IV. The Cost of Providing Basic Universal Service in Pennsylvania

A. Use of Auctions

1. Position of Parties

GTEN was the only party in this proceeding to propose that the cost of BUS be determined through a "market-based approach" or an auction process. GTEN proposes that eligible carriers or COLRs be selected through a competitive bidding process. See GTEN Stmt. 1.2 (Williams) pp. 29, 30. Under GTEN's proposal, the auction process would actually determine the level of BUS support for high cost areas. GTEN Stmt. 1.2 (Williams) p. 29. The bidding process would determine the lowest level of universal service support that any carrier would be willing to accept while undertaking COLR responsibility. GTEN Main Brief at p. 27.

The Commission would first establish an affordable universal service rate. Each qualified bidder would then bid the support amount required to provide universal service to a designated area, wherein the support amount is the difference between the market price for the service and the rate established by the Commission. GTEN Stmt. 1.2 (Williams) pp. 31-33; Tr. 87.

GTEN witness Williams explained the procedure as follows:

Each qualified bidder would bid the support amount it requires, per subscriber, to undertake the COLR obligation defined by the Commission for a time period specified in the request for bids issued by the Commission. The Commission should establish a preset time period for the COLR obligation in order to give prospective COLRs a clear indication of the responsibility for which they will be bidding. This process would let qualified bidders know how long they have to recover any required capital investments, and when the next opportunity to revise the level of US support will occur. ...The Company recommends that the COLR obligation period be set for no less than five years and no more than ten years. The new level of support would be determined by the lowest bid of a qualified firm, which would then be committed to serve as COLR in the given area.
Each of the other qualified bidders would then be authorized to decide whether to become a COLR at the support level determined by the auction process. If another carrier chose to "opt-in", it would assume the same obligation as the provider that submitted the winning bid. Carriers that "opt-in" will become eligible to receive the same per-customer level of support that the winning bidder receives for each customer to which it provides BUS.
A service provider that elects to "opt-out" would not bear any of the responsibilities of a COLR, nor would it be eligible for COLR funds. All carriers who "opt-out" in a given area should be afforded streamlined, symmetric, non-dominant regulation.
At the end of the commitment period specified in the Commission's auction procedure, the COLR obligation for that CBG would again be put up for bid. At that time, the new auction process would be based upon the current definition of US, including any revisions to the definition that the Commission may have adopted since the previous auction process. The US definition on which the support levels were initially bid should remain in effect for a given CBG during the commitment period. Each bidder would then know with certainty what is expected of it during that period. Any changes to the definition of US should be introduced at the next auction cycle for each area. The recommended auction process will establish a market-based "price tag" for any new functions or features that have been added to the definition.

GTEN argues that the auction process is superior to a procedure that calculates universal service funding support based upon cost studies. GTEN Reply Brief at p. 21. GTEN states that there is vigorous disagreement among the parties with respect to the validity of the various cost studies that have been proposed and that they reach drastically different conclusions. Id. at p. 21. GTEN also argues that because the auction process is market based it will result in adoption of the best technology and network structure and, perhaps most importantly, will encourage bidders to focus on the total business opportunity of high cost areas rather than limiting their analysis based upon the cost of providing BUS. Id. GTEN also urges adoption of the auction process because it would eliminate the need to address one of the most hotly contested issues in this proceeding -- the appropriate treatment of the cost of the local loop. Id.

MCI witness Bryant testified that GTEN's proposal may not be workable because it would require a company other than the ILEC which has the winning bid to instantly build wholly new facilities to provide service to customers in all of the area subject to auction. Meantime, the incumbent LEC would be under no duty to continue providing service as the COLR. This would create the potential for disruption of service. MCI Stmt. 4.0 (Bryant) at p. 4. Similarly, Bell witness Emmerson explained that in recognition of the fact that a LEC may lose its bid, two practical problems arise:

1) the LEC will no longer have a good use for much of its loop and other infrastructure investments, but also may not want to make them available to the new competitor at fire sale prices because the new COLR is also a competitor to the LEC; and
2) the new COLR must somehow instantaneously be ready to provide service to all under its newly acquired obligation.
Bell Stmt. 3.2 (Emmerson) at p. 24.

Bell witness Emmerson also testified that the incentives created by the auction may further distort investment decisions and accounting practices in ways which reduce economic efficiency. Emerson pointed out that "[a]t auction time there is always a cost of exit for the incumbent and therefore its network planning is certainly going to be affected by the timing of the next auction because there is a risk of losing the bid." Bell Stmt. 3.2 (Emmerson) at p. 23.

If the Commission determines not to adopt the auction process, then GTEN recommends using the BCM with modifications to determine the market price for providing universal service.

b. Discussion

While GTEN's proposal is facially attractive because it would minimize the size of the fund, we cannot endorse the use of auctions at this time for obvious reasons. First, we agree with the arguments of OCA and AT&T that the auction process would not be competitively neutral at this time because of the inherent advantages inuring to the ILEC due to its ubiquitous network and status as a ubiquitous provider. Both OCA witness Johnson and AT&T witness Darrah point out that the COLR auction process would constitute a barrier to entry since the incumbent would have an insurmountable advantage by virtue of having facilities already in place. OCA Stmt. 1.2 (Johnson) p. 35; AT&T Stmt. 1.1 (Darrah III) pp. 26-27. OCA witness Johnson testified: "[t]he incumbent carriers enjoy numerous benefits as a result of their historic quasi-monopoly position. No other carrier has enjoyed these benefits, and no other carrier can realistically be expected to take their place as the carrier of last resort, at least for the near term future." OCA Stmt. 1.2 (Johnson) at p. 35. We agree that an existing ILEC which already has in place a ubiquitous network throughout its service territory would have a distinct advantage and could underbid most other competitors at this point in time.

Further, we are convinced that without more study GTEN's proposal for COLR auctions will be complicated to implement and administer and could increase the costs of providing BUS. AT&T Stmt. 1.1 (Darrah III) pp. 26-27; OCA Stmt. 1.2 (Johnson) p. 36; OTS Main Brief at p. 9. For instance, AT&T witness Darrah testified that depending on how the service territories were defined, the Commission could be required to conduct a large number of "auctions." AT&T Stmt. 1.1 (Darrah) at p. 26. GTEN contemplates administering its proposal on the basis of CBGs which on average have only 400 households, which could conceivably mean that this Commission would have to conduct over 11,000 auctions, a task we have no desire of undertaking. Id.

We recognize however, that there are aspects of the auction proposal that warrant further consideration and that the use of auctions may be more feasible once competition takes hold and there is more than one established carrier in any given market. Like the Joint Board, we find recognize that there may be important advantages in using an auction process or market-based approach in particular the potential for reducing overall BUS funding levels.

Nonetheless, because of the likely anti-competitive consequences that have been identified by the parties to this proceeding, we underscore our finding that the use of auctions is inappropriate at the present time.

B. Multi-Fund Approach

1. Position of the Parties

Only one party, Bell, urges that the Commission implement multiple universal service funds, ranging anywhere from two funds to 40 funds representing each ILEC service territory in the state. Bell argues that a state-wide fund would perpetuate and potentially extend the present inefficiencies associated with cross-subsidization. Bell Main Brief at p. 57. Bell also argues that a state-wide fund might not be "sufficiently responsive to variations in telecommunications policy goals and needs among local communities." Bell Main Brief at 57. Bell would limit funding to the providers in a particular serving area so as not to require carriers to support the serving territories of other LECs. Bell Main Brief at p. 57. Finally Bell argues that the definition of universal service will evolve over time and may need to recognize special concerns which could apply to different portions of the state. Bell Main Brief at p. 58.

GTEN and the other parties oppose the establishment of multiple universal service funds in Pennsylvania. GTEN urges that there should be a single statewide fund for all universal service programs. GTEN Main Brief at p. 22. It states that a statewide structure would be competitively and structurally neutral and would be based on the broadest possible base of telecommunications services and providers. GTEN Main Brief at pp. 22-23. It also argues that if the universal service funding mechanism was done on a LEC by LEC serving area, the Bell proposal would require the customers of the LEC serving the high cost area to, in effect, subsidize themselves. GTEN Reply Brief at p. 30.

PTA witness DeFalco agrees and points out that the funding of BUS is a statewide obligation and produces statewide benefits of increased accessibility and increased toll and local usage which cross LEC boundaries. Additionally, with one fund, the burden associated with administering multiple funds will be eliminated and the costs of funding the pool will be spread over a much larger base. PTA Stmt. 1-R (DeFalco) p. 10. DeFalco sums up the argument against the multi-fund approach in the following excerpt:

If the Commission were to fragment the funding administration by incumbent local exchange carrier serving area, the funding costs could be inordinately applied to certain participants in the pooling process which would levy a greater burden of the funding costs eventually back to end-users in high cost areas. A better plan is to spread such costs over the entire base of Pennsylvania and have all funders and eventually all end-users contribute some very small share to keeping local exchange rates reasonable throughout the state. PTA St. 1-R at 9. Having a greater number of people on the network in Pennsylvania, will benefit all Pennsylvanians. Id.
PTA Main Brief at p. 36.

GTEN offered the following reasons why the universal service funding should be statewide and not based upon the serving area of any particular LEC. GTEN states:

Because it will certainly be the case that CLEC serving areas will not precisely track incumbent LEC exchanges, it would simply be impossible to administer a universal service fund mechanism that attempted to provide universal service support funds to a particular serving area based only upon funds collected from that area. Attempting to do this would also be contrary to the underlying theoretical basis for providing universal service support, i.e., the existence of the "network externality."
The network externality is used to support the subsidization of customers who would not otherwise choose to, or be able to, pay for access to the network. The theory is that all those on the network benefit from the fact that more people are on the network.
Because the underlying rationale for subsidizing universal service is the externality, it is consistent to collect the universal service funds on a statewide basis and distribute them similarly. After all, a subscriber in a particular serving area benefits from the fact that more people are on the network throughout the state, and not just in the subscriber's own serving area.

GTEN Main Brief at p. 26.

Finally, many parties believe that multiple funds could run afoul of several provisions of the TA-96. GTEN witness Williams testified that Bell's proposal would unreasonably narrow the support base and would not be competitively neutral since all providers would not be contributing to universal service on an equitable and nondiscriminatory basis. He further expressed concern that the Bell proposal conflicts with §§ 254(d) and 254(f) of the Federal Act. GTEN Stmt. 1.1 (Williams) at p. 33.

2. Discussion

Once again, we agree with the majority of commenters that a multi-fund approach is not in the public interest and is not supported by the evidence in the record of this proceeding. As discussed in more detail below, we find that the evidence in the record establishes that a multi-fund approach would be antithetical to the three primary goals of Chapter 30, competition, infrastructure development and universal service, for the following reasons.

First, we agree with many of the commenters that the multi-fund approach would have obvious anti-competitive impacts. It would unreasonably narrow the support base and all providers would not be contributing on an equitable and nondiscriminatory basis. For instance, carriers competing in Bell's service territory would be contributing a different amount than carriers serving GTEN's territory. If fund eligibility is on an exchange basis, conceivably the amount assessed could be different for each exchange. By contrast, a statewide fund ensures the broadest possible funding base available which is consistent with the requirements of the TA-96. The same assessment rate is used for each carrier. Additionally, a large funding base will have the effect of lessening the funding obligations of individual carriers, ensuring that the mechanism does not in itself act as a barrier to entry.

Bell's argument that the definition of universal service will evolve over time and may need to recognize "special concerns" which apply to different portions of the state is contrary to the objectives of both Chapter 30 and the Federal Act. While the Federal Act recognizes that each state may necessarily have its own separate definition of BUS, it in no instance contemplates or condones the notion advanced by Bell that the Commission may want to define BUS differently in different regions of the state. We believe that neither Chapter 30 or the Federal Act would condone such a result. In fact, the provisions of both require comparability of service between rural and urban areas of the state. The primary purpose of the state funding mechanism is to eliminate barriers to entry and ensure affordable rates in high cost areas; not to create barriers to entry or create a system of telecommunications "haves" and "have nots."

Second, the use of multiple funds, particularly one fund for every existing ILEC service territory, would lead to absurd results. It makes little sense to establish a separate fund for every LEC service territory in the state particularly since this would lead to high cost areas funding their own universal service requirements, a scenario which would have little practical benefit for high cost areas. For instance, consider the impact of Bell's proposal when a company that serves only high cost exchanges is involved. Under the multi-fund approach, no carriers would be willing to serve high cost areas because the amount necessary to ensure universal service in those areas would be exorbitant, and rates in high cost areas would suffer as a result. Hence, none of the primary benefits of the funding mechanism would be realized.

Third, a multiple fund approach would lead to an administrative nightmare. The Commission and/or the third party administrator which it chose would have to deal with a different assessment rate for each service territory or exchange, depending upon how eligibility is ultimately determined. Each service territory or exchange would have a different group of carriers serving the area. Assessment rates would have to be based upon the various carriers serving each market and the percentage of gross intrastate operating revenues each carrier received. The Commission would likely have to employ several third party administrators just to implement and carry out this scenario.

Finally, we agree with GTEN that the multi-fund approach fails to recognize the inherent benefits of a state-wide network. "A subscriber in a particular service area benefits from the fact that more people are on the network throughout the state, and not just in the subscriber's own serving area." It also fails to recognize that the inherent obligations and advantages in being a ubiquitous service provider in the state. The principal reason behind the multi-fund approach appears to be a concern with one company subsidizing another company. We do not find this concern to be a persuasive reason for abandoning the one fund concept. Where one carrier through regulation has obtained the envied position of being a ubiquitous service provider, there are certain obligations which it must undertake. We also note that no other state has considered or adopted a multi-fund approach.

Based upon the above discussion, we reject the multi-fund approach. A single universal fund will be more conducive to effective competition in Pennsylvania, will ensure compliance with the TA-96, will be easier to administer, and will best promote the three primary goals of Chapter 30. We find that these are all compelling reasons for adopting a one-fund approach at this time.

D. Costing Models

1. Design Assessment Criteria

At least four separate models were introduced into the record, with their respective sponsors each claiming that their model alone most accurately measured the costs associated with providing BUS. Sprint/United introduced the Benchmark Cost Model ("BCM"); and an improved version entitled the BCM 2. AT&T and MCI jointly sponsored a model by Hatfield Associates ("Hatfield"), called the "Hatfield Model". They also introduced an improved version of the model which incorporated aspects of the BCM. The OCA utilized a proxy model constructed by OCA's expert witness Dr. Ben Johnson which for purposes of this proceeding we will call the "Johnson Model". Additionally, Bell submitted its own model for use in its own service areas.

For purposes of our evaluation and comparison of the various models presented for our review, we will use the following generally accepted design assessment criteria:

(1) Is the model nonproprietary and replicable. Can it be independently tested and validated? Can the inputs and assumptions be readily verified?
(2) Can the model be applied on a statewide basis? Is it equally applicable to both large and small LECs alike?
(3) Does the model enable analysis on a highly disaggregated basis?
(4) Does the model use the best available technology and information? Does it utilize forward-looking technology and information?
(5) Is the model readily adaptable to changes in the definition of basic universal service?
(6) Does the model design accurately reflect the costs of putting in place the loop and switch necessary to provide BUS?

We note that the evidence and/or testimony of the various parties on the individual cost models proposed for our review appears to coincide with the above-listed assessment criteria. In other words, without specifically identifying the criteria listed above, many parties nonetheless evaluated the various models using these standards. In addition to the assessment criteria just discussed, we will examine whether the model has been considered for Federal funding purposes. In other words, is there evidence of wider use of the model.

2. The Bell Cost Model and Study

The Bell cost studies and methodologies are based on largely proprietary cost models that have been developed by Bell Communications Research ("Bellcore"). Bell states that it has followed the TSLRIC approach suggested by the Commission in its September 5, 1996, Order. Bell defines TSLRIC as "the difference between the costs to the firm to produce all of its products or services, and the cost to the firm if it does not produce the particular service, family of services, or group of services under study." Id. at 4-5.

Bell states that its cost study contains the forward looking costs of [Bell's] ubiquitous network structure and the true economic costs of providing BUS including: (1) the TSLRIC of basic universal service at the wire center level; (2) shared-fixed costs; (3) common overhead costs; (4) community service costs; and (5) other costs." Bell Stmt. 1 (Sanford) at 3. Bell states that its cost study "reflects the forward looking costs of [its] ubiquitous telecommunications network, rather than the fictitious costs of a hypothetical network." Id. at 4, emphasis in the original.

Bell's study disaggregated the relevant cost output information to the wireline center level. Bell's Witness Sanford points out that the wireline center "can encompass one or more end office switches." Id. at 5, n. 2. Although Bell "does not believe that business lines should be included in the definition of Basic Universal Service", because they are not included in the Commission's definition of BUS, Bell has "included single-line business service in determining [its] Basic Universal Service Obligation and Subsidy." Id. at 6. Bell's cost study includes shared and common costs. Bell's shared-fixed costs were allocated based on the proportion of direct investment while its common overhead costs were allocated through the use of a proprietary common overhead factor applied to Bell's BUS obligation. Id. at 6-7.

Bell's TSLRIC study calculates the costs of BUS based on the following proprietary Bellcore models. The Bellcore Ultimate Allocation Area Analysis model ("UAAA") and the Loop Cost Analysis Model ("LCAM") are utilized in order to simulate the necessary facilities between Bell's central office ("CO") switch and the customer's location and to compute associated costs. The UAAA model develops various loop characteristics and cable investments in an Ultimate Allocation Area ("UAA"), which is the section of a wireline center that is used for planning and engineering purposes. Id., Appendix A, at 2-3, n. 4. The LCAM model determines the total forward looking loop cost per month for each individual wire center. Inputs to the LCAM model include factors such as the output from the UAAA model, loop electronics equipment investments and capacities; utilization; pole, conduit, land and building investment factors; switch termination costs per line for each wire center; billing cost per line; directory listing cost per line; and annual and operating cost factors. The LCAM model arrives at the total copper and fiber optic investment per line for each wire center. The utilization factor is then used to derive Bell's spare cable capacity. The LCAM model then computes the monthly costs of these investments on a per loop basis through the use of annual cost factors developed in the capital and cost model and divided by twelve. Bell Stmt. 1.0 (Sanford) Appendix A, at 5-6.

Bellcore's Local Usage module calculates the forward looking cost to set-up a call and the per minute cost of a local call. The components of local usage include switching, interoffice trunks, Signaling System 7 ("SS7"), and billing for measured usage. Id. at 9. In turn, the cost of originating and terminating local calls in the end offices and tandem switches of Bell's network, is calculated on the basis of switching investments per call and per minute derived from the Bellcore Switching Cost Information System ("SCIS") model for each CO switch type. Correspondingly, the Bellcore Common Channel Signaling Cost Information System ("CCSCIS") is used to develop the SS7 investment per call. The investment per set-up and per minute for switching, interoffice trunking, and SS7 are then multiplied by the annual capital and operating expense cost factors to arrive at a cost per set up and per minute of use. The billing cost for measured calls is then added to the cost per set-up. Id., Appendix A, at 9-10.

The Bellcore Capcost+ model is used to calculate forward looking annual capital and operating expense factors for each investment account for each service. The Capcost+ model uses as inputs such financial data as the cost of money, depreciation lives, the debt and equity ratio, material and labor inflation factors, labor productivity factors, maintenance factors, administration factors, tax factors, and demand. Id., Appendix A, at 10-11.

The "community service cost" module captures certain costs that Bell accrues for free operator and directory assistance calls, the annoyance call bureau, unreimbursed 9-1-1 administrative costs, and unreimbursed administrative expenses relating to Pennsylvania's Telecommunications Relay Service ("TRS"). Id., Appendix A, at 11-14. As it has been explained previously, Bell's cost model accounts for certain "other costs" (Bell's revenue reduction from the February 1996 implementation of its Lifeline Service and bill inquiry and account maintenance costs), and for shared-fixed and common overhead costs. The subsidy module in Bell's cost model computes the amount of Bell's alleged subsidy for the provision of BUS. Id., Appendix A, at 14-17.

Bell's cost study also accounted for a certain "other costs" category. This category includes Bell's "...revenue reduction from Lifeline service on an annual basis", and Bell's "...costs for bill inquiry and account maintenance associated with local access, which are shared costs that are not captured in any other category above and are incurred as a result of providing Basic Universal Service." Id. at 8. Bell's cost study incorporated updated assumptions regarding the deployment of its loop facilities and its commitment for broadband facilities deployment under its Chapter 30 Alternative Regulation and Network Modernization Plan ("NMP"), that was originally adopted by this Commission through the June 28, 1994, Order at Docket No. P-00930715. Id.

Bell concludes that its residence BUS is subsidized in the majority of its wireline centers. Bell draws this conclusion by comparing the TSLRIC results for BUS with the revenues received on a wireline center basis. In this analysis, Bell includes single-line residence and business revenues from dial-tone lines, the FCC-mandated subscriber line charge ("SLC"), and touchtone revenues. Bell does not include the revenues for local usage in its subsidy analysis since it supports the inclusion of only a small level of usage within the local calling area (if usage is to be included at all), and Bell argues that it does not receive any revenues from its Budget service, where the Budget service provides the smallest usage package to residence customers. Furthermore, the local usage of Bell's business customers is measured, therefore, no costs or revenues were included in Bell's subsidy analysis. Id. at 10.

Bell argues that only its model can accurately depict its costs of providing BUS. Bell states that it is not necessary that all companies use the same model. Rather, Bell advocates that each of the parties use their own model in calculating BUS. Bell Main Brief at p. 25. No party, other than Bell itself, supports the use of the Bell model to calculate BUS cost in Pennsylvania.

a. Criticisms of the Bell Model

1. Model Design

One of the major criticisms of Bell's study is that it reflects "embedded" costs rather than forward-looking costs. Bell has taken the position that although a TSLRIC cost study should be on a "looking forward" basis, it nevertheless must reflect the actual telecommunications network of the carrier in question and the actual associated network costs. Characteristically, Bell states that its "cost study -- unlike the BCM [Benchmark Cost Model] and Hatfield proxy model or Dr. Johnson's model -- is based on [Bell's] actual cost and actual network", and that Bell's "model is not based on the fictitious costs of a hypothetical network, nor is it a proxy for the actual network." Bell Reply Brief, at 14-15, emphasis added.

Bell's insistence on "actual network costs" has engendered the criticism of other parties to the proceeding, especially those who have based their respective TSLRIC cost studies on proxy models. For example, MCI witness Mercer points out that Bell's use of information regarding existing investments in loop plant "...provides a historical view of loop facilities, not a forward-looking treatment of what is actually required to provide universal service." MCI's Witness Mercer expresses a similar criticism regarding Bell's use of information regarding its interoffice facilities for purposes of its TSLRIC cost study. AT&T/MCI Stmt. 1.1 (Mercer) at 3. Bell denies that its cost study is an embedded study and maintains the position that its cost study is a forward-looking TSLRIC study. Bell Reply Brief at 22.

Another major criticism levelled against Bell's TSLRIC cost models relates to the use of the Bellcore proprietary cost models. The Bellcore models and their underlying computational logic are not available in the public domain, neither are they open to the degree of scrutiny by other parties in this proceeding. AT&T/MCI witness Mercer testified that:

Just as important, and in sharp contrast to both the Hatfield Model and the BCM, the ...[Bell] study is premised on a proprietary model that is susceptible to interpretation and use only by ...[Bell], and that appears to be applicable only to ...[Bell's] network. ...This is not simply a result of the study's use of [Bell] specific proprietary information. Rather, it is a function of the study's reliance on internal models and algorithms that are proprietary to Bellcore, and thus apparently not even subject to release by ...[Bell]. Under these circumstances, the Commission has no opportunity to even evaluate the usefulness of the model in ...[Bell's] territory, much less apply ...[Bell's] methodology to any other Pennsylvania ILEC. AT&T Main Brief at 44, citing AT&T/MCI Stmt. 1.1 (Mercer) at 4.(8)

Bell counters that its cost study should not be discounted "...because it contains material that is proprietary to Bellcore." Bell Reply Brief at 18, footnote omitted. Bell states in support of its argument that the "Bellcore models,... unlike all the other models before the Commission, are widely accepted and have had their general accuracy independently confirmed before the FCC", where such "FCC determination was itself affirmed in federal court." Id., n. 54, citing Commission Requirements For Cost Support Material to be Filed with Open Network Architecture Tariffs, 7 FCC Rcd. 1526, 1533 (Com. Car. Bur. 1992); Allnet Communications Services, Inc. v. FCC, 800 F.Supp. 984 (D.D.C. 1992, aff'd Allnet Communications Services Inc. v. FCC, No. 92-5351 (D.C. Cir. 1994).

Bell distinguishes the evidentiary treatment of its TSLRIC cost study and of the Bellcore models in the Chapter 30 Competitive Safeguards Investigation at Docket No. M-00940587, from the treatment that we should accord to its TSLRIC cost study in the instant proceeding.(9)

Bell states that neither itself or Bellcore "refused to answer a single question posed by [OCA's witness Dr. Johnson..." and that "...all the cost models presented in this docket contained some proprietary information that the sponsoring party refused to disclose." Bell Reply Brief at 20, emphasis in the original, footnotes omitted.(10)

2. Model Inputs

Many parties disagreed on the proper allocation of the joint costs of Bell's network. This particular issue receives more analytical discussion in other parts of our Order. However, additional issues regarding Bell's calculation of its BUS subsidy involve the inclusion of the appropriate costs and revenues.

Teleport claims that the costs which "taint" Bell's subsidy quantification are Bell's costs for buried cable, its costs to deploy broadband facilities for residential local exchange service, and its level of depreciation rates. Teleport Main Brief at p. 22. REA costs are $1.39 while Bell's costs are significantly higher. TCG Stmt. 2.1 (Gabel) at 10-11; TCG Main Brief at p. 22. Concern was also raised regarding Bell's loop costs due to excessively high depreciation rates adding significantly to subsidy levels. Teleport Main Brief at p. 23. TCG Witness Gabel testified that while the technical operating requirements of broadband facilities could make some older metallic and analog switching facilities obsolete, the operating requirements of basic local service have not changed. Thus there is no need to accelerate plant retirements for facilities that provide this service. TCG Stmt. 2.1 (Gabel) at 12-13; Teleport Main Brief at p. 23. Overall, Teleport asserts that Bell's revenues exceed its costs to provide local exchange services, and that there is no subsidy.

Teleport also objects to Bell's inclusion of broadband facilities in its network as a cost of basic local exchange service. TCG witness Gabel testified that these facilities are unnecessary for the provision of voice grade local exchange service. TCG Stmt. 2.0 (Gabel) at 14; PCTA Stmt. 1.1 (Townsend) at 11; Teleport Main Brief at p. 22. TCG objects to Bell's having used a minimal revenue generating usage service but the most expensive available technology for costing services. TCG Main Brief at p. 23.(11)

3. The Hatfield Cost Model

AT&T and MCI are the sponsors of the Hatfield Cost Model. The ultimate output of the Hatfield Cost Model is the monthly cost per line of BUS based on capital investment per line, expense factors based primarily on Uniform System of Accounts ("USOA") definitions, and capital cost and depreciation figures. AT&T/MCI Stmt. 1.0 (Mercer) at 5. The Hatfield Model's application to Pennsylvania ILECs indicates that relevant population densities have a substantial impact on the monthly per line cost.

The Hatfield Cost Model is based on a local network module which contains an engineering module of the local telephone network infrastructure that would be used to provide BUS in the particular area being studied. AT&T/MCI Stmt. 1.0 (Mercer) at 6. Key inputs to the local network module include the demographic characteristics of the area being studied and the capacity limits for the various network elements that make up the local network. The local network module determines as outputs the type and amounts of network equipment required to provide BUS in the area studied, including distribution (local loops) and local end office switching. In turn, these outputs, along with relevant unit costs, are used as inputs into the Hatfield Model's investment module.

The investment module develops an estimate of the capital investment required for the various types of network equipment. The outputs of the investment module are subsequently used in two ways. First, they become inputs to the Hatfield Model capital cost module, which also uses as inputs such financial parameters as the cost of capital (cost rates for common equity and debt and the relevant capital structure common equity and debt ratios), the economic life of the network equipment, and the combined state and federal income tax rate. The capital cost module then derives the monthly per line capital carrying costs. Second, the outputs of the investment module become inputs into the expense module of the Hatfield Module. The expense module produces the monthly per line Operations, Administration and Maintenance ("OA&M") expenses. The addition of the monthly per line carrying costs with the corresponding figure for OA&M expenses produces the estimated monthly cost per line of BUS. Id. at 7.

Since its original formal presentation to the Commission in July 1995, the Hatfield Cost Model has undergone a series of changes. Most significantly, the Hatfield Model has incorporated the BCM jointly developed by MCI, Sprint, NYNEX and US West, which is being sponsored in the instant proceeding by Sprint/United. In particular, the Hatfield Model now utilizes the BCM data base which assigns each CBG to specific ILEC wire centers. Consequently, the results of the Hatfield Cost Model can be disaggregated to the wire center level. The results can also be aggregated to a higher level, such as by population density zones, by telephone company service territory or by entire state. The BCM's use in the Hatfield Model also allows incorporation of terrain data by CBG. In turn, this facilitates the estimation by CBG of the feeder and distribution cable distances and terrain factors, which are utilized in the calculation of amounts and associated structure multipliers of copper distribution cable, copper feeder cable, and fiber feeder cable.

Incorporation of the BCM in the Hatfield Model has also led to improvements with regard to estimating the distribution portion of the local exchange network. For example, the BCM considers both an architecture that utilizes digital loop carrier ("DLC") in the feeder portion of the network and an architecture that extends individual wire pairs all the way back to the serving wire center, selecting the appropriate architecture according to total loop length. The BCM has also added factors measuring the difficulty of network installation based on the type of terrain for each CBG, which is extremely useful for the estimation of outside plant installation costs. Id. at 10.

The Hatfield Model has also redefined its method of simulating the central office ("CO") switching investment through its integration of the BCM. According to AT&T/MCI witness Mercer, previous versions of the Hatfield Model utilized a "desert start" approach which "...simply determined the ideal number and location of switches independently of the existing wire centers." Through the incorporation of the BCM, the Hatfield Model is now able "...to locate the switches at actual LEC wire centers, and more properly size the switches at a given wire center based on the total population of the CBGs served by that wire center." AT&T/MCI Stmt. 1.0 (Mercer) at 11. AT&T/MCI witness Mercer goes on to explain that although the "effect on investments of using actual wire centers instead of the `desert start' wire centers is not clear... using data for the actual CBGs, which are much smaller areas, and using actual wire center locations which have presumably been selected by LECs in a manner that reflects the location of the customers served by each wire center, the BCM may tend to produce less costly arrangements than did our model." Id.

The Hatfield Cost Model, however, still differs from the BCM in various respects.(12) The Hatfield Model modifies the BCM estimates of access lines in each LEC wire center in order to account for business lines and multi-line residences. The Hatfield Model accounts for business lines and for residences with multiple access lines, i.e., residences with a second access line, by using the FCC Common Carrier Statistics, and adjusts the size of the local network loop plant accordingly. These types of access lines are not included in the loop facilities calculations of the BCM.

In contrast to the BCM, the Hatfield Model utilizes significantly lower investments per line for DLC equipment. The Hatfield Model is capable of utilizing intermediate BCM results in order to disaggregate the network loop investment into categories and to compute more discrete capital carrying costs for each category as well as operational costs through the application of FCC Automated Reporting Management of Information System ("ARMIS") expense factors. On the other hand, the BCM makes overall calculations of expenses and capital carrying costs based on the total investment in the distribution network. Id. at 13. The Hatfield Model has also adopted certain BCM methodological parameters relating to the line fill factors for outside plant and the investment in underground conduit that carries cable in areas of high population density. AT&T/MCI witness Mercer acknowledges that the BCM use of low line fill factors -- 25% in areas of low population density -- may lead to higher cost results due to the excess amount of outside plant that is required to serve a given number of customers and that this is a concern. Similarly, the fact that the BCM makes no provision for underground conduit investment in high population areas, may not allocate the proper amount of investment to BUS, and may underestimate the material and installation costs associated with the distribution network in urban areas. Id.

The Hatfield Model local network engineering module utilizes as input the end office switch capacity in terms of the maximum lines served. The Hatfield Model is able to calculate the number of CO switches that are required to serve each wire center and also determines the size and cost of these switches. This determination is made in accordance with the number of access lines served by each wire center and the capacity limits of the switching system. Id. at 14. However, the BCM does not currently provide information necessary to determine the location of tandem switches or the way in which the various wire centers are interconnected. It should be pointed out that although the Hatfield Model is attempting to deal with this problem, AT&T/MCI witness Mercer states that "...the investment in interoffice transmission, tandem switching and signaling was only a small fraction -- less than 3% -- of the total investment..." and that the "3% figure translates into an average addition to the cost per line per month of Basic Universal Service of less than a dollar." Id. at 14-15.

The Hatfield Cost Model utilizes various assumptions for its calculations. Equipment capacities for DLC and cable optical and copper facilities used by the BCM are based on systems that are widely used by LECs, and stated in various industry publications, including those of Bellcore and AT&T, as well as in FCC reports. Assumptions regarding the capacity of switching systems were developed after discussions with industry sources and through the use of promotional and technical literature from CO switching equipment manufacturers. Similar approaches and information sources were used for the development of the associated equipment investment and installation costs. Id. at 16.

The capital investment carrying costs include depreciation expenses, the cost of capital, and state and federal income taxes. A composite depreciation life of 18 years has been used in the Hatfield Model for the calculation of the depreciation expense through the straight-line depreciation method, i.e., 1/18th of the total capital investment is taken as depreciation expense each year over the life of the investment. The overall cost of capital is determined to be 11.4% and it is based on a 40:60 debt to common equity ratio with a cost of debt at 7% (this implies a before-tax 14.3% cost of common equity capital estimate). Since depreciation leads to the reduction of the capital investment amounts over time, the required return amounts have been levelized over the assumed life of the investment through a present value calculation. Id. at 16-17. The Hatfield Model utilizes an assumed combined federal and state income tax rate of 40%. This combined tax rate is utilized to increase the pre-tax return amounts so that the after-tax return is equal to the assumed cost of capital since the common equity component of the return is subject to federal and state income tax.

The Hatfield Model includes the following network-related expenses in its calculations: Network Support, General Support, Central Office Switching, Central Office Transmission, Cable and Wire, Provisioning, Network Operations, Call Completion and Billing and Collection Expenses. A factor for Uncollectibles is also included. The Hatfield Model excludes expense categories associated with the provision of operator services, directory assistance, and "911" services or functions. The developers of the Hatfield Model rely on the Commission's definition of BUS which specifies that access should be provided to these services but does not include the services themselves. The Model also excludes installation costs since customers are separately charged for installation. Similarly, Marketing Expenses and some Customer Services Expenses are also excluded. Id. at 18-19.

The Hatfield Cost Model relies primarily on Bell's historical data as recorded in the relevant FCC 1994 ARMIS No. 43-03 Reports, for the development of certain expense factors. These expense factors, when applied to the investment outputs of the Model, provide the relevant operational expense estimates. This approach is used in lieu of the determination of labor and material costs of major operating functions for each telephone company. Certain expense categories are assumed to vary directly with capital investment, e.g., expenses related to Cable and Wire facilities, while others are assumed to vary directly with the number of access lines, e.g., expenses for Network Operations. Id. at 19. The Hatfield Model, however, does not absolutely rely on historical expense factors for the development of operational expenses.

One exception is the development of OA&M expenses for central office switching. There, the Hatfield Model utilizes an expense factor from a publicly available New England Telephone cost study rather than relying on Bell's historical expense data. The reasons are that Bell's historical data may include software expenses that the Hatfield Model capitalizes, and that Bell's network still utilizes some older switches with more expensive maintenance than current switching technologies. Id at 20.

The Hatfield Model has excluded marketing expenses from its computations since such expenses also reflect advertising and sales costs for competitive or discretionary services. The Hatfield Model initially utilized a 10% overhead loading factor in its calculations based on Bell's historical overhead expenses, however, this was subsequently adjusted down to 6% based upon an examination of revenues and overheads in the auto manufacturing and airline industries. Id. at 21-22.

The investment module of the Hatfield Cost Model does not calculate capital investment in Furniture, Office Equipment or General Purpose Computers. Instead, the Model utilizes actual 1994 Bell figures to calculate the ratio of the above categories of investments to the overall Bell level of investment. This ratio then has been multiplied with the Hatfield Model total BUS investment estimate, and the result has been treated as an additional investment needed to provide BUS. Id. at 22.

The February 1996 Hatfield Model submission calculated BUS costs in the following study areas:

1) The same six population density zones for Bell as they were modeled in the Hatfield July 1995 study; namely, 0-10, 10-100, 100-500, 500-1,000, 1,000-5,000, and greater than 5,000 people per km2;
2) The same six density zones for GTE North and United Telephone Company of Pennsylvania; and
3) The cities of Philadelphia, Pittsburgh and Punxsutawney.

The Hatfield Model has also produced TSLRIC results for Bell by excluding joint, shared and common costs. These results were produced through the elimination of investment and expenses associated with: 1) the distribution network, 2) overhead, and 3) furniture, office equipment, and general purpose computers.

a. Criticisms

1. Model design

Bell faults the Hatfield Model's method for including business lines and the access lines of multi-line residences and sizing the local network. According to Bell, the Model assumes constant ratios between total lines to residence lines in each CBG for a given density cell with the result that "downtown business districts" are treated "as sparsely populated rural areas rather than the most dense (in terms of access lines) area of the state." Id. at 30, footnotes omitted.

In addition, GTEN criticizes the following two aspects of the Hatfield Model's design:

-- The Hatfield Model fails to calculate an incremental cost of BUS, but instead calculates a stand-alone cost of a group of services; the inclusion of the additional lines associated with the other services substantially reduces the average cost per line calculated by the Hatfield Model;
-- The Hatfield Model's use of density zones or wire center "averages" obscure TSLRIC BUS cost differentials on the CBG level.
GTEN Main Brief at 15-17, footnotes omitted.

AT&T responds to Bell's criticism that the proxy models in general produce "fictitious results" by pointing out that three ILECs developed the BCM, and that both the BCM and the Hatfield Model thus reflect realistic network designs based on typical LEC design and engineering practices. AT&T Main Brief at p. 41.

2. Model Inputs

Bell argues that the Hatfield Model uses an unrealistically low annual charge factor of 22.97% (as opposed to the BCM annual charge factor of 31.67%) through the use of ARMIS figures. Bell Main Brief at 29. Bell references PTA's suggestion that the Commission should use company-specific carrying factors rather than the ARMIS factors. Id. n. 90.

Bell also criticizes the Hatfield Model's use of an 18-year life for the development of its composite depreciation expense. Bell argues that an 18-year period is not an economic life and does not reflect the probabilities of technological obsolescence, equipment failures and usage limitations. Similarly, Bell criticizes the Hatfield Model for utilizing the incorrect composite tax rate for Bell in the Model's computations. Bell's conclusion is that the Hatfield Model is "even more flawed than the BCM." Id. at 31.

GTEN argues that the Hatfield Model suffers from the following input problems:

-- The Hatfield Model incorrectly estimates switching costs by grossly underestimating the cost of small central offices;
-- The Hatfield Model calculates unreasonably low structure costs as the result of its use of deeply discounted materials prices;
-- The Hatfield Model uses a composite average capital equipment life of 18 years, which is artificially long and results in artificially low costs;

-- The Hatfield Model excludes the costs of establishing new customer accounts;

-- The Hatfield Model recommends an overhead loading factor of six percent which is based on a comparison with firms in the auto manufacturing and airline industries -- industries that are totally unsuitable for comparison with the regulated market in which LECs operate.

AT&T acknowledges that some refinements are necessary for the omission of interoffice facilities and modifications to the switch algorithm that will permit a separate identification of fixed and usage-sensitive switching costs. AT&T Main Brief at p. 40. However, AT&T states that none of these matters provides any reasons to question the overall accuracy of the cost estimates provided by the Hatfield Model.

Finally, in response to the criticisms levelled by GTEN, AT&T states that Dr. Mercer explicitly adopted a conservative approach in selecting the inputs to the model in order to avoid underestimating the cost of BUS. AT&T states that if GTE or any ILEC can provide data that is more accurate than the inputs now used in the Hatfield Model, that new data can easily be accommodated in the model. However, according to AT&T, no party has provided this information. AT&T Main Brief at p. 42. Finally, AT&T states that the Hatfield Model, particularly with the inclusion of the BCM, clearly uses real data to derive accurate estimates of the cost of BUS, and provides a powerful, adaptable tool for the Commission's use on a going-forward basis. AT&T Main Brief at pps. 42-43.

4. The Benchmark Cost Model

According to United/Sprint witness Dunbar, the BCM "was developed as a joint effort by Sprint, NYNEX, MCI, and US WEST (joint sponsors) in response to the FCC's expressed interest in considering a model which develops `proxy' costs for the provision of basic telephone service at the CBG level." The BCM "was filed on September 12, 1995 with the FCC by the joint sponsors for consideration in CC Docket No. 80-286 (USF proceeding)" and it represents a considerable improvement over earlier local loop and CBG models. Sprint/United Stmt. (Dunbar) 2.0 at 4. Since the inception of the BCM, its joint sponsors have presented the BCM model results for various states and before state utility regulatory commissions and the FCC. United/Sprint Main Brief at 25-26.

The BCM produces a benchmark cost range for a defined set of basic residential telephone services assuming efficient engineering and design criteria and the deployment of current-state-of-the-art loop and switching technology. The BCM uses the current national local exchange network topology. The BCM provides a benchmark measurement of the relative costs of serving customers residing in a given area, i.e., the Census Block Group ("CBG")(13) level. Sprint/United Stmt. 2.0 (Dunbar) at 4.

The stated purpose of the BCM is to identify CBGs in which the cost of providing basic telephone service is so high that some form of explicit, high-cost support may be necessary as part of a universal service solution at both the Federal and state levels. Sprint/United Main Brief at 26. The BCM does not define the actual cost of any telephone company, nor the embedded cost that a telephone company might experience in providing telephone service today. Id. at 28.

The BCM included only residential lines in its analysis, because business line source data was not readily available. However, Sprint/United argues that because the primary purpose of the study is to identify high cost CBGs, the impact of excluding business lines from the BCM calculations in those CBGs is de minimis. Sprint/United Main Brief at 28.

The BCM modeling process includes four major steps. First each CBG is associated with the nearest central office ("CO"), using the distance between the centroid of the CBG and the CO location. The second step involves the determination of the appropriate feeder and distribution plant for the relevant CBG location. The BCM assigns all CBGs in a quadrant to a single shared feeder and selects the appropriate loop technology for each CBG. The Model then sizes and prices the feeder and distribution cables. Sprint/United Main Brief at 29-30. The third step relates to the development of the appropriate placement costs for the BCM derived network. The BCM utilizes for this purpose U.S. Government data for terrain and density in order to develop the estimates of loop placement costs within the CBG. Id. at 30.(14)

The BCM then develops the appropriate switching costs associated with serving each CBG. Those investments are calculated based on current CO locations, and the use of capital investment in Nortel (formerly Northern Telecom) DMS 100 fully digital CO switching equipment. In addition, the BCM divides the switching costs on the basis of common switch costs, such as the central processor frames and equipment dedicated to billing and data recording, and the costs that vary by access line, such as line cards. Sprint/United Main Brief at 30, citing Sprint/United Stmt. 2.0 (Dunbar) at pps. 6-7, 26.

The BCM contains the following underlying assumptions relating to customer dispersion. The BCM assumes that households are uniformly distributed throughout the CBG. Sprint/United Stmt. (Dunbar) 2.0 at p. 7. This assumption plays a role in the siting of the local loop facilities and the associated feeder routes. In using the CBG, it is assumed that feeder facilities begin at the CO and generally end at the end of the CBG unless the CBG overlaps the main feeder. In that case, the plant that diverges from the main feeder is considered to be distribution plant. The BCM utilizes four equally sized distribution cables within the CBG. Id. at 8.

The BCM also utilizes the following assumptions in developing its feeder and distribution plant architecture:

-- Cable and fiber feeder systems share structures such as conduit systems and poles along main feeder routes and are placed at the same time.
-- The cable size of each feeder segment (both fiber and copper) is determined based upon the number of households in the CBGs served by that feeder segment as well as the fill factor applicable to the CBG.
-- Feeder cable sizes range from 100 pair cable to 4200 pair cable for copper facilities and from 12 strands to 144 strands for fiber optic facilities.
-- Analog copper technology is assumed for all distribution facilities.

-- Four equal distribution legs are used to service a CBG.

-- Distribution cable sizes range from 50 pair cable to 3600 pair cable for copper facilities.
The following loop technologies are incorporated into the BCM:
-- Copper. All distribution of loops less than 12,000 feet long are assumed to be carried on copper pair cable. All distribution plant is assumed to use copper loop technology.
-- Fiber -- Subscriber Loop Carrier 2000. If the loop is greater than 12,000 feet in length and CBG density is greater than 5 households per square mile, the loop technology is assumed to be Digital Subscriber Line Carrier Series 2000 with fiber feeder. With this technology, a remote terminal is located at the edge of the CBG where the feeder plant ends. Each CBG uses a minimum of four dedicated fibers to provide up to 672 voice grade paths. The 672 figure is reduced by whatever fill factor is entered into the model.
-- Fiber -- Advanced Fiber Communications ("AFC") Next Generation. If the loop is greater than 12,000 feet in length and the CBG density is less than 5 households/mi, the loop technology is assumed to be Digital Fiber Loop Carrier Bus Technology (Advanced Fiber Communications next general digital loop carrier system) with fiber feeder. With this technology, a remote terminal is located at the edge of the CBG where the feeder plant ends. Each CBG uses a minimum of four fibers to provide 672 voice grade paths, but the same four fibers may continue to more remote CBGs within capacity constraints. Again, the 672 is reduced by whatever fill factor is entered into the model.
Sprint/United St. 2.0 at 8-10 (Dunbar).


The BCM utilizes the following major assumptions regarding CO switching technologies:

-- The source for switching costs is Telecommunications Policy for the 1990s and Beyond,by Walter G. Bolter, James W. McConnaughey, and Fred J. Kelsey (M. E. Sharpe, Inc., Armonk, New York, 1990).
-- The Nortel (formerly Northern Telecom) DMS 100 switch is assumed in the BCM since this was the only public cost information found on switch technology.
-- The Nortel DMS 100 costs are split between common costs and per line costs through regression analysis.
-- The common cost per switch is $647,526 and the per line cost is $238.87.
-- Common costs represent central processor frames, frames and equipment dedicated to billing and data recording, miscellaneous power equipment and emergency backup power systems, main distributing frame plus miscellaneous frames required for overall system testing, and various software expenses to allow the switch general operating capabilities.

The BCM utilizes certain user adjustable inputs in order to derive a Basic Local Service Factor which is used to allocate the CO switching investment. The BCM assumes that the Nortel DMS 100 CO common costs consist of a 70% non-traffic sensitive ("NTS") portion and a 30% traffic sensitive ("TS") one. The BCM further assumes that 30% of the TS costs are allocated to local switching. Id. at 28. Similar techniques and assumptions are utilized in order to derive similar factors for other types of BCM network equipment and facilities.

The BCM households per square mile (density) of a CBG determines the fill factor -- which is a user-adjustable-input -- that is used for plant serving that CBG and the mixture of underground, buried and aerial plant. The BCM utilizes six density groupings for households per square mile. CBGs with densities greater than 850 households per square mile are considered urban, while CBGs with densities less than or equal to 850 are considered rural. Urban CBGs have higher placement costs than rural CBGs, if other terrain indicators are equal. The BCM model also contains numerous assumptions regarding terrain and placement engineering practices. Id. at 10- 11. An additional set of assumptions governs the respective sizes and costs of copper and fiber optic cable, the prices for cable, switching and circuit equipment, and the costs for various structures related to network facilities deployment, e.g., conduit systems, interduct, poles, and the capitalized cost for cable placement. Id. at 12-13.

Throughout the BCM process, all cost calculations are derived in terms of investment. In order to determine a monthly cost for basic local service by CBG, the individual investments for the piece parts must be summed to include loop and structure investments, electronic circuit equipment investments and CO switching investments. An annual cost factor is applied to total investment and divided by 12 to estimate a monthly cost of basic local service. Id. at 28.

The BCM utilizes two different annual cost factors that represent two views of the appropriate level of expenses attributable to basic local service, and provides upper and lower boundaries for the discussion of the monthly cost of basic service. The first annual cost factor of 31.6765% is based on certain 1994 ARMIS Form 43-01 source data. The second annual cost factor of 22.97% is based on the inclusion of limited expense categories and limited expense amounts, and uses the assumptions that are also utilized by the Hatfield Cost Model. Some of these assumptions are:

-- Investment-related expenses of depreciation and after-tax return on investment ("ROI"):

- ROI = 9.5%
- 45/55 Debt to Common Equity Ratio
- 11.0% Cost of Common Equity
- Combined Federal and State Tax Rate of 39%
- 18-year Weighted Average Service Life in Calculating Depreciation

-- Operating and Maintenance Expenses Partially Attributable to Basic Local Service: Network Support; General Support; Central Office Switching; Central Office Transmission; Cable and Wire; Provisioning; Network Operations; Call Completion; and Billing and Collection.

-- Excluded Expenses include certain Customer Services Expenses; Marketing Expenses; and Product Development Expenses.

-- General and Administrative Expenses are based on a 10% Gross-Up level for Overhead. Id. 28-30.

The original BCM -- or BCM-1 -- has undergone significant enhancements and improvements as a result of the intentional exposure of its assumptions and computational logic to public scrutiny and testing. The improved and enhanced version of the BCM, or BCM-2, has been utilized in the present Investigation. In response to our June 6, 1996, Order and July 3, 1996, Secretarial Letter in the instant Docket, on or about August 15, 1996, Sprint/United submitted the results of the BCM-2 computational runs for Pennsylvania ILECs in the instant proceeding. Sprint/United also submitted an accompanying document summarizing the BCM-2 improvements and enhancements. This document had been previously submitted on or about July 3, 1996, to the FCC in the context of the Federal-State Joint Board's Universal Service proceeding at Docket No. 96-45. The enhancements and the improvements of the BCM-2 are outlined below:

-- Enhancements have been made in computing the cost in sparsely populated rural areas. These include:

- BCM-1 assumed a uniform population distribution throughout the CBG. While this assumption is reasonable in some areas, many CBGs contain large non-populated areas. To better identify populated areas, the road network within CBGs of less than 20 households per square mile has been analyzed. In these areas a buffer of 500 feet on either side of the roads has been created to define the populated area. Areas which fall outside of this buffer are excluded from the BCM-2 analysis. The original number of households are assumed to be uniformly distributed in the reduced CBG area.
- The BCM-2 has adopted a maximum investment per wireline loop. This has been done in order to recognize that some telephone service customers may be more reasonably served through wireless loop facilities.

-- Enhancements have also been made in identifying the cost in urban environments. Urban distribution architectures have been modified to better reflect the placement of plant in dense suburban and urban environments. Several network elements not included in BCM-1 are now included in BCM-2. Some of these enhancements include:

- BCM-2 utilizes a public information source in order to determine business access lines in each CBG and includes business lines in the outside plant architecture.
- BCM-2 now includes the cost of the pedestal, the drop wire and the network interface device in its local network modeling. These elements are necessary for the provision of telephone service and add approximately $200 of investment per household. In addition, BCM-2 now includes the costs for engineering, splicing, cross-connects and inter-office trunking which were not included in BCM-1.

-- Enhancements have been made to provide more accuracy and flexibility in the processing of the Model.

- An enhanced switching module has been developed which more accurately determines the cost of central office switching, and better addresses the cost in a CO host/remote switching architecture.
- A variable to account for the impact of terrain slope on outside plant costs has been added.
- The computation of expense elements has been enhanced. BCM-2 has been modified to recognize that some expenses are related to investment, e.g., maintenance, depreciation, return, etc., but other expense categories are related to number of lines, e.g., billing, overheads, etc.


The BCM, with modifications, is also supported by GTEN and PTA. GTEN states that it supports the BCM because "...it is the most relevant, up to date model in the public domain; it is forward looking, long run, incremental from a zero level of service; pertains to actual existing telephone companies; has switching nodes where telephone companies currently have them; and utilizes current outside plant technology and engineering practices." GTEN Initial Brief at p. 13. GTEN goes on to state that "...the BCM with certain minor adjustments provides a reasonable estimate of the cost of the local loop component of universal service." Id. at p. 13. PTA states that in view of the Commission's rejection of embedded cost techniques, it now leans toward endorsement of the BCM. PTA Stmt. 2.0 (Curry) at 3. PTA believes that is the best, overall approach and, with some modifications discussed below, the PTA would endorse its adoption by the Commission.

GTEN believes that one of the most significant advantages of the BCM is its utilization of CBGs for its geographic level of disaggregation. GTEN Initial Brief at p. 13.

a. Criticisms of the Benchmark Costing Model

1. Model Design

The OCA criticizes the BCM on the grounds that it does not produce the results of a true TSLRIC cost study in accordance with the directives of the relevant Commission Orders in the instant proceeding. OCA argues that:

A properly conducted Total Service Long Run Incremental Cost study must reflect the incremental cost and not the average cost of providing service. There is obviously a difference between the average cost of offering residential service on a network where no other service is being provided, and the incremental cost of adding a service to a fully functioning network already in place. As the incremental cost is generally less than the average cost, any model which uses average costs will report costs which are higher. Accordingly, such models will overstate any calculated "subsidy".
OCA Main Brief at 18, footnote omitted, emphasis in the original.

The OCA "emphasizes that the PUC should make certain that the cost models that will be used to calculate costs for a universal service fund must be true incremental cost models." Id. at 20. The OCA argues that this approach "requires the PUC to determine that in any TSLRIC study that will be used for universal service purposes the study must calculate the incremental cost of providing the service in question", and that this "means that the costs studied should only reflect the additional costs incurred to provide service on a network that is capable of providing all other services." Otherwise, according to the OCA, "the costs calculated will not truly reflect incremental costs and will likely overstate the residential universal service costs at issue." Id. emphasis in the original.

OCA witness Johnson also attacked the BCM-1's exclusion of business lines in the following excerpt from his testimony:

Consider the Benchmark cost study submitted by Sprint/United, for example. The BCM study calculates the cost of providing one loop to each household; it excludes any consideration of the loops that would be installed to serve business and private line customers. Thus, it ignores the economies of scope that can be achieved when the same network serves business, private line, and residential customers. In effect, Sprint/United has computed an estimate of the cost of providing BUS to residences, as if no other services were being provided, rather than the incremental cost of providing this service assuming all other current services continue to be provided. This is a fundamental flaw in their study, and it results in significantly higher costs than a true TSLRIC approach.
OCA Main Brief at 18, citing OCA Stmt. 1.2 (Johnson) at 8-9 (emphasis original).

Bell urges the Commission to refrain from utilizing the BCM 1 for determining the cost of BUS 1 unless modified as discussed below. Bell Main Brief at 25-26, footnotes omitted.

-- The BCM should be modified to include an equitable and sustainable portion of shared and common costs of the firm. According to Bell, the "current version of the BCM severely understates the amount of shared and common costs legitimately incurred by carriers of last resort."
-- The BCM (BCM-1) should be modified to size the network based on all lines, not just residential lines, taking into account economies of scale.
-- Customers should be assigned to their correct wire center and to the correct company serving them. The BCM improperly assigns many CBGs to the nearest wire centers, regardless of whether that wire center actually serves any of the customers in the CBG or even belongs to the company serving the area.
-- The BCM should be modified to resemble feasible network designs which may serve the actual distribution of population. Currently, the BCM is severely limited in its engineering assumptions. For example, the BCM assumes each CBG is square in shape rather than the actual CBG boundaries; the feeder cable distances are not based on terrain (the appearance of water bodies, mountain ranges, or other natural features) but consist of straight lines drawn at right angles from the wire center to the CBGs; the actual number of trenches in both the feeder and distribution plant are fixed at four each (four feeder routes per wire center and four cable runs per distribution area); and a population which is uniformly spaced within the square area (that ranges in size from about a city block to hundreds of square miles).
-- The model should include the costs to provide carrier of last resort obligations that are included in the Commission's definition of universal service.

Bell Main Brief at 25-27, footnotes omitted.

Bell states that it "is possible that the new version of the BCM [BCM-2] will correct many of the problems identified by ...[Bell] and other parties." Id. at 28, footnote omitted. Bell also argues that "[c]ost proxy models like the BCM and the Hatfield model do not calculate the actual costs of actual companies, but are surrogates for cost for purposes of assessing magnitudes of funding", and, thus, these "proxy models were not designed to be used -- and should not be used -- for pricing purposes." Id., n. 86. Bell goes on to point out that "Sprint/United witness Mr. Dunbar explained that the BCM model was not designed to be used for pricing but can establish state-wide benchmarks for costing, and that United Telephone does not use the BCM for pricing purposes." Id., citing Tr. at 725 (Dunbar).

With respect to design, the PTA objects to the following:

-- PTA argues that with the exclusion of business lines in the cost modeling of BCM-1, "...there is a disproportionate overstating of the cost per line in CBGs, exchanges...for companies that have a higher than average business to residential ratio... because the model does not recognize the additional economies gained in both feeder and distribution plant for business lines." Id. citing PTA St. 2 at 4. PTA also argues that in "some low density, rural exchanges the cost will be so high that business lines may qualify for USF support to maintain a reasonable local business rate." Id. at 23, citing PTA Stmt. 2.0 (Curry) at 5.

-- PTA argues that the local exchange service "rates are set at the exchange level," thus "it would seem logical that the universal [service] rates, benchmark costs and universal service funding be at the exchange (wire center) level also." Id. at 24 citing PTA St. 2 at 5. PTA goes on to state that based on the experience of the National Exchange Carrier Association ("NECA") in administering 1,450 study areas on a nationwide basis, the "...Pennsylvania USF [Universal Service Fund] with 11,688 CBGs will be extremely complex, difficult to verify, and very expensive to administer" and, therefore, "costs should be calculated using the exchange as the smaller unit of study, rather than the CBG." Id. citing PTA Stmt. 2.0 (Curry) at 6.

2. Model Inputs

PTA has the following concerns regarding the BCM's inputs for small rural telephone companies.

-- BCM-1 utilizes a 31.6765% annual carrying charge factor for cost development that has been derived from ARMIS data. PTA argues that this annual carrying charge factor "...does not appear to be representative for small telephone companies." According to a PTA study, the "...annual carrying charges for PTA companies range from 30.28% to 71.05% with an overall weighted average of 38.25%" and that only "two of the twenty-nine PTA companies studied had an annual carrying charge lower than the 31.6765% used in the Benchmark Cost Model."(15) Thus, PTA argues "...for the inclusion of company specific annual carrying charge factors for calculating each company's benchmark costs." Id. at 22.

-- PTA states that the "BCM accepts only one discount rate for [telecommunications] equipment purchases", where "...larger companies receive significantly bigger discounts on their equipment purchases than do the smaller telephone companies", and that "...if the Universal Service Fund is set on a single percentage point above average benchmark costs, small companies with rural exchanges will be at a distinct disadvantage because discount levels are not reflected on a company specific basis." Id. citing PTA Stmt. 2.0 (Curry) at 5.

In addition, Bell advocates the following input modifications:

-- The Commission should require the use of actual and verifiable data for engineering and planning feeder routes and distribution facilities and for investment and expenses associated with certain loop components, including serving area interfaces, the terminal and the line drop.

-- Plant specific expense should be developed based on realistic estimates. According to Bell, the current version of the BCM (BCM-1) misrepresents annual expenses.

Finally, GTEN advocates certain additions/modifications, including the addition of certain costs that were included within the Commission's universal package (EAS and a white pages listing), different switching costs, a different annual charge factor and a switching startup cost. GTEN Main Brief at p. 14.

5. The Johnson Cost Model

The OCA presented a TSLRIC study that was prepared by OCA's witness Ben Johnson and was based on Dr. Johnson's own cost model for BUS in Pennsylvania. The Johnson Model was made available to all the parties in the proceeding, as well as to the Presiding Administrative Law Judge and to the Commission's legal and technical staff. Since the Johnson Model is based on the Quattro Pro electronic spreadsheet software application package for personal computers, the OCA undertook the extra step of purchasing individual Quattro Pro packages for the Commission's use. Id. at 24, n. 12.

The Johnson Model is structured so that it remains faithful to the TSLRIC definition espoused by Dr. Johnson in his direct testimony:

The total service long run incremental cost (TSLRIC) of a service (or a group of services) is equal to the firm's total cost of producing all its services including the service (or group of services) in question, minus the firm's total cost of producing all its services EXCEPT the service (or group of services) in question. Thus, it is a particular form of long run incremental cost (LRIC), in which the specified increment is the entire volume of output of a particular service, while all other services remain unchanged. OCA St. 1.0 at 7-8, emphasis in the original.

In general terms, the Johnson Model calculates the costs that are associated with two distinct simulated telecommunications networks. The difference between the two cost figures provides the TSLRIC cost figure that is sought for a particular telecommunications service or group of services.

The simulated configurations of the two alternative telecommunications networks that are used for the TSLRIC cost computation are essentially influenced by the Model's user input assumptions. As Dr. Johnson explains in his direct testimony, "...the user specifies the carrier's share of the business and residence markets within two geographic zones -- one close to the central office, the other farther away", where, "in the second category, the [Model's] user specifies whether the carrier will build a network to serve only business customers, only residential customers, or both." OCA Stmt. 1.0 (Johnson) at 25. Thus, the use of "these input assumptions allows the model to simultaneously calculate the costs associated with two different networks, thereby providing direct contrasts of alternative situations" and where "...the user can develop the costs of a network designed to serve only business customers and half the residential customers and place that scenario side by side with that of a ubiquitous network designed to serve everyone." A comparison of "the total costs of these two network configurations directly computes the TSLRIC of adding ubiquitous residential service to a network that would otherwise only serve business customers and price-inelastic residential customers." Id. at 25-26.

In a generally similar fashion, the Johnson Model "can compare the incremental cost of adding service for a [specific geographic] zone two (farther from the central office), to a network that would otherwise serve the [geographic] zone nearest to the central office."

OCA witness Johnson strongly supports the premise that the fundamental methodological approach of his Model is designed to "...directly answer one of the questions of primary concern to the Commission in this proceeding: What is the total incremental cost of providing basic universal service to price elastic customers (those that could potentially abandon the public switched network?" Id. at 26, italics in the original. OCA witness Johnson goes on to explain that "...the focus of this proceeding should not be residential service in the aggregate", but, rather, "the focus should be on residential service as it is provided to those incremental customers that might abandon the public network in the absence of appropriate regulatory policies designed to maintain universal service." Thus, according to witness Johnson, "...in evaluating the issues to be resolved in this proceeding, it is the cost of serving less profitable, more price elastic customers that must be considered." Id.

Like the other models presented for our review, the Johnson Model is based in a number of discrete logic modules. These modules include the Loop module, the Miscellaneous Model (which contained the cost of termination, switching, trunking, billing and collection, and other miscellaneous costs), the Annual Charge Factor Module, and the Summation Module (which combines the results of the earlier modules and applies appropriate adjustments for joint and common costs). The Johnson Model focuses on particular wire centers and produces the corresponding cost results (the Model contains approximately 375 actual wire centers of Bell Atlantic- Pa.).

The Loop Module calculates both the TSLRIC and the ASIC of typical loops within a wire center having user-specified characteristics. The user can custom-define an individual wire center, based upon input values for a wide variety of different factors. For instance, the Model can be adjusted to estimate the extra cost of placing cable facilities under difficult conditions. The additional time required for placing cable in congested urban areas can also be built into the Model through the selection of appropriate input assumptions. The user can set plant utilization factors appropriate to the study in question. The user can also control the assumptions that determine the mix of fiber optic and copper cable and can specify an all-copper network, an all-fiber/digital network, or a combination. Id. at p. 28. The Model also contains data inputs concerning the cost of materials, labor time requirements, and other underlying relationships and calculations that drive the model. Id. at p. 29.

In the Miscellaneous Module, the user inputs a loop cost and allocation factors for joint and common costs, along with a few other key assumptions for a composite cost of BUS. The module contains additional assumptions and detailed calculations that are used in developing the composite cost figure. Termination estimates the cost of the facilities located at or adjacent to the customer's premises. SwitchTrunk provides a simplified estimate of the cost of the end office, as well as tandem switching and interoffice trunking within the local calling area. BillColl contains an estimate of the cost of billing, collection, and segregates this estimate into joint and direct cost items. AnnCost contains the annual charge factors used in this module.

The Annual Charge Module computes the annual carrying charge for a variety of expected plant lives which can then be paired to specific plant items such as copper cable or poles. The module relies upon debt/equity ratios and annual cost rates which can be input by the user, as well as a tax rate.

Dr. Johnson states that while the Model is long run and forward looking, it does assume that central offices will remain in approximately the same location as the incumbent carrier's. Dr. Johnson refers to this assumption as the "scorched node" approach as contrasted to the "scorched earth" one. However, the Johnson Model is adaptable in simulating the market entry of a relatively new BUS provider, where the new entrant will utilize a lesser number of central offices in serving a particular exchange (which amounts to network optimization through the use of the "scorched earth" modeling approach). Id. at 32. Dr. Johnson states that these modifications could potentially yield some reduction in costs when modeling a new entrant with a relatively small market share, requiring fewer central offices to efficiently serve a particular exchange.

The Johnson Model estimated certain TSLRIC BUS costs for the "marginal" or "incremental" customers who will connect with the public switched network through appropriate universal service policies. These costs were calculated by assuming that an incumbent BUS provider serves 100% of the business customers and 50% of the residential customers in an initial network configuration. The incumbent's subsequent network configuration serves all of the residential customers. Id. at 41-42.

The Model has the capability for user-modified variables which are organized by subject area into six categories. The first two categories, Market Shares and Network Type, allow the user to define the size and type of carrier being modeled. Id. at 25.

The Johnson Model allocates 50% of the local loop costs to BUS. Notably, Dr. Johnson states that the deployment of broadband network facilities and services among residential customers, such as video dial tone, is likely to lead to further cost reductions of BUS. Dr. Johnson points out that Bell Atlantic's plans to offer high quality video services on demand, using a full 45 megabits per second ("Mbps") of bandwidth, is equivalent to the provision of 672 voice grade circuits to a single household. Id. at 48-49.

a. Criticisms of the Johnson Model

1. Model Design

Bell attacks the Johnson Model on the basis that it "does not attempt to quantify the total cost of providing universal service as defined by the Commission, but instead measures -- and erroneously labels as 'TSLRIC' -- only the cost of serving the 50% of residential customers which he characterizes as 'marginal customers' or 'the additional customers who can be enticed onto the network through carefully crafted pricing and regulatory policies'." Bell Main Brief at 31-32, citing OCA Stmt. 1.0 (Johnson) at 42 and Tr. at 1526-27 (Mercer). Bell goes on to argue that "Dr. Johnson's model may very roughly calculate how much ...[Bell] would save if it did not have to serve 50% of its customers, but it does not calculate how much it costs ...[Bell] to provide Basic Universal Service to its customers, and it seriously understates the cost of ...[Bell's] network." Id. citing Bell Stmt. 3.2 (Emmerson) at 17.

Bell also advances a series of criticisms relating to the Johnson Model calculation of costs that relate to various categories of telecommunications plant and facilities, especially for the subscriber outside plant. Bell asserts that the Johnson Model improperly excludes certain telecommunications infrastructure network costs from the numerator of the long run incremental unit cost calculations. Id. Bell further argues that the Johnson Model produces illogical results since it calculates a lower cost per line for longer residential loops than for shorter average loop lengths. Id. at 34. Bell concludes by stating that the Johnson Model is unreliable and its results are "meaningless." Id. at 36. The OCA responds that the Bell arguments should be dismissed since they do not represent any substantial difficulties with the actual design of the Johnson Model. OCA Reply Brief at 8-11.

2. Model Inputs

GTEN also levels a series of attacks on the Johnson Model inputs. See GTEN Main Brief at pps. 18-20. GTEN claims that Dr. Johnson's model most significant error is his arbitrary allocation of 50% of the local loop costs to universal service. In addition, he utilizes average wire center loop length for both business and residence, and the model is unable to identify high-cost and low-cost areas within a wire center. GTEN claims that this is a serious drawback because cost differences within a wire center can be substantial. GTEN Reply Brief at p. 18.

GTEN also claims that Dr. Johnson's model suffers from the infirmity of using inputs that lack source documentation. Additionally GTEN claims that additional mathematical errors were discovered during cross-examination including the annual charge factor, an understatement of the tax rate and cost of money.

Bell attacked the Johnson Model assumption for allocating 50% of the local loop costs to BUS. Bell also argues that the Johnson Model inputs relating to the computation of certain annual costs and carrying charges are flawed resulting in substantial reductions on the calculated costs for basic universal service. Id. at 33-34.

6. Comparative Model Assessment

In choosing a model and its various inputs, the Commission must exercise extreme care to neither over-size or under-size the fund. If the Commission over-sizes the fund, an unnecessarily large subsidy will result. If the fund provides windfalls, the subsidized providers will have a competitive advantage over other providers since they will be able to use their windfalls to protect profits while the subsidized provider aggressively competes in competitive markets. Understated costs could cause financial harm to the ILECs and force them to continue to rely upon traditional subsidies that may not be sustainable with competition. See Sprint/United Stmt. 1.0 (Jamison) pp. 7 and 15.

With these principles in mind, we turn to our evaluation of the various models under the seven assessment criteria set out above.

a. Is the model nonproprietary and replicable. Can it be independently tested and validated? Can the inputs and assumptions be readily verified? Is the model user-friendly?

Complete disclosure as to how the universal service subsidy is calculated is essential. The cost formulas used will shift millions of dollars to and from universal service payers and payees. OCA Main Brief at p. 22. The cost formulas may also affect the extent to which consumers actually receive any benefits from the fund. Id. Thus, the effect of what cost formulas are used will be material and critical to all participants. Id.

Most of the models presented contain some proprietary data. However, it appears that virtually all, with the exception of the Bell model, could still be independently tested, replicated and validated.

Bell states that it is unable to disclose its model due to a contract with Bellcore. Tr. 637-38. Specifically the following models were identified as being proprietary to Bellcore: Ultimate Allocation Area Analysis model, Loop Cost Analysis Model, SCIS model, Common Channel Signaling Cost Information System model, and Capcost model. In addition, no information on the input unit costs of various network components, such as switching, loop electronics, cable, and structures were provided. AT&T/MCI Stmt. 1 (Mercer) at p. 4.

As to the proprietary nature of Bell's model, we do not find this to be a fatal flaw with respect to Bell sustaining its burden of proof. However, at the same time, it must be realized that the relative utility of Bell's TSLRIC cost study, as it is premised on Bellcore's proprietary cost models, is of limited value in the context of this investigation. Indeed, it is easy to ascertain that any intended utilization of the Bell TSLRIC cost study within a broader Universal Service costing and funding mechanism applicable throughout the Commonwealth, would suffer from the obstacles of the Bellcore cost model proprietary protection. Arguably, if this Agency and the Administrator of the funding mechanism were to utilize Bell's TSLRIC study and Bellcore's cost models, we would have to continuously weigh Bell's and Bellcore's proprietary protection interest against the needs of wider dissemination and constant validation testing for Bell's TSLRIC study and Bellcore's models.

Even Bell's argument that it willingly responded to all questions put to it is not enough to overcome the burdens associated with a proprietary model given the need for continuous monitoring and periodic adjustment of the TSLRIC cost studies and models and of the associated Universal Service Fund contributions and disbursements, which mandates the utilization of a TSLRIC cost model that could easily be subjected to the widest possible degree of public scrutiny, especially for its computational logic. We believe that such an approach not only will contribute to the improvement of the model's quality and predictive power over time, but will also establish public confidence regarding the Universal Service cost measurements and Fund support mechanisms instituted through the actions of this Commission. The highly proprietary nature of the Bell TSLRIC study and of the underlying Bellcore cost models clearly does not comport with these goals.

OCA witness Johnson aptly summarized the problems with utilizing the Bellcore model, given the considerable proprietary information contained therein, to determine BUS costs on a statewide basis:

While in some cases I found it was possible to trace numbers from one page to the next, I could not work all the way from the initial inputs to the final loop cost estimate. The lack of formulas or algorithms, as well as the absence of detailed calculations prevented me from replicating the methodology, or determining the impact of a change in assumptions. As should be apparent from my earlier testimony, I wanted to evaluate the impact of the underlying assumptions concerning which customers or services were included within the total set of services being produced, and which customers or services were part of the specific increment being analyzed (included or excluded). Using the offered documentation, I wasn't able to calculate the impact of changing these key assumptions. In other words, I could not calculate the impact of changing their assumptions concerning which customers and/or services were included within the increment being analyzed, and which customers and/or services were served by the firm in any event.
OCA Stmt. 1.2 (Johnson) at p. 33.

AT&T argues that the public nature of the Hatfield methodology is one of the major strengths of this model, especially in comparison to "proprietary" studies, such as that put forward by Bell, in that it permits issues regarding the model and its inputs to be raised, examined, and resolved by the Commission. AT&T states that to that end, Hatfield has made full disclosure of the model's processes, assumptions and inputs, as required by the Commission. According to the sponsors of the Hatfield Model, the results of the AT&T/MCI Hatfield Cost Model study demonstrate that "...it is possible to measure the [TSLRIC] cost of basic universal service with reasonable accuracy" and that the Model "is flexible and user-friendly because it would allow costs to be measured at various levels of disaggregation at the Commission's discretion and would allow any errors in assumptions or inputs to be corrected with company-specific data." MCI Main Brief at 6.

Similarly, Sprint/United witness Dunbar testified that the BCM has been made publicly available along with a complete copy of all tables, assumptions, calculations, but with data limited to 50 CBGs. Sprint/United Stmt. 2.1 (Dunbar) p. 11. It will run on any standard PC with Windows and EXCEL version 5.0 or later. The BCM model, input file, output file, intermediate loop cost results file, and summary reports are all on two 3.5" diskettes according to witness Dunbar. Id. He also testified that these diskettes have been made available to all regulators and interested parties. Others also note that the BCM is available to all parties, was developed by a coalition of LECs and IXCs, uses no company proprietary data, and requires no burdensome special data collection. GTEN Stmt. 8.1 (Cole) at p. 4.

Dr. Johnson testified that an important advantage of the Johnson model is that it has been created entirely within Quattro Pro spreadsheet software and can be readily operated, reviewed or studied by anyone using industry-standard software on a standard IBM-compatible personal computer. Dr. Johnson also testified that the Commission and the parties can review all of the input assumptions and calculations in a straightforward, efficient manner and can make changes to the assumptions, or use the model to study alternative scenarios, markets or situations. OCA Stmt. 1.0 (Johnson) at p. 2.

However, Dr. Johnson's "user-friendly" characterization was disputed by several parties who stated that the Johnson model required significantly more research and model intervention than does the BCM. The number of inputs apparently required by the user are many times that of the BCM. For example, Sprint/United stated that the Johnson model requires inputs by exchange and by competitor. Pennsylvania has 836 wire centers that are rate centers and many non-rate center host or remote switches which would put his input quantities into the thousands. Sprint/United Stmt. 2.0 (Dunbar) at p. 7. Witness Dunbar also pointed out that the Johnson model uses entries by exchange that include density, the number of residence loops, the number of single line business loops, the residence loop length, and the single line business loop length that would be up to the user to locate. Id at pp. 7-8. Additionally, Dr. Johnson conceded that certain inputs of the Johnson Model did not lend themselves to ready verification.

Accordingly, from the information contained in the record discussed above, it appears that the BCM 2 and Hatfield models are most conducive to verification and reproduction. Additionally, of the four models presented for our review, the BCM 2 and Hatfield also appear to be the most user-friendly of the models at this time.

b. Can the model be applied on a statewide basis? Is the model capable of generating results that are applicable to both large and small LECs?

The universal service fund should proceed under one method to calculate universal service costs for all recipients. OCA Main Brief at p. 22. Using different methods for various recipients could produce inequitable and discriminatory results in violation of the TA-96.

As OCA notes, most of the parties to this proceeding did not have access to portions of the Bellcore model including important assumptions and algorithms underlying it. Thus, it would be difficult to apply in the context of a statewