.5% for copper feeder, which is almost identical to the 76% rate adopted in the prior OANAD order. SBC-CA proposes a fill rate for copper feeder of 66.2%.

SBC-CA's proposed fill rate for copper feeder is based on its current network experience. According to SBC-CA, "network fills have been stable over time and represent the best estimate of utilization on a going-forward basis." (SBC-CA/Smallwood 3/12/03, p. 45.) JA respond that SBC-CA's actual records reflect a work in progress, namely the transition from a basic exchange network to the upgraded Project Pronto DSL-capable network, which deliberately places duplicate facilities. In other words, when SBC-CA places fiber, it does not remove copper. (JA/Donovan-Pitkin-Turner 2/7/03, p. 72.) Thus, SBC-CA's actual records reflect lower usage of copper feeder caused by duplicative facilities. JA contend that if actual fill levels are used for TELRIC modeling, fill is understated, costs are overstated, and the model does not accurately reflect that only one feeder technology would be used in a forward-looking network to serve each distribution area. (Id., p. 73, and paras. 141-156.) SBC-CA responds that even with the addition of fiber facilities under Project Pronto, SBC-CA's copper feeder utilization levels have remained constant because the incremental capacity made available was small. (SBC-CA/Bash 3/12/03, p. 24.)

JA claim that the 77.5% copper feeder fill rate proposed in HM 5.3 allows for 2.5 years of growth in feeder usage. (JA/Donovan-Pitkin-Turner, 2/7/03, p. 137.) SBC-CA criticizes this fill rate as unreasonable because SBC-CA has never had a copper feeder utilization level above 70%. When levels did approach 70%, SBC-CA experienced increased network maintenance costs. (SBC-CA/McNeil, 2/7/03, pp. 19-21.)

ORA/TURN echo the views of JA that SBC-CA's copper feeder utilization rates, which are based on actual network fill rates, are unreasonably low. ORA/TURN note that SBC-CA has argued to the FCC that fills in excess of 82.5% often result in plant rearrangement costs, which implies that the fill levels proposed by SBC-CA can be raised without resulting in higher costs for provisioning spare capacity. (ORA/TURN/Roycroft 2/7/03, p. 34, n. 27, citing SBC comments in the FCC's 1999 Universal Service docket.)

Discussion. Once again, we find SBC-CA's long term experience with copper wires dispositive, and expect that in the future fill rates will continue to look like current fill rates. Thus, we adopt a fill rate for copper feeder of 66.2% based on the observed fill factor in SBC-CA's network.

3. Fiber Feeder

JA propose sizing factors that achieve a fill level for fiber feeder of 79.6%. SBC-CA proposes a fill level for fiber feeder of 16.22%.

According to JA, HM 5.3 models 4 fibers to each DLC site, two of which are for redundancy. HM 5.3 then uses the next largest fiber cable size available, which results in additional spare capacity at each DLC site. (JA/Donovan 10/18/02, p. 50.) As a result, the proposed fiber fill rate of 79.6% includes duplicative facilities, so the effective utilization of the fiber strands is at or below 50%. (Joint Comparison Exhibit, 12/3/03, p. 3, contained in Attachment 4 of ALJ Ruling of 4/1/04.) JA claim that this is consistent with the approach endorsed by the FCC in its universal service proceeding, and supported by SBC-CA at the federal level as well. (JA/Donovan 3/12/03, p. 92, citing Inputs Order, para. 208.) Moreover, JA contend that its fiber feeder fill rates are based on the engineering concept of "fill at relief," as documented in SBC-CA's loop deployment guidelines, but reduced to allow for 2.5 years of feeder growth. (JA/Donovan-Pitkin-Turner, 2/7/03, pp. 134-137.)

SBC-CA characterizes JA's fiber feeder fill inputs as a 100% cable sizing factor and criticizes this as unrealistic because it sizes the network perfectly to meet demand today, but leaves no administrative spare capacity to perform maintenance and accommodate customer moves and relocations. (SBC-CA/Murphy 2/7/03, p. 53.) SBC-CA also contends that the feeder fill rate proposed by JA has never been achieved and operating the network at that fill level would inflate "network health costs." (SBC-CA/McNeill 2/7/03, p. 20.) JA respond that SBC-CA misinterprets the 100% sizing factor in the model as a 100% fill rate. Rather, JA contend that because HM 5.3 models two redundant fibers to each site, it results in an achieved fill of 50% or less. (JA/Donovan 3/12/03, p. 91.)

In contrast, SBC-CA once again proposes a fill rate for fiber feeder based on its actual experience, which it says captures the true fiber utilization in its network today and is designed to accommodate growth over a five-year period. (SBC-CA/Bash 10/18/02, p. 20.) SBC-CA maintains that the appropriate fill level must consider the capacity of the DLC equipment connected to the fiber, a concept it calls "channel fill." (Joint Comparison Exhibit, 12/3/03, p. 4.) Specifically, SBC-CA arrived at its fiber fill rate of 16.22% by multiplying the percentage of "lit" fiber strands that are actually in use by the average utilization of the working strands. (SBC-CA/Smallwood 3/12/03, p. 58.)

JA counter that SBC-CA's calculation of fiber feeder fill incorrectly determines the percent of active fiber strands needed to serve a DLC system, and thereby significantly overstates fiber costs. (JA/Donovan-Pitkin-Turner, 2/7/03, p. 186.) JA maintain that SBC-CA's "channel fill" calculation is meaningless because fiber strands do not have channels and can accommodate nearly unlimited capacity depending on the electronics deployed on each end of the fiber. (Joint Comparison Exhibit, 12/3/03, p. 4.) ORA/TURN states that SBC-CA's fiber feeder fill rates are unreasonably low, particularly given SBC-CA's own advocacy to the FCC for fiber fill factors as high as 100%, as long as fiber redundancy is maintained to allow upgrades and equipment modifications without disrupting customer service. (ORA/TURN/Roycroft, 2/7/03, p. 34.)

Discussion. For our model runs, we will use the fiber feeder fill rate proposed by JA because it mirrors the approach used by the FCC in its modeling, and it provides full redundancy and spare capacity for 2.5 years of growth.

We find that SBC-CA's proposed fiber feeder fill level of 16.22% is not forward-looking for several reasons. First, SBC-CA's proposed fill rate is not consistent with the FCC's findings in its universal service cost modeling, which SBC-CA supported. Second, we find SBC-CA's discussion of "channel fill" unclear. While SBC-CA's discussion of channel fill is no doubt an accurate depiction of the percentage of fiber strands actually in use, we are not convinced that the channel fill concept is useful or relevant to designing and deploying forward-looking fiber facilities. The calculations SBC-CA describes appear to be useful in determining what percent of fiber strands are actually used today. Nevertheless, SBC-CA has not met its burden of proving that an efficient forward-looking network would necessarily be designed to achieve this same usage level of fiber strands. In other words, SBC-CA has not justified why it makes sense to design a network with more than 80% spare capacity in fiber facilities.

Finally, SBC-CA's fill rate is contradicted by statements of its own witnesses. According to Bash, the optimal fill rate for feeder plant is higher than for distribution because feeder is usually placed in underground conduit, on existing poles, or buried along major rights-of-way, which makes it easier to reinforce. (SBC-CA/Bash 10/18/02, p. 20.) SBC-CA's witness McNeil contends that standard engineering principles recognize that because feeder facilities cost less per unit of length than distribution facilities, the objective is to minimize the size of the DA and achieve a reasonable fill of the feeder facilities. (SBC-CA/McNeil 2/7/03, p. 16.) However, SBC-CA has proposed a fiber feeder fill rate less than half of its copper distribution fill rate, without explaining this inconsistency. For these reasons, we reject the 16.22% fill rate proposed by SBC-CA and adopt the JA's proposed fiber feeder fill rate.

4. DLC Common Equipment

JA propose a fill level for DLC common, or "hard-wired," equipment of 74.6%. SBC-CA proposes a fill level of 47.4%.

In the HM 5.3 model, the investment required for DLC common equipment starts with the number of lines for that DLC, inflates that by a "channel unit sizing factor," and then chooses the next larger DLC Remote Terminal size to match the number of lines. HM 5.3 incorporates a choice of DLC systems from 24-line up to a maximum of 8,064 lines in a controlled environmental vault (CEV). (JA/Donovan 10/18/02, pp. 59-60.) JA contend that a 75% fill level allows 3% growth for 10 years (Id., p. 10).

In contrast, SBC-CA proposes a DLC fill factor based on its actual network operations. Like JA, SBC-CA has sized DLC common equipment to allow for ten years of spare capacity at the outset. (SBC-CA/Smallwood 3/12/03, p. 54.) However, SBC-CA's ten-year projection involves more spare capacity than presumed by JA. According to SBC-CA, JA's common equipment fill level ignores real-world constraints by modeling too many DLC sizes. SBC-CA contends that JA's assumptions result in less spare than is required in actual operating conditions, by assuming that each DLC can be perfectly sized to match the number of lines. SBC-CA claims that sufficient reserve must be maintained to allow for the various services supplied within the DLC. (Id., p. 53.) In addition, SBC-CA claims that high reinforcement costs make it more economical and practical for SBC-CA to size DLC remote terminals for a ten year period when it initially places them. (Id. p 54.)

JA criticize SBC-CA for modeling only four DLC configurations, which does not reflect the range of DLC sizes SBC-CA actually deploys. JA contend SBC-CA's modeling choices are less efficient, understate DLC capacity, and lead to more spare capacity than is required. (JA/Donovan-Pitkin-Turner, 2/7/03, pp. 176-177.) Specifically, JA maintain that LoopCAT does not use the correct line capacity for a 6x16 CEV, and SBC-CA admits to this. (Id., para. 356-7.) In addition, JA allege that SBC-CA has double counted fill factors in its costs for CEV structures by assuming a fill factor for the CEV structure itself, on top of the common equipment fill factor. (Id., para. 362.) Finally, JA contend that SBC-CA's historic fill levels for DLC equipment are not indicative of an efficient, forward-looking utilization level because SBC-CA has placed new fiber and DLC facilities on top of copper plant, but is only gradually moving customers to DLC facilities. (Id., para. 141.)

In comments on the Proposed Decision, AT&T contends that the common equipment fill level is a modeling output, and its results depend on the chosen level of DLC plug-in fill. AT&T explains that the primary input that affects DLC common equipment is the DLC plug-in sizing factor, which is set at 75% in Section VI.E.5 below. When DLC plug-in equipment is modeled at 75% utilization, DLC common equipment is then sized to meet this level of demand. Larger DLC common equipment is modeled, and this creates more spare capacity and a lower common equipment achieved fill level. The bottom line, according to AT&T, is that when plug-in equipment is modeled at a 75% fill factor, the resulting DLC common equipment fill level in HM 5.3 is 62%. (AT&T, 6/1/04, p. 15.)

Discussion. We will use a common equipment fill factor of 47.4% proposed by SBC-CA for several reasons. First, JA's proposed 74.6% fill factor does not allow for sufficient growth over a 10-year period, which SBC-CA agrees is a reasonable growth horizon. Second, we find SBC-CA's real world experience in currently installing this equipment dispositive and believe that its future forecast more accurately reflects a forward-looking deployment of this technology.

We note that in HM 5.3, it is not possible to independently set this value - the model derives this factor from the DLC Plug-In Equipment number. Thus, we note that our HM runs depart from our adopted fill rate.

5. DLC Plug-In Equipment

JA propose a fill level for DLC plug-in equipment, i.e., line-cards, of 89.9%. SBC-CA proposes a fill level of 53.1%.

JA support their proposal by contending that their 89.9% sizing factor provides adequate spare for a six to twelve month period, in line with SBC-CA's engineering guidelines which stress minimizing spare plug-in levels. (JA/Donovan 10/18/02 p. 10; JA/Donovan 3/12/03. p. 95.) JA maintain that their proposed fill level is reasonable because DLC plug-in equipment can be placed on an as needed basis. (JA/Donovan-Pitkin-Turner, 2/7/03, para. 350-52.) Further, they note that SBC-CA admits that plug-in equipment is placed to accommodate growth over a six to twelve month period. (SBC-CA/McNeil 2/7/03, p. 22, n. 29.)

SBC-CA's proposal derives from its claim that current fill levels based upon historical utilization of DLC equipment are the best estimate of forward-looking usage, since fill levels have remained static over time. (SBC-CA/Bash 10/18/02, p. 21.)

SBC-CA contends that JA's proposed fill is unrealistic because it ignores inventory management and travel time whenever service orders are placed. According to SBC-CA, there is a trade-off between deploying spare line cards up front and bearing the expense of extra technician trips each time a line card needs augmenting. (SBC-CA/Smallwood 3/12/03, pp. 54-55.) Along the same lines, SBC-CA contends it is unable to manage DLC channels on a single pair basis, and the highest DLC utilization level SBC-CA ever achieved was only slightly above 70%. (SBC-CA/McNeil 2/7/03, pp. 21-22.)

Discussion. We will use a DLC Plug-In fill factor of 53.1% for our model runs. Once again, we find SBC's real world experience obtained through currently installing this new equipment as dispositive.

.

6. SAI Equipment Fill

Serving Area Interfaces (SAIs) refers to the equipment in the loop network that connects feeder and distribution facilities. JA propose a fill factor for SAI equipment of 67.8%, while SBC-CA proposes a fill factor of 47.2%.

The fill level proposed by JA is based on modeling 3.5 lines per residential living unit (two of which are for distribution termination, and 1.5 for feeder termination), and two lines per business. (JA/Donovan 3/12/03, para. 95, n. 61.) In addition, HM 5.3 models SAI equipment sizes that are currently available and may be larger than equipment in place in SBC-CA's network. As JA's witness Donovan explains, SBC-CA's current equipment is based on 1972 guidelines to size SAI equipment for 200 to 600 living units, but equipment available today can serve many more living units. (Id., p. 43.)

SBC-CA criticizes JA for modeling too many SAI sizes, including sizes that are rarely used, thereby artificially reducing the spare that is required in a real-world network. (SBC-CA/McNeill 2/7/03, p. 16.) McNeill maintains that ILECs still use the 1970 era guidelines of sizing distribution areas to serve 200 to 600 living units. He explains that:

Because feeder facilities cost less per-unit of length than distribution facilities, the objective is to minimize the size of the DA and achieve a reasonable fill of the feeder facilities. (Id.)

Thus, SBC-CA claims that JA's larger distribution area assumptions contradict SBC-CA's established practice to minimize the size of distribution areas.

Similar to the approach used by JA, SBC-CA sizes SAI equipment based on an estimation of the distribution and feeder terminations required per loop. In response, JA contend that SBC-CA has made an error in its calculations and double counted the spare terminations required at each SAI by applying a distribution fill factor for two-thirds of the SAI terminations. (JA/Donovan-Pitkin-Turner, 2/7/03, para 289.) SBC-CA's witness admits that it made this error in developing its SAI fill factor, but he does not supply a corrected fill calculation. Rather, SBC-CA maintains that its fill levels are more appropriate than JA's. (SBC-CA/Smallwood, 3/12/03, p. 47.)

Discussion. We will use JA's 67.8% fill factor and find that it is reasonable and forward looking because it is based on 3.5 terminations per line at the SAI, essentially identical to the number of terminations that SBC-CA has proposed. We find JA's proposed SAI fill factor more reasonable than SBC-CA's because SBC-CA admits that it made an error in further applying a distribution fill factor to the terminations it determined were needed, without offering a correction. We will use the fill factor proposed by JA because it has been adequately supported and it allows an adequate number of terminations per line, while still allowing 33% spare capacity for maintenance, churn and growth.

7. Premise Termination Equipment

Premise termination equipment refers to the equipment that terminates a local loop at each customer location and includes the drop-wire from the distribution network to the "network interface device" (NID) on the customer premise. JA have proposed fill factors for business and residential customers' premise termination equipment of 57.5% and 54.2%, respectively. SBC-CA has proposed premise termination fill factors for business and residential customers of 45.6% and 17.5%, respectively. These fill factors are highly dependent on assumptions regarding the number of lines that will be terminated at each business and customer location.

In HM 5.3, JA assumed a two-pair termination for each residence and a 6-pair termination for each business. (JA/Mercer 10/18/02, Exhibit RAM-5, p. 16.) In contrast, the SBC-CA's LoopCAT modeled a 6-pair termination for each residence as a single dwelling unit, and business premise termination equipment was modeled in a range of sizes depending on information on lines per business customer. (SBC-CA/Smallwood 3/12/03, p. 37.) To understand these fill percentages, we must take a look at the assumptions that underlie these fill factors.

For residential terminations, SBC-CA claims that JA's approach undersizes premise termination equipment by ignoring the network design standard of building more than 2 pairs to each customer location. (SBC-CA/Murphy 2/7/03, p. 52.) In contrast, JA charge that SBC-CA is inflating loop costs by installing equipment to serve 6 lines at each location, when only 2 or less are needed. (JA/Donovan-Pitkin-Turner, 2/7/03, para. 199.) Likewise, JA contend SBC-CA's approach is flawed because it has not taken into account that many residential premises are in multiple dwelling units. (Id., para. 226.)

For business terminations, SBC-CA used its billing system database for information on lines per customer to propose its premise termination fill factors. (SBC-CA/Smallwood 3/12/03, p. 36.) JA contend SBC-CA's method was flawed for business customers because it focused on lines per customer rather than lines per location, thus ignoring the fact that multiple business could be located at the same location. (JA/Donovan-Pitkin-Turner, 2/7/03, para. 230-32.)

ORA/TURN analyzed current information on SBC-CA's actual premise terminations and state that this data is not consistent with SBC-CA's assumptions in its modeling proposals. Thus, ORA/TURN conclude that SBC-CA's premise termination fill factors are not supported by actual SBC-CA drop and NID purchases and thus are not reflective of current SBC-CA practices or efficient forward-looking ones. (ORA/TURN/Roycroft 2/7/03, p. 31.)

Discussion. We are troubled by the assumptions made in both HM 5.3 and the SBC-CA models in this area, particularly for residential terminations. SBC-CA charges that HM 5.3 undersizes premise termination equipment by modeling only two pairs per residence, which leaves no room for spare if a residence orders a third line. JA charge that SBC-CA seriously inflates loop costs by installing more costly equipment to serve six lines at every residence when current data suggests less than two lines per residence are needed on a forward-looking basis. Both sides charge that neither model determines the appropriate NID size by basing it on the reality of multiple dwelling units.

We find that all of these criticisms are valid. JA have assumed the minimum NID required for each residence and underestimated costs in this area. Further, JA criticize the SBC-CA models for not taking into account the economies in premise termination equipment from sizing them for MDUs, but HM 5.3 does not do this either, other than to economize on drops per location. On the other hand, SBC-CA has assumed a high cost NID with too much excess capacity, which results in a low achieved fill for premise termination equipment, and it has ignored the economies to serve MDUs. For business terminations, SBC-CA has used current information to size NIDs to serve businesses, while HM 5.3 has simply assumed that all business can be served by a six-pair NID. While JA are correct that SBC-CA has not accounted for multiple businesses at one location, HM 5.3 has not done this either.

While we would like to modify the models to account for MDUs and better assumptions regarding premise termination equipment, the parties have only provided potential solutions on this issue for the SBC-CA models. There is no reason to refine the SBC-CA model with regard to MDUs when the HM 5.3 model did not model the network to this level of detail either. Our initial intent was to take the midpoint of the results of the two models and, therefore, we preferred that both models run with similar inputs and assumptions. For this reason, we do not adopt any of the suggested MDU solutions proposed by the parties.

Rather, we modified both HM 5.3 and the SBC-CA models to run with similar assumptions regarding the NID and premise termination fill factors for residential customers. We ran both models with the assumption of a two pair NID and the premise termination fill factors proposed by JA. On the other hand, we lengthened our assumed business and residential NID install time to one hour. This is a generous time estimate for NID installation, and indeed JA contend NID installation can be done in half that time. Nevertheless, we used one hour as a conservative assumption, recognizing that real world constraints, including but not limited to travel and set up time for individual NID installations rather than mass statewide installation, could make a twenty-five minute installation assumption too optimistic. We did not modify NID sizes in either model for business premise terminations, since both models currently run with similar, although not identical, assumptions.

8. Correlation of Fill Factors and Maintenance Costs

SBC-CA argues that it makes intuitive sense that when the network is run with less spare capacity, i.e.,. at a higher fill level, there is a corresponding increase in repair and maintenance activity. (SBC-CA 3/12/03, p. 45.) According to SBC-CA, higher fill levels lead to complicated conditions that could delay service or create outages. (SBC-CA/Bash 10/18/03, p. 34, and Attachment CMB-13.) SBC-CA's witness Bash claims that the preferred range of fill factors is between 30% and 50%, and that above a 50% level, costs rise due to the need to rearrange plant for installation or repair purposes. (Id., p. 35 and Attachment CMB-6.) The SBC-CA cost model allows the user to assume a linear relationship between maintenance cost factors and fill factors so that as a higher fill level is assumed, the maintenance and other expense factors will increase automatically. (SBC-CA/Makarewicz, 3/12/03, p. 27.)

JA criticize SBC-CA's assumption of a linear relationship between maintenance and other expenses and utilization levels. (JA/Donovan-Pitkin-Turner, 2/7/03, pp. 208-09, and 217.) SBC-CA assumes this linear relationship for all fill factors even though Bash's analysis only pertains to copper distribution. According to JA, SBC-CA provides no other analysis to justify linking higher utilization for all facilities to higher expense levels. (JA/Brand-Menko, 2/7/03, p. 106.) JA contend that Bash's analysis is flawed because a proper TELRIC analysis should consider the least total cost of a network element and not optimize one cost, such as maintenance expenses, without considering the full range of economic tradeoffs. (JA/Donovan-Pitkin-Turner, 2/7/03, p. 208.) Further, JA question Bash's underlying data and whether it truly supports her conclusions. (Id., pp. 212-213.)

ORA/TURN criticize Bash's assumption that higher fills lead to higher operating expenses, noting that other factors beyond fill could be affecting operating costs. (ORA/TURN/Roycroft 2/7/03, p. 39.) ORA/TURN maintain that Bash's data relates only to distribution maintenance expenses although SBC-CA tries to correlate all maintenance expenses to higher fill levels. (Id., p. 40-1.) TURN's witness Roycroft criticizes Bash's analysis for focusing on expenses at such a granular level rather than at the wire center level. Roycroft's own correlation test using SBC-CA loop maintenance expenses for feeder and distribution showed a slight negative relationship between fill level and maintenance expense (i.e., higher fill leads to lower expenses), although his results were not statistically significant. (Id., p. 42.) In response, SBC-CA says that Roycroft's analysis is an "apples to oranges" comparison to Bash's work because he focused on trouble-related maintenance expenses while she focused on provisioning costs. (SBC-CA/Bash 3/12/03, p. 40.)

XO joins JA and ORA/TURN in criticizing SBC-CA's assumed link of higher fill levels to higher maintenance expenses. XO claims that the data submitted by SBC-CA do not show a linear relationship of costs and fill levels, and cost increases only appear when fill rises over 50% on average. (XO 2/7/03, p. 30.) Moreover, XO notes that SBC-CA's most recent Loop Deployment Guidelines, which urge loop engineers to maximize plant utilization, contradict Bash's concerns that higher fills raise maintenance expenses. (Id., pp. 28-29.)

Discussion. We will not employ the feature in the SBC-CA models that links fill factors and maintenance costs for several reasons.⁸³ First, we agree with JA that Bash has only analyzed the effect of fill levels on one aspect of loop costs, rather than total loop costs. An efficient carrier would not necessarily optimize only this one cost, without doing a comprehensive analysis of the effect of higher fill on all costs.

Second, JA and ORA/TURN raise concerns with the adequacy of Bash's analysis. We agree with ORA/TURN that there could be other factors driving operating expenses higher, and Bash's analysis does not prove a causal relationship between fill and operating expense. Further, Bash's own analysis shows the rise in costs is steeper when fill levels rise above a 50% level. If we use a fill level of 51.6% in our model runs, SBC-CA's analysis would not indicate vast increases in maintenance expenses at our chosen fill level. Therefore, we see no need to link fill factors to maintenance expenses because even if we relied on SBC-CA's analysis by Bash (which we are not inclined to do), it only applies at generally higher fill levels.

Third, ORA/TURN and JA are correct in pointing out that Bash's analysis is only applicable to copper cable fill, and not other fill rates. Bash only shows this correlation for distribution cable, and not for the other modeling areas where SBC-CA has extrapolated that higher fills lead to higher expenses. There is no basis to assume a link between higher fill for feeder, DLC or switching to higher maintenance costs in all of these areas.

Finally, it is unclear what is proven by Bash's table listing service order dates and fill levels. Bash contends her exhibit illustrates that service delays correspond to higher fill levels, and that 14 of 48 delayed orders corresponded to terminal, or SAI, fill levels of 92%. (SBC-CA/Bash 3/12/03, p. 38, describing Attachment CMB-13 of the Bash Declaration, 10/18/02.) We hesitate to rely on any conclusions from a sample of 48 loops in a network of 17 million. Moreover, since we are modeling terminal fill far below 92%, we are not convinced there is any cause for concern.

J. Structure Sharing

"Structure sharing" refers to the modeling assumption that poles and conduit modeled in a forward-looking network may be shared with other utilities. It also refers to the assumption that even within one company's network, feeder, distribution, and interoffice facilities may share the same poles and conduit. In the cost models, a lower structure sharing percentage indicates that more structure costs are shared with other utilities.

For its models, SBC-CA assumes that forward-looking structure sharing will match the levels that are reflected in the cost factors it has calculated as modeling inputs. In contrast, JA contend that state regulatory commissions and the general public may require more structure sharing among utilities in the future (i.e. lower modeling percentages), to reduce costs and prevent disruptions from excavation and other construction. Thus, JA contend that on a forward-looking basis, SBC-CA's engineers will implement more structure sharing. JA criticize SBC-CA's structure sharing assumptions as merely invoking its embedded network.

JA's witness Donovan explains that HM 5.3 varies the percentage of underground structure sharing depending on the density zone and whether the structure is for feeder or distribution facilities. Sharing percentages for feeder range from 50% in low-density areas to 33% in high-density zones (i.e., more costs are shared with other utilities in higher density zones). Sharing percentages for distribution range from 100% in the lowest density zones to 33% in the highest density zones. (JA/Donovan 10/18/02, p. 17-19.) For aerial structure, sharing percentages range from 50% in low-density zones to 25% in the higher density zones. Donovan hypothesizes that as population densities increase, so do the opportunities for sharing of pole space. (Id., p. 19.)

Donovan explains that HM 5.3 also reflects sharing of structure between feeder and distribution cable by assuming a default value of 55% for sharing of feeder and distribution facilities. Donovan claims this percentage is supported by cost models in other states, the FCC's Synthesis Model, and SBC-CA's own loop deployment guidelines. (Id., pp. 19-21.) According to Donovan, it would be illogical for a carrier to place poles on the north side of a street to handle distribution cable, and the south side of a street to handle feeder cable. In his view, feeder cable can be economically routed onto distribution structure. (JA/Donovan, 3/12/03, pp. 96-97.) For interoffice facilities, HM 5.3 assumes that interoffice cable shares structure along existing feeder routes 75% of the time. (Id., pp. 97-98.)

SBC-CA criticizes JA's structure sharing assumptions because they ignore SBC-CA's actual experience and rely on speculation by JA's witnesses. JA erroneously assume that all networks, including those of utility and cable providers, are rebuilt simultaneously, so that each provider would be ready and willing to share structure costs with the hypothetical new entrant in a TELRIC model. Specifically, JA assume that other service providers will finance up to 75% of pole costs, up to two-thirds of SBC-CA underground construction costs, and 75% of the cost to bury cable. (SBC-CA, 2/7/03, p. 58.) SBC-CA contends that the FCC has rejected JA's assumptions regarding structure sharing between utilities. (SBC-CA/Murphy 2/7/03, p. 46.)

Discussion. With regard to structure sharing between utilities, we will not adopt either the proposal of JA or the sharing percentages embedded in SBC-CA's cost factors. We do not find it reasonable to assume that on a forward-looking basis, a carrier building an efficient network would be able to achieve the structure sharing percentages with other utilities that are assumed by the Joint Applicants. Donovan provides little basis other than his own speculation for these sharing percentages. Neither do we find the sharing percentages proposed by SBC-CA to be reasonable, mainly because we cannot identify what these percentages actually are since they are embedded within SBC-CA's various cost factors for conduit, poles, and other structure. If we could have identified SBC-CA's current structure sharing practices, we might have considered using them in our modeling.

Given our dissatisfaction with the structure sharing modeling inputs of both SBC-CA and JA, we will instead use the structure sharing percentages adopted by the FCC in its Synthesis Model, as set forth in its Inputs Order. (Inputs Order, para. 243.) We will use these percentages as inputs in our HM 5.3 model run. When adopting these percentages, the FCC noted that SBC-CA concurred with these percentages and claimed they reflected its current practice. (Id., para. 244.) As we have already noted in our discussion of SBC-CA's modeling flaws, we are unable to make structure sharing modifications in the SBC-CA models.

With regard to intra-network structure sharing, we find that JA's assumption of a 55% sharing percentage between feeder and distribution networks is realistic on a forward-looking basis, and within the range of percentages adopted in other states and by the FCC. It is reasonable to assume that an ILEC would make efforts to economize by sharing networks that it controls. We will adopt this assumption for our runs of HM 5.3.

K. Plant Mix Assumptions

"Plant mix" assumptions refer to the percentages of aerial, buried, and underground plant assumed in the loop network. SBC-CA assumes that a forward-looking plant mix matches its experience in its current network. In contrast, JA use ARMIS data to develop plant mix assumptions for HM 5.3.

JA criticize SBC-CA for relying on embedded rather than forward-looking plant mix data.

SBC-CA contends that HM 5.3 assumes a plant mix that could never be achieved in California because it assumes away the constraints faced by providers operating in the real world. (SBC-CA 2/7/03, pp. 60-61.) According to SBC-CA, JA rely on statewide ARMIS data, but allocate it across density zones based on the opinion of JA's witness Donovan. Further, JA rely on averages of data dating back eleven years rather than more recent data. This results in JA understating the amount of underground facilities that could reasonably be expected on a forward-looking basis given new local ordinances that mandate "out-of-sight" placement of new telecommunications outside plant construction. JA's assumptions are counter to recent trends toward greater use of underground facilities throughout California. (SBC-CA/Murphy 2/7/03, p. 50.) Further, the FCC has assumed higher percentages of underground structure in higher density zones than JA have assumed in HM 5.3. (Id., p. 51.)

Discussion. We find it more reasonable and forward looking to use SBC-CA's plant mix assumptions in our model runs rather than the assumptions developed by JA, primarily because JA arrive at their plant mix assumptions based on information dating back eleven years rather than recent trends.

In comments on the Proposed Decision, SBC-CA contends we erred in translating its plant mix assumptions into HM 5.3. (SBC-CA, 6/1/04, p. 9-10; Workshop Transcript, 6/14/04, p. 990-1002.) Upon review, we agree with SBC-CA that plant mix assumptions should be based on an approximation of the quantity of loops that are aerial, buried or underground. This means that we will base plant mix inputs on pair feet, as SBC-CA suggests, rather than sheath feet, as suggested by JA. We have amended our calculations and our final HM 5.3 runs incorporate this change.

L. Labor Cost Assumptions

A critical input in TELRIC modeling exercises involves the forward-looking cost of labor to install, operate and maintain the network. Labor costs are manifested in the TELRIC models through not only hourly wage rates, but also assumptions regarding crew size and the time it takes to perform a given task. We now address the key criticisms of the labor cost assumptions in both the HM 5.3 and SBC-CA models.

1. HM 5.3

CWA and SBC-CA criticize labor costs in HM 5.3 as substantially understated. According to CWA, HM 5.3 does not use actual labor cost data provided by SBC-CA, but instead develops its own labor cost inputs without empirical data and based solely on the opinion of a group of industry experts. The opinion of the experts is not supported by adequate documentation. (CWA 2/7/03, p. 6.) CWA states that HM 5.3 is therefore inaccurate with regard to inputs for 1) the number of persons required to operate and maintain certain network elements, 2) the number of hours it takes each person to complete those tasks, and 3) the hourly labor rates paid by SBC-CA to each person who performs these tasks. (Id., p. 5) For example, CWA contends that JA have used a flat hourly labor rate that differs substantially from the rate actually paid by SBC-CA in California. (Id., p.8.) In addition, JA have assumed a two person crew for the installation of buried or underground copper cable, when three persons are actually required to ensure worker safety and promote maximum efficiency. (Id., p. 9.)

SBC-CA criticizes HM 5.3 for its labor productivity assumptions that fail to use California-specific labor factors. (SBC-CA, 2/7/03, p. 62.) For example, SBC-CA contends that HM 5.3 includes unreasonably low assumptions for the installation of an RT cabinet, assuming two technicians can perform the work that normally requires a three-person crew, and unrealistic productivity rates for placing and splicing loop cable. SBC-CA's witness McNeill states that HM 5.3's productivity assumptions could only be achieved in rare circumstances, and not on a consistent basis. (SBC-CA/McNeill 2/7/03, pp. 46-47.) For example, McNeill alleges that HM 5.3 does not include adequate set-up and break down time, does not account for other operating realities and constraints, and fails to incorporate additional labor beyond the outside plant engineer. (Id., pp. 48-49.)

Joint Applicants respond that the labor cost estimates in HM 5.3 are conservatively high. JA witness Donovan defends the assumptions he uses in HM 5.3 regarding the size of labor crews and notes that SBC-CA cannot point to any specific reasoning or government safety mandates for larger crew sizes. (JA/Donovan 3/12/03, pp. 24-26.) Donovan contends that SBC-CA's own witness validates the HM 5.3 assumptions regarding cable placement per day. (Id., pp. 27-28.) Further, Donovan defends his assumptions regarding splicing times as more detailed than SBC-CA's assumptions. (Id., p. 31.)

2. SBC-CA Models

JA criticize SBC-CA for using a labor rate that includes numerous "loadings" that JA allege do not comply with forward-looking cost principles, but instead rely on embedded costs. (JA/Flappan, 2/7/03, p. 4.) These loadings increase the average hourly wage to account for non-productive time, benefits, support assets (such as computers, furniture, and tools consumed in the course of providing labor services), clerical support, and supervisory support. (Id., p. 8.) If SBC-CA's model is used, JA's witness Flappan recommends numerous adjustments to these labor "loadings," including adjustments based on labor information from the U.S. Dept. of Labor Bureau of Labor Statistics (BLS) and adjustments to remove assumed wage increases. (Id., pp. 10-11.) Flappan states that when SBC-CA actual hourly wages are normalized by removing excessive loadings added by SBC-CA, the resulting hourly wage is below the inputs used in the HM5.3 model. Thus, he reasons that HM 5.3 labor costs are conservatively high. (JA/Flappan 3/12/03, p. 3.)

SBC-CA and CWA dispute JA's proposed adjustments to SBC-CA's labor cost loadings. First, SBC-CA and CWA contend that the BLS statistics used by JA are not derived from companies comparable to SBC-CA. (SBC-CA/Makarewicz, 3/12/03, pp. 35-36, CWA 3/12/03, p. 5.) Second, SBC-CA contends it is inconsistent to argue for a lower labor rate while modeling a hypothetically efficient, supra-modern network. (SBC-CA/Makarewicz, 3/12/03, p. 39.) According to CWA, JA unreasonably assume cutting edge technologies in the modeled network, while assuming that SBC-CA can "employ garden variety electricians to operate and maintain its networks and pay them accordingly." (CWA 3/12/03, p. 11.) Third, SBC-CA and CWA contend SBC-CA faces increasing wage, healthcare and pension-related costs, so it is unrealistic to reduce estimates in these areas. (SBC-CA/Makarewicz, 3/12/03, p. 41.)

Similarly, CWA supports the use of SBC-CA labor data because it is derived from the company's collective bargaining agreement with CWA and thus represents the most accurate data available on labor costs. (CWA, 3/12/03, p. 2.) In contrast, CWA contends that the labor-loading adjustments advocated by JA ignore SBC-CA's actual labor cost data and AT&T's own collective bargaining agreements, which commit to specific wage and benefit levels, and increases. (Id., p. 7.)

3. Discussion.

With regard to the labor assumptions in HM 5.3, we agree with SBC-CA that many of the assumptions are based purely on the opinion of JA's witness Donovan and Flappan. SBC-CA attacks HM 5.3 inputs for wage rates, crew sizes, and assumptions on the length of time to perform certain tasks. We find that it is more reasonable to use SBC-CA's fully-loaded hourly wage rate, as modeled by SBC-CA, rather than Donovan's opinion, wherever possible. In certain circumstances, we agree with SBC-CA that the crew sizes in HM 5.3 may be understated. We prefer to adopt SBC-CA's more conservative assumptions regarding the crew size for cable installation activities based on the testimony of SBC's witness McNeil. (SBC-CA/McNeil, 2/7/03, p. 42-45.) We will remove the increases we made to crew sizes for splicing or NID installation, based on comments on the Proposed Decision that these are generally one-person activities. In addition, crew sizes for DLC installation are handled separately in our discussion of DLC installation costs.

On the other hand, we do not agree with SBC-CA's criticisms of Donovan's estimates for the amount of time for certain installation activities, particularly cable installation per day and splicing times, as SBC-CA offers no clear information on the amount of time for these installation activities. We find that Donovan has credibly defended his estimates for cable installation per day. (JA/Donovan 3/12/03, pp. 27-28.) Donovan also provides adequate support for his splicing assumptions by showing that they result in higher cost estimates that SBC-CA's own internal costing tool. (Id., p. 32.) Because we are increasing the crew size for cable installation, we consider it even more reasonable to leave Donovan's estimates for the amount of cable installed per day at the levels assumed by JA, as the model now assumes more hands to do the work.

As a result, we find it necessary to run HM 5.3 with a different fully-loaded hourly wage assumption and with larger crew sizes for cable installation activities. As we have already discussed, HM 5.3 did not allow us the ability to make hourly wage modifications in all areas. We have made the modifications where possible, and otherwise, we note that it was not possible to adequately modify the labor cost assumptions built in to HM 5.3 in all circumstances. In comments on the Proposed Decision, both MCI/WorldCom and SBC-CA suggest new approaches to modify the labor rate embedded in HM 5.3 inputs.  (MCI/WorldCom, 6/1/04, p. 8; SBC-CA, 6/1/04, p. 4.)  MCI/WorldCom suggests the Commission assume, based on HM 5.3 inputs, that labor costs are approximately 50% of terminal and splice investments.  Using this information, the Commission can increase that portion of terminal and splice investment to match SBC-CA's hourly loaded labor rate.  We will adjust HM 5.3 inputs for terminal, splice and SAI investments based on information from the record indicating that approximately 50% of these investments relate to labor costs and can be increased to match SBC-CA's hourly loaded labor rate.  We will not modify labor rates for trenching and buried drop installation because HM 5.3 inputs are based on price quotes from contractors and do not assume installation by SBC-CA personnel. (JA/Mercer, 10/18/02, Exh. RAM-5, p. 22.) 

With regard to the SBC-CA models, we do not agree with the labor loading adjustments suggested by Flappan because they are based on nationwide, historic information for companies that we are not convinced are reasonably similar to SBC-CA. Therefore, we will not make any adjustments to SBC-CA's labor loadings or other labor input assumptions.

In other comments on the Proposed Decision, MCI/WorldCom urges the Commission to remove SBC-CA's loadings for overtime and support assets. (MCI/WorldCom, 6/1/04, p. 27-28.) We decline to adopt these labor-loading changes. With regard to support assets, JA have not shown that a limitation on support assets for nonrecurring costs adopted in D.98-12-079 is applicable to recurring costs loadings here. With regard to overtime loadings, we agree with MCI/WorldCom that SBC-CA's overtime assumptions appear excessive when compared with BLS nationwide overtime data. Nevertheless, MCI/WorldCom admits that a reduction in SBC-CA's overtime assumptions implies the company should hire more full-time employees. (MCI/WorldCom, 6/1/04, p. 28.) It is not appropriate to reduce the overtime loadings SBC-CA has assumed without a corresponding change in the model to assume more employees. For this reason, we will not change SBC-CA's overtime assumptions.

M. Crossover Point

The crossover point refers to the feeder route length at which fiber feeder facilities become less costly than copper feeder. SBC-CA models a crossover point from copper to fiber at 12,000 feet. In other words, LoopCAT assumes that copper feeder loop segments longer than 12,000 feet convert, or "crossover," to fiber after 12,000 feet. According to SBC-CA, copper loops in excess of 12,000 feet are not consistently capable of supporting many services such as DSL, and longer loops introduce inefficiencies into SBC-CA's provisioning processes. (SBC-CA/McNeil, 2/7/03, p. 9.) McNeil further claims that an 18,000-foot loop, as modeled in HM 5.3, cannot provision all the UNEs at issue in this proceeding and would present compatibility problems by not adhering to industry equipment standards. (Id., p. 9.) Nevertheless, as JA point out, SBC-CA's LoopCAT model generates approximately 100,000 copper loops that are longer than 18,000 feet. (JA, 2/7/03, pp. 74-75.)

JA's HM 5.3 model employs the concept of an economic crossover point. The model varies the transition from copper to fiber depending on its selection of the economic point for that distribution area. The maximum copper loop length modeled in HM 5.3 is 18,000 feet. (JA 3/12/03, p. 34.)

JA criticize SBC-CA's use of a 12,000 foot maximum copper loop length because they contend SBC-CA is modeling a more expensive "special services" loop that has higher technical standards than are required for the loop that has been designated as the UNE in this proceeding, thereby increasing loop prices. JA contend that an 18,000 foot maximum copper loop is capable of supporting both POTS voice service and advanced services. Donovan asserts that the FCC concluded that loops as long as 18,000 feet can support those advanced services that are "eligible for universal service support." (JA/Donovan 3/12/03, p. 36.) He also maintains that other technical problems of 18,000-foot loops alleged by SBC-CA are not applicable for the UNEs at issue in this proceeding. (Id., p. 37.)

Discussion. The parties dispute two concepts, namely the crossover point and the maximum copper loop length. With regard to the crossover point, the Commission adopted 12,000 feet as the economic crossover point in the prior OANAD proceeding. (D.96-08-021, mimeo. at 61.) We find no reason to deviate from this modeling approach and we will employ 12,000 feet as the crossover point in our model runs of both HM 5.3 and LoopCAT.

With regard to the maximum copper loop length, SBC-CA contends that copper segments over 12,000 feet pose a problem, yet its own model does not limit copper segments to 12,000 feet. Upon review, we find that LoopCAT contains numerous copper loops longer than 12,000 feet, including approximately 100,000 copper loops longer than 18,000 feet. Nevertheless, the rigidities of LoopCAT should not drive our modeling of loops in HM 5.3. We therefore limit HM 5.3 to a maximum copper length of 12,000, which is our economic cross over point.

N. Switching Inputs

1. Switch Vendors

One of the key differences between HM 5.3 and SBC-CA's switching module, SICAT, is the input assumption regarding the vendors that will provide switches in a forward-looking environment. SBC-CA bases its price per line for switching investments on its contracts with two switch vendors, Lucent and Nortel. In contrast, HM 5.3 assumes that the price per line is based on current prices offered by Siemens, a switch vendor that SBC-CA does not currently use in California, although SBC-CA does purchase switches from Siemens in other states. (Hearing Tr., 4/16/03, p. 680.) JA contend that the Siemens switch prices are the best indication of forward-looking switch prices.

SBC-CA contends that HM 5.3 inappropriately assumes that Siemens switches will be deployed on a going-forward basis, despite the fact that SBC-CA does not deploy a single Siemens switch in its California network today. (SBC-CA 8/1/03, p. 18.) According to SBC-CA, Siemens switches will not interface or operate properly with existing SBC-CA systems because, unlike the Lucent and Nortel switches, they do not have an OC-3 optical interface that allows digital switches to interface with the interoffice network and they do not handle tandem switch functions. (SBC-CA 2/7/03, p. 67.) Thus, because a Siemens switch with SONET-based optical interface capabilities is not currently available in North America, the Siemens switch prices cannot be relied on for TELRIC modeling. (SBC-CA 8/1/03, p. 14.) Furthermore, SBC-CA alleges that deploying Siemens switches will impose other costs on the company, such as training and network modification, which are not accounted for in HM 5.3. (Id., p. 15.)

In response, JA explain that they use the Siemens switch price as a surrogate for forward looking switch prices. In other words, they believe that forward-looking switch prices for all vendors will move towards the Siemens price. (JA 8/1/03, p. 18.) JA admit the Siemens switch modeled in HM 5.3 does not include a SONET optical interface, but is only equipped with an electrical interface. (Id., p. 17.) Nevertheless, JA contend that the Siemens switch price provides the best forward-looking estimate of switch prices based on their speculation that even if the switch did contain an optical interface, it would not cost any more than, and might even cost less than, the Siemens switch used as a proxy in HM 5.3. (Id., p. 18.) Moreover, JA contend that HM 5.3 incorporates all necessary multiplexing costs to connect to optical switches. (Id.) Finally, JA maintain that SBC-CA's arguments regarding extra costs to deploy Siemens switches in its embedded network are inappropriate to consider in a TELRIC study. (JA, 3/12/03, p. 19.)

Discussion. We conclude that we should rely on Lucent and Nortel switch prices to determine unbundled switch costs in our model runs. The problem with JA's argument to use the Siemens contract price as a surrogate for the forward-looking switch cost is that it ignores the fact that the Siemens switch available in North America does not have the optical interfaces necessary to operate in the SBC-CA network. JA conjecture that the newest Siemens switches will come with the optical interface at the same price, but they admit that the switch on which they base forward-looking investment costs does not have the same equipment as the switches that exist in SBC-CA's network today.

Because we cannot rely on JA's argument that the Siemens switch provides all the same functionalities and capabilities as the switches currently deployed in SBC-CA's network, we will only use Lucent and Nortel switch vendor contracts in the switching portion of the model. We do not accept JA's claim that the Siemens switch price should serve as a proxy for a forward-looking prices when we cannot be assured that the Siemens switch is fully compatible with our other forward-looking network assumptions. Additionally, it is not reasonable to expect that a carrier would purchase all of its switching requirements from solely one vendor.

2. Percentage of New, Growth and
Replacement Lines

In order to model switching costs in either SBC-CA's SICAT module or HM 5.3, we must determine the appropriate mix of lines purchased over the modeling period. According to SBC-CA's witness Bishop, "[SBC-CA's] switching contracts reflect a negotiated price per line based upon the composite forecast demand for each type of line or switching system that [SBC-CA] expects to install over the term of each contract." (SBC-CA/Bishop, 2/7/03, p. 5.) There are three basic types of lines provided under switching contracts - "new" lines, "growth" lines, and "replacement" lines. A new line refers to switching components for a line that is completely new to the site of installation, whereas a growth line refers to a line added to existing facilities. A replacement line refers to switching components that replace an existing switching system (either analog or digital). (Id.) Both SICAT and HM 5.3 make input assumptions about the percentage of lines that are new, growth or replacement in order to model switching investment costs.

SBC-CA criticizes JA for assuming in HM 5.3 that 92.6% of lines in the forward-looking network will be purchased at today's new line price, which incorporates a substantial discount over the price for growth lines. According to SBC-CA, it is inconceivable that a switch vendor would provide an entire network of switches at the new switch discount with no further purchases for switch growth or upgrades. (SBC-CA/Lundy Decl., 3/12/03, pp. 7-10.) Instead, a real-world carrier increases capacity and incorporates new technology by adding new, growth and upgrade equipment over time and its network should reflect a mix of such technologies. (Id., p. 19, SBC-CA/Mandella 3/12/03, pp. 2-3.) Furthermore, SBC-CA contends that this Commission's prior OANAD decisions, the FCC, and other states have rejected HM 5.3's unrealistic approach of basing switching costs on too high a percentage of new lines. (SBC-CA/Lundy 3/12/03, pp. 15-16, SBC-CA/Mandella 3/12/03, p. 3.) The Commission has already rejected a similar assumption that Pacific could purchase 90% of its digital lines at the new or replacement price, finding this assumption unrealistic. (SBC-CA/Lundy 3/12/03, p. 17, citing D.98-12-106, pp. 103-04.)

JA defend their input line assumptions as TELRIC-compliant based on the assumption that given current growth projections, a carrier would purchase the vast majority of the lines it needs from switch vendors at the new price. (JA 3/12/03, p. 22.) In turn, JA criticize SBC-CA's SICAT model for including costs to upgrade older generation digital switches and not reflecting a forward-looking mix of digital lines. Specifically, JA criticize SBC-CA for including costs related to switch upgrades and "Other Replacement" lines for miscellaneous purchases under its current contracts. (JA/Ankum 2/7/03, p. 34-35, pp. 50-51.) JA claims these costs are not appropriate for a TELRIC analysis because they relate to growing or upgrading the embedded network, rather than accurately estimating the costs that would be incurred in a forward-looking network.

Discussion. We agree with SBC-CA that JA have assumed too high a percentage of lines can be purchased at the new switch discount. In our own prior OANAD proceeding, we rejected a similar assumption. We find it unrealistic to assume that vendors would sell over 90% of the lines needed for a forward-looking network at the same price that they have negotiated with SBC-CA, when SBC-CA has actually purchased far fewer lines at the new line price. We find merit in the testimony of SBC-CA's witness Bishop, who is involved with switch vendor negotiations, when he states that the price per line reflects a composite forecast demand of what the vendors expect SBC-CA to purchase over the contract period. Similarly, JA have assumed too low a percentage of lines purchased at the growth, or upgrade price. We find it more reasonable to include a higher percentage of growth line purchases to reflect that in a forward-looking environment, a carrier would not be able to buy all of its switches at the new line discount price currently applicable to SBC, and would have to upgrade its switches and incur growth line costs.

We find it more reasonable to run both models using an average of the weighting of new and growth lines reflected in SBC-CA's SICAT model (which is based on SBC-CA's actual percentages of lines purchased during the 5 year study period) and the forecast of JA. This results in a percentage of new lines of 70.8% new and 29.2% new.

A related issue involves JA's criticism that SBC-CA has inappropriately included "other replacement costs" and switch upgrade costs in its switching investment calculations. We find that SBC-CA has not provided adequate justification for its "other replacement costs" and we will remove them. On the other hand, we will not remove the upgrade costs that are included in SBC-CA's SICAT. We find that SBC-CA has provided sufficient justification that over the life of its switches, there will be some costs to upgrade a switch. We do not agree with the JA assertion that the current generation of switches will not require upgrades.

3. Vertical Features Costs

Another aspect of the unbundled switch UNE involves costs for vertical features such as caller ID and call waiting. JA propose that any costs for switching features be included in the port price for the switch whereas SBC-CA proposes that vertical switching features should be priced individually.

a. JA Position

JA claim that it is not possible to identify feature hardware or software costs that are specific to the activation of individual features in SBC-CA's switching contracts. (JA 8/1/03, p. 71.) In other words, hardware and software for features are not purchased on a per feature basis. Rather, as discussed further below, feature hardware comes with all new lines and feature software is purchased for a suite of features in bulk through a "buyout fee." (JA/Ankum 2/7/03, p. 140.)

Regarding feature hardware, JA claim that these costs are captured in the overall switch price per line because vendors provide feature hardware for new lines in the standard per-line price for the switch. SBC-CA's witnesses confirm that switch prices for new lines cover feature hardware, but line prices for growth do not include hardware. (SBC-CA/Lundy 3/12/03, p. 50, SBC-CA/Bishop 2/7/03, p. 10.) For growth lines, JA calculated what it describes as a de minimis one cent per month per line cost for the necessary feature hardware. (JA/Pitts 3/12/03, p. 19.) JA later revised this estimate downward to a half cent per line per month based on their assertion that at least one vendor provides hardware for growth lines in the growth line price. (Hearing Tr., 4/17/03, p. 824.)⁸⁴

Regarding feature software, JA include these costs in the expense portion of their model because software costs are based on "buyout" fees that allow SBC-CA the right to use feature software on either an "uncapped" basis for unlimited lines, or "capped" to a certain number of lines. According to JA, at least one switching contract states that "it is expected that the caps will not be exceeded." (JA 8/1/03, p. 77 n. 212; Hearing Tr., 4/17/03, p. 861.) Moreover, SBC-CA's own SICAT model does not attempt to calculate costs if the caps are exceeded. Although SBC-CA did identify specific feature-related software charges in its contracts, JA claim these references are irrelevant because they are expired, not applicable to SBC-CA's California operations, or refer to the generic software for the overall switch operation. (JA 8/1/03, pp. 77-78.)

JA assert that SBC-CA's proposed feature costs are anticompetitive and violate cost causation by imposing costs on a per-feature basis on competitors when SBC-CA does not incur costs to activate features on individual lines. (JA/Ankum 2/7/03, p. 144.) JA note that although SBC-CA has proposed individual feature costs in California, SBC-CA used the same switching contracts to propose including feature costs in the monthly recurring costs for the port in other SBC-CA states such as Texas, Indiana and Wisconsin. (JA 8/1/03, p. 72.)

b. SBC-CA Position

In contrast, SBC-CA claims that it does incur additional costs for specific feature hardware and software installed in its switches and that UNE rates must allow SBC-CA the opportunity to recover these costs. SBC-CA states that its contracts contain separately identifiable costs for the feature hardware and software that are required to offer vertical features to customers. (SBC-CA 8/22/03, p. 49.) SBC-CA develops costs for feature software, feature specific hardware, and queries for caller ID. The monthly cost for these items is based on software investment and cost factors for sales taxes and capital and operating expenses. (SBC-CA/Currie, 10/18/02, p. 22.)

Although SBC-CA admits that some feature hardware costs are included in per-line prices when new lines are purchased, it asserts that costs for hardware for growth lines and the costs for sufficient memory and switch processor capability must also be incorporated in the cost models. (SBC-CA 8/22/03, p. 51.) SBC-CA contends that memory and processor costs are analogous to upgrading a desktop computer from a 386 processor to a Pentium processor as more software is added to the computer. (SBC-CA 8/1/03, p. 34.) With respect to feature software, SBC-CA says its contracts are replete with individual feature prices based on per-line prices, per-switch prices, or buyout arrangements.

c. Discussion.

Although both SBC-CA and JA agree that SBC-CA incurs costs for feature hardware and software, they do not agree on whether features should be priced individually or whether the costs should be incorporated in the monthly port price. The dispute centers around whether demand by a customer for a specific feature on a discrete line causes SBC-CA to incur the hardware and software costs it has identified. JA state, "If SBC-CA incurs feature costs whether or not a customer requests features, these costs are more properly assigned to the switch port because they are not caused by a request for features, rather, they are caused by the provision of switching in general." (JA 8/22/03, p. 67.)

We agree that costs for feature hardware and software that are incurred up front as switches are purchased through per line, per switch, or buyout charges, are more appropriately included in the monthly port price rather than per feature charges. We also find several problems and inconsistencies in SBC-CA's feature cost study that cast doubt on its ability to accurately determine feature costs on a per line basis. Therefore, we will incorporate switch feature hardware and software costs into the monthly port price for our model runs of both SBC-CA's SICAT and HM 5.3.

We find this is appropriate for several reasons. First, there is no dispute that that feature hardware is included as part of the per-line cost for new switches. Further, SBC-CA acknowledges that "the discrete feature hardware, that would be identified in the [switch] contract, are indeed, ... very limited components ...." (Hearing Tr., 4/17/03, p. 849.) Given these two facts, we find it more reasonable to take the limited feature hardware costs that can be separately identified, above and beyond those already incorporated into the flat per line switching price, and roll them into port rates.

Second, JA's witness Ankum contends that SBC-CA's feature cost study double counts feature hardware costs by calculating feature rates based on costs that are included in the flat per line price for new switches. (JA/Ankum 2/7/03, pp. 144-145.) Given the general agreement that the majority of feature hardware is included in the flat per line price and the lack of response by SBC-CA to Ankum's charge, we find Ankum's position plausible. To avoid potential double counting, we will roll the identified feature hardware costs into port rates.

Third, SBC-CA suggests that flat per line switching prices do not include the memory and processor costs attributable to features. SBC-CA contends the new line price includes only a limited amount of feature hardware, but additional equipment to meet future demands must be ordered separately. As feature use grows, memory and processor capacity must grow too and there is a cost for these upgrades as specified in the contract. (SBC-CA 8/22/03, p. 51.) Despite this qualification, SBC-CA's own feature cost study does not incorporate memory and processor costs, and it is not clear how we would do this on a per feature basis. JA make a credible argument, using SBC-CA's own data showing average processor utilization well below 100%, that memory and processor utilization are sufficient as engineered for new lines and there is little risk that more is needed. (JA 8/22/03, pp. 63-64, 68.) Furthermore, we have already included switch upgrade costs in our per line switching investment calculations that we described in our discussion of input assumptions for new and growth lines. If we included upgrade costs for features as well, this could potentially lead to double counting.

Fourth, SBC-CA contends that any modeling must account for software costs related to feature use that may exceed the caps in the switching contracts. According to SBC-CA, it does not purchase all features under buyout arrangements and JA ignore specific contract provisions discussed at hearings identifying a number of feature hardware and software charges. (SBC-CA 8/22/03, p. 49.) Despite this assertion, SBC-CA does not incorporate costs exceeding the software caps in its own feature cost study and the switching contracts indicate that software caps will most likely not be exceeded. Furthermore, SBC-CA's feature study does not account for uncapped feature software, i.e., SBC-CA's ability to activate features on future lines at no extra cost. (JA/Ankum 2/7/03, p. 144.) We find that because feature software costs appear to most often be incurred through buyout arrangements, i.e., SBC-CA buys software in bulk and incurs the software cost whether or not a feature is ordered, and SBC-CA's own feature study does not take this into account in calculating per feature rates, it is more reasonable to roll feature costs into the monthly port price.

Finally, SBC-CA points to specific contract terms as indicative of feature hardware and software costs. (SBC-CA 8/1/03, pp. 35-36.) JA allege that SBC-CA inappropriately includes activities to maintain and upgrade old switches in a TELRIC analysis and that many of the contract references provided by SBC-CA are out of state, zero cost, or refer to contract provisions for far more functionality than simply end-user features and it is not appropriate to assign all of these costs to features. (JA, 8/1/03, p. 76; JA 8/22/03, pp. 69-72.) We do not find SBC-CA's list of contract terms persuasive given the doubts raised by JA. We prefer to take identified feature hardware and software costs, and incorporate them into the port price rather than rely on SBC-CA's feature cost study.

In comments on the Proposed Decision, SBC-CA contends the flat port rate can only include the 31 feature costs that were studied and priced in the OANAD proceeding. (SBC-CA, 6/1/04, p. 28.) We disagree with this limitation because the scoping ruling established that this proceeding would examine "all elements within the unbundled switching UNE, including ports, features, usage and termination"⁸⁵ and the FCC has defined the unbundled switching UNE to include "the basic switching function" and "all other features that the switch is capable of providing...."⁸⁶ Our understanding of the parties' cost studies is that they included all features currently available and were not limited to the features previously reviewed in the mid 1990's during the prior OANAD proceeding.

4. Rate Structure

Yet another issue surrounding switch pricing is whether the Commission should adopt a new, simplified rate structure for unbundled switching that abandons the set-up and minutes-of-use rate elements that were adopted in the prior OANAD cases in favor of a monthly flat price per line. In other words, all switching costs would be incorporated into the switch port price. JA contend that a flat price per port is a more accurate representation of the way that SBC-CA incurs forward-looking switching costs than SBC-CA's current UNE rate structure of usage-sensitive rates. (JA/Pitts, 10/18/02, p. 11.) SBC-CA maintains that rates should remain usage-based because switches are inherently usage sensitive.

d. JA Position

According to JA, TELRIC principles require that rate structures reflect the manner in which costs are incurred. (JA 8/1/03, p. 63.)⁸⁷ JA contend that a flat per port price is more appropriate than a usage-based rate for several reasons. First, JA claim that SBC-CA does not incur any measurable usage-sensitive costs in its current switch vendor contracts, which are used to estimate forward-looking switching costs. JA acknowledge that SBC-CA incurs costs as the number of lines served increases, and growth in the number of lines is accounted for in the HM 5.3 model. However, growth in the number of lines can be distinguished from growth in usage per line, which is often measured in "centi call seconds," or CCS.⁸⁸ JA maintain that increases in usage on a per line basis, or CCS, does not impact SBC-CA's costs. Indeed, "SBC-CA has deliberately negotiated its contracts so that it does not have to worry about usage variations among its switches." (JA 8/1/03, p. 64, citing SBC-CA's witness Bishop at Hearing Tr. 4/16/03, p. 716-718.) The contracts indicate the same price per line up to a proprietary "breakpoint" CCS level that has been negotiated with the switch vendors. (Hearing Tr., 4/16/03, pp. 716-718.) When new or replacement lines are placed, SBC-CA and the switch vendor jointly engineer the switch for each central office based on a 10-year forecast of CCS requirements so that it is unlikely SBC-CA will encounter additional costs for increases in per line usage for at least 10 years. (JA 8/1/03, p. 64-5; citing to Hearing Tr., 4/16/03, pp. 692-695.)

SBC-CA's witness agrees that vendors do not expand the switch processor every time SBC-CA orders additional lines. (JA 8/1/03, p. 65; Hearing Tr., 4/17/03, p. 869.) SBC-CA negotiated to pay the same amount per line for any switch, regardless of actual usage levels, unless usage exceeds a "breakpoint" CCS level, thus the pricing is not usage based. Therefore, JA contend that SBC-CA cannot justify charging CLCs usage-related switching costs, i.e., charges that vary based on the minutes of use per line, "because any conceivable usage-related costs that SBC-CA might incur under those contracts are too trivial to justify a separate charge." (JA 8/1/03, p. 63.)

JA witness Ankum analogizes that SBC-CA's switch purchases are like a bridge that is engineered with 4 lanes for peak hour capacity. (Hearing Tr., 4/17/03, pp. 791-2.) There is a fixed cost to building a bridge to serve the peak hour usage. But there is no incremental cost for one car to cross the bridge in the middle of the night. Similarly, the switch is engineered for the peak hour usage based on a 10 year forecast. SBC-CA pays one price to the switch vendor, whether the switch is engineered for a high usage or low usage level, up to the pricing breakpoint in the contract. Ankum points out that switches engineered for areas with higher usage per line cost the same as switches in areas with lower usage per line. Likewise, switch vendors charge the same price per line whether the line is used for a residential or business customer.

According to JA:

[T]he critical fact is the switching costs incurred by [SBC-CA] under its switching contracts are not usage sensitive. (JA 8/1/03, p. 71.)

...

A switch for a downtown high-density urban area will cost [SBC-CA] the same for a given number of lines as a switch in a rural community. That is, [SBC-CA] incurs the same costs per line for a high volume customer in a high volume, high usage urban area as it does for a low volume customer in a rural community." (JA/Ankum 2/7/03, p. 122.)

Ankum explains that although SBC-CA's SICAT model makes it look like SBC-CA purchases each switch component separately, in fact SBC-CA purchases switching facilities on a per line basis. (Id., p. 102.) All of the switching costs that SBC-CA's model identifies as usage-sensitive are, in fact, included in the fixed per line price. (Id.)

Second, JA maintain that SBC-CA's own "Infrastructure Deployment Guidelines" specify minimum capacity requirements for switches that are far in excess of SBC-CA's current usage levels. (JA 8/22/03, p. 62 citing PHE-11, p. 104 and PHE-12, p. 95.) Using Ankum's bridge analogy, each switch, or bridge, is designed to handle a minimum level of peak hour traffic, even if actual traffic is less. JA claim that according to SBC-CA data, the average level of switch processor utilization in California, including all calling and feature demand, is far below the threshold CCS pricing breakpoint specified in the vendor contracts and below the minimum capacity requirements listed in SBC-CA's infrastructure guidelines. (JA 8/1/03, p. 67, citing JA/Pitts Declaration, 3/12/03, Exh. CEP-REB-10.) JA allege this same data shows that average CCS in California is holding steady or declining slightly, so there is no danger of exceeding switching capacity if switches are provisioned at the minimum capacity specified in SBC-CA's deployment guidelines. (JA 8/22/03, p. 64, and Hearing Tr., 4/17/03, p. 795.)

According to JA, per line switch usage may decline as a result of high-speed connections such as DSL and cable modems that take internet traffic off the circuit switches. (JA 8/1/03, p. 67.) Although SBC-CA's Mandella argues that internet traffic has increased network usage,⁸⁹ JA rebut this with several arguments. First, when customers use second lines for dial-up access, the increase in usage is associated with line growth rather than more usage per line. Further, Ankum says most of the growth in dial-up access has already occurred and CCS levels per line are actually decreasing due to such developments as increased DSL lines and cable modems that take Internet traffic off the circuit switches. Any CCS per line increase due to dial-up internet traffic has already occurred and should be reflected in the data JA collected from SBC-CA on statewide average CCS per line. SBC-CA admits there is no longer an upward trend in CCS per line. (JA/Ankum 2/7/03, pp. 117-119, citing SBC-CA/Mandella deposition of 11/20/02.) SBC-CA does not dispute the current average statewide CCS levels other than to say that they may be depressed because of the economy and that individual switch CCS levels may be higher. (Hearing Tr., 4/16/03, p. 722.) SBC-CA also does not dispute that switches are installed based on 10-year forecasts of CCS levels, and SBC-CA's own data indicate that average processor utilization is far below maximum levels. (JA 8/22/03, p. 64.)

Third, JA assert that SBC-CA's testimony regarding usage considerations in engineering and maintaining older switches is not relevant to forward-looking costs. While SBC-CA's witness Mandella discusses engineering considerations regarding switch usage, he does not show that switching costs are usage-sensitive. Even if switch components are sized to handle specific amounts of traffic, this does not prove that SBC-CA's switch costs are usage-sensitive if all of the components described by Mandella are purchased as part of the per line price (JA/Ankum 2/7/03, pp. 111-13). SBC-CA's witness Lundy confirms that contracts have a per line pricing structure under which the vendors supply all necessary switch components. (JA 8/1/03, p. 69.) According to JA, Mandella's reference to load balancing and managing line concentration ratios refer to maintenance expenses on older, outdated switches which would not be required under a TELRIC model that assumes investment in new, forward-looking switches provisioned based on a 10 year usage forecast. (Id., p. 70.) JA contend these normal maintenance activities are incorporated into any modeling as switch maintenance expenses. (Hearing Tr., 4/17/03, p. 791.)

Finally, JA contend that SBC-CA's proposed usage rate structure discriminates against competitors by forcing them to pay incremental costs per minute of use that SBC-CA itself does not incur. This impairs CLCs' ability to offer flat-rated services at prices that are competitive with SBC-CA's flat rated residential services. (JA 8/1/03, p. 63.) JA note that usage-sensitive rates have been rejected in other SBC-CA states, namely Illinois, Wisconsin, and Indiana. (Id., p. 64, and JA/Ankum 3/12/03, paras. 6-21.)

e. SBC-CA Position

SBC-CA supports its proposal for usage based switching rates with the testimony of witness Mandella, who describes how switches are engineered based on the estimated usage of the switch, typically measured in CCS. "The switch usage design is engineered to meet the expected call volumes for the busiest hour of the busiest period of the year and is based upon both historical trends and forecasted call volumes for the particular switch being examined." (SBC-CA/Mandella 10/18/02, p. 11.) More usage per line means SBC-CA has to increase the usage portions of the switching equipment. (Id., p. 11.) This can be done by modifying the line concentration ratio,⁹⁰ adding more trunks, increasing the links, or "umbilicals," between the host and remote switch, or increasing the memory of the switch processor. (Id., pp. 14-17.)

SBC-CA maintains that evidence shows that switch vendors do not install excess capacity for free. Engineers for the switch vendors and SBC-CA evaluate traffic and demand information, and forecast data, to engineer the appropriate amount of capacity unique to each central office switch, based on CCS requirements. Vendors do not install the maximum amount of CCS capacity in every central office, even if contract prices reflect one price for all capacity up to a specific CCS capacity level. (SBC-CA 8/1/03, p. 24, and SBC-CA 8/22/03, p. 47, n. 155.) SBC-CA cannot install the maximum capacity in every switch, and it must pay for installation of additional capacity in its switches. (SBC-CA 8/1/03, p. 30, and Hearing Tr., 4/16/03, p. 714.) Thus, SBC-CA reasons that usage plays a major role in determining its switch costs.

SBC-CA's witness Aron counters the JA proposal by suggesting that even though vendors recover usage-sensitive costs through a fixed per line price, and even though they may provide enough capacity to handle most situations for this fixed price in the short run, this does not mean that incremental switch usage is free. (SBC-CA/Aron 2/7/03, p. 37.) Aron suggests that usage-sensitive rates are justified under a long-run view, which requires that costs incurred to provide capacity be identified and recovered from customers in proportion to their share of switch capacity consumption. (Id., SBC-CA/Aron 3/12/03, p. 77.) SBC-CA witness Lundy argues that in the long run, the price that switch vendors charge is affected by utilization because long-run usage of the switch affects vendors' production costs for switch components. (SBC-CA/Lundy 3/12/03, p. 36.)

Aron explains that there are great variations in customer usage levels and a flat rate for switching would be inefficient because it would increase usage at the margins. (SBC-CA/Aron 3/12/03, pp. 74-75.) Aron says:

I would expect that flat-rated port prices would result in much higher usage levels on CLEC ports that one would see under usage-sensitive rates, even aside from the responsiveness of individual customers to different price levels. This is inefficient because it leads to the migration of the highest-usage customers to CLECs on the basis of a price structure that fails to reflect the costs that these high usage customers cause to the system, rather than on any basis of relative efficiency or other economic fundamentals. Hence, the fact that there is, in light of the Internet, a substantial variability in usage across customer types in today's economy and the fact that CLECs can target customer types so as to attract those that impose the greatest usage cost burden on the network, make flat-rated pricing inappropriate and inefficient from an economic perspective. (Id., p. 76.)

Aron reasons that CLCs that can control customer type will target high volume customers with a low rate, despite their usage burden on the switch. (Id.) This will be inefficient because the price CLCs will be charged for usage by their high volume customers will not reflect the costs those customers cause in the long run.

SBC-CA alleges that a flat rate for switching is contrary to common industry practice. SBC-CA notes that the FCC has endorsed usage based switching rates in several other states when reviewing Section 271 applications for in-region long distance entry. (SBC-CA 3/12/03, p. 65, n. 254.)

f. PacWest Position

PacWest opposes JA's proposal for flat rates for end-office switching and explains that its own and other CLCs' interconnection agreements specify that SBC-CA's UNE price for switching is used to determine what PacWest charges SBC-CA for termination of its customers' local traffic. JA's flat rate option leaves no mechanism for carriers to compensate each other for usage on each other's network, and leaves CLCs who invest in their own network facilities with no means of cost recovery. (PacWest/Wood Decl., 3/12/03, p. 23.) PacWest says the Commission cannot ignore the effect of a flat-rated price on the CLCs who apply this rate for various forms of inter-carrier compensation, also known as reciprocal compensation. PacWest maintains there are several economic and policy reasons to reject JA's proposal.

First, PacWest claims that a flat per port rate is inconsistent with the definition of economic cost and violates several of the Commission's CCPs. (Id., p. 4.) According to PacWest, JA are departing from cost causation principles when they argue that costs should follow vendor contracts and not engineering principles. PacWest asserts that JA's reasoning behind a flat rate is flawed because switches do not have infinite capacity. (PacWest/Wood, 3/12/03, p. 5.) Also, PacWest contends that the JA proposal is limited to one category of usage, thus ignoring switched access and toll, which continue to be priced on a usage-sensitive basis. This violates CCP 8, which states that "the cost methodology implementation should ensure that costs for services which use the network in the same way are treated consistently." (Id., p. 11.)

Second, PacWest maintains that the JA proposal ignores the historical treatment of 70% of costs as "traffic sensitive" and 30% non-traffic sensitive." (PacWest 2/7/03, pp. 11-12.) PacWest claims that JA have provided no reason to abandon this traditional split. While the JA proposal implies all costs are non-traffic sensitive, the JA have not explained why this 70/30 split was appropriate in the past, but is no longer. (PacWest/Wood 2/7/03, p. 11.)

PacWest urges that if the Commission does adopt a flat per-port rate structure, the flat rate structure should not be applied in the context of "reciprocal compensation" to CLCs for use of a CLC switch. PacWest says that "given the nature of the services for which reciprocal compensation is charged" a flat per port rate is not appropriate. Thus, the Commission should limit use of the flat rate to what SBC-CA charges CLCs, and SBC-CA should simultaneously calculate a usage-sensitive unbundled switching rate for reciprocal compensation in interconnection agreements. (PacWest 8/1/03, pp. 1-2.) In the three states where a flat per port rate has been adopted, a usage sensitive rate has been retained for reciprocal compensation purposes. (Id., p. 7.) PacWest explains how the HM 5.3 model can calculate a usage-sensitive unbundled switching rate even if a flat per port rate is adopted for SBC-CA. (Id., p. 8.)

g. Discussion

This issue highlights that economists' opinions can differ greatly when faced with rate design questions. In this case, experts for both parties reviewed the exact same switch vendor contracts and drew vastly different conclusions. After intense scrutiny of the switching contracts, both sides agree that the contracts set prices for breakpoints of CCS, or usage per line, levels. The record shows that in SBC-CA's switching contracts, the primary basis used by switch vendors to charge SBC-CA for its switches is a flat price per line. JA say this proves that usage per line plays no part in switch costs because all usage levels are provisioned at the same price. SBC-CA says the exact opposite--each switch is provisioned based on its usage and there is a cost to go above that usage level.

We find the issue comes down to whether it is reasonable to assume that the capacities provisioned at the prices in the switching contracts will be exceeded. In other words, is it likely that SBC-CA will have to incur an additional cost to accommodate growth in usage per line beyond the 10 year forecast used to provision a new switch? We conclude that given our TELRIC modeling assumptions in this proceeding, it is unlikely that SBC-CA would have to buy additional capacity because switches are provisioned based on a 10-year forecast of capacity requirements. JA have shown that the current statewide average CCS level is well below the minimum quantity CCS provisioned under the contracts and SBC-CA's current CCS levels are below switch maximums. Even if individual lines have higher usage, normal switch operating expenses should cover "grooming" and other switch maintenance work without SBC-CA incurring an additional cost to increase the CCS level of the switch.

Ultimately, we agree with SBC-CA's witness Aron that long-run usage is not free. But we find that given how SBC-CA incurs costs through its switching contracts, switch pricing over the TELRIC modeling period reflects a series of step increases rather than a smooth upward curve. Essentially, the cost SBC-CA incurs for its switches is flat until SBC-CA jumps to a higher level of usage per line. Therefore, we agree that because SBC-CA's cost to obtain switches is a flat per line rate regardless of the usage level provisioned, at least up to the pricing breakpoint, the cost that is passed on to CLCs who use SBC-CA's switch should also be a flat monthly charge. As usage on individual lines increases or decreases, SBC-CA's costs do not change as long as usage does not rise above the maximum CCS level in the switch contract. Although we agree with SBC-CA's Aron that this may cause CLCs to target high volume customers, SBC-CA already does this itself through its own flat-rated pricing plans. Further, as JA's witness Ankum has pointed out, SBC-CA will incur higher costs for high volume users only if three conditions are met simultaneously (i.e. all increased usage occurs at the peak hour, the usage exceeds the CCS limitations engineered into the line, and usage increases for a large number of lines and not just a few lines). (JA/Ankum, 2/7/03, p. 107.) We agree with JA that for all of these conditions to be met simultaneously would be extraordinary. Thus, the only time SBC-CA's switching costs rise is when the entire switch must jump to the next CCS level, which is unlikely given that switches are engineered based on a 10-year forecast of usage per line.

We find that charging a flat rate for switching is consistent with TELRIC guidance that rate structures should reflect the manner in which costs are incurred, and consistent with our CCP 2 regarding cost causation. SBC-CA incurs switching costs on a predominantly per-line basis, therefore, there is no basis for it to charge its wholesale customers a usage sensitive rate.

We disagree with PacWest that the flat rate proposal violates cost causation for the reasons we have already explained. In addition, we do not agree that a flat rate for local switching violates CCP 8. It is not clear from this record that tandem, toll, and switched access services use the network in the same way as local switching and that the same rate structure must be used for those services. Moreover, we find that JA have provided ample reasons to deviate from the former 70/30 split of traffic sensitive and non-traffic sensitive costs, based on their analysis of the vendor contracts and CCS provisioning practices. As JA note, SBC-CA itself did not use the former 70/30 methodology either in its own switch costing proposal. (JA/Murray, 3/12/03, p. 46.)

Finally, we do not find that PacWest's concerns over reciprocal compensation should prevent us from adopting a flat rate structure for unbundled local switching. We agree with JA and PacWest that we can employ the same method as other states and retain a usage-sensitive rate for reciprocal compensation purposes. We have calculated usage based end-office switching rates using HM 5.3 and SBC-CA's SICAT which are set forth in Appendix C.

As a side note, the market has rapidly evolved to flat rate monthly pricing for local service, and many other telecommunications services as well. Allowing SBC-CA to collect usage costs from its wholesale customers, who are its competitors in the flat-rated residential market, would place CLCs at a disadvantage. The FCC's Wireline Competition Bureau (Bureau) came to a similar conclusion when it examined the issue of usage based versus flat rates for switching in an arbitration case between AT&T, MCI/WorldCom, and Verizon Virginia.⁹¹ In its decision, the Bureau determined that several categories of switching costs should be recovered on a per line port basis, and that charging a per line port price "recovers these costs on a competitively neutral basis, thereby potentially extending to many different subscribers the benefits of competition." (VA Arbitration, para. 464. See also paras. 471, 472, and 477.)

5. Switch Fill Factors

HM 5.3 models switching UNE costs based on a 94% analog and digital line fill factor. SBC-CA proposes an analog fill factor of 82% and a digital fill factor of 20%.

SBC-CA maintains that the fill factors it uses for digital line ports, analog line ports, and digital trunk ports are forward-looking and based upon the best available data representing conservative estimates of average switch utilization. (SBC-CA, 3/12/03, p. 68.) According to SBC-CA, "[T]he fill factors used in SBC-CA California's cost studies, which are based on current, efficient utilization levels, are the most appropriate values to use." (SBC-CA/Lundy 3/12/03, p. 42.) Further, SBC-CA explains that digital line utilization is lower than analog line utilization because digital "is a newly deployed technology, and as such would be expected to have lower utilization." (SBC-CA/Mandella 3/12/03, p. 16.) As SBC-CA's witness Mandella explains, digital equipment placed today "is sized to meet the projected size of the tract and the anticipated rate of construction," and "utilization of the equipment will be minimal in the early stages of actual service activation and increase over time." (Id.)

With regard to HM 5.3, SBC-CA contends that the switch fill factors it uses are based on nothing more than speculation and ignore the need for administrative spare, economic order quantities, and forecast uncertainties. SBC-CA disputes JA's assertion that a 94% fill level is appropriate because switches are highly modular and can be expanded on short notice. SBC-CA alleges that JA fail to account for standard switching equipment order intervals, and that at least one switch vendor has stated that a small switch installation will take 25 weeks from start to finish. (Id., p. 17.)

In contrast, JA argue that SICAT fill factors are based on embedded data, and lead to a digital fill factor that is too low and entirely inconsistent with SBC-CA's assumption of wide deployment of DLC in its loop study. SBC-CA's proposed 20% digital fill factor derives from the fact that SBC-CA has deployed little DLC in its current network during a time when it is implementing a new technology. (JA/Ankum, 2/7/03, p. 99.) Thus, SBC-CA's model reflects actual utilization rates from today's switching network rather than forward-looking usage assumptions. JA contend this violates TELRIC modeling standards. According to JA, SBC-CA's SICAT model has ignored SBC-CA's own engineering guidelines that recommend a far higher digital line "fill at relief" level. (Id., p. 100.)

Discussion. First, we find that SBC-CA has not met its burden in justifying why its current digital fill level is forward-looking. SBC-CA relies on actual fill levels from its network today. In particular, SBC-CA proposes a 20% digital line fill factor as forward-looking, but admits that current digital fill rates are low because the technology is newly deployed. (SBC-CA/Mandella 3/12/03, p. 16.) Moreover, we find SBC-CA's testimony in this area confusing. On the one hand, SBC-CA criticizes JA for assuming that SBC-CA's California network is growing "when, in fact, [SBC-CA's] overall customer base in California is declining, and is expected to decline in the future." (Id., p. 14; and SBC-CA 3/12/03, p. 70, n. 278.) On the other hand, SBC-CA's witness Mandella explains that SICAT is based on a fill factor of 20% for digital lines "consistent with the forward-looking expected utilization of a newly deployed, growing, and efficiently managed network." (Id., p. 15, emphasis added.) We are puzzled why SBC-CA has used such a low digital fill factor, which presumably leaves 80% spare capacity for growth in the network, when it contends that its customer base is declining.

Second, we see no reason why there should be such a drastic disparity between analog and digital fill levels. As JA have explained, switching equipment is highly modular and can be expanded within months rather than years, which means minimal spare is warranted. (JA/Ankum 2/7/03, p. 87-88.) SBC-CA's testimony supports this, given Mandella's estimate of 25 weeks for switching equipment installation.

Third, we find that JA's 94% fill levels may be too optimistic and not leave enough room for administrative spare and growth. Instead, we find it appropriate to use SBC-CA's 82% analog fill factor and apply the same rate to digital line assumptions. We will therefore use this 82% fill level in our model runs.

Cost of equity = Risk free rate + (Market risk premium) x (Beta)

5.8% risk free rate + (9.1% x .83) = 13.35% Avera claims that recent evidence supports this 13.35% estimate. He updates this portion of his analysis based on an update to the Harris & Marston 1992 study. The update shows an average equity risk premium of 7.14% which he adjusts upwards by 3.25% based on interest rate declines since the study period. This results in an equity risk premium of 10.1% and an implied cost of equity of 15.22%. (SBC-CA/Avera, 10/18/02, p. 25.)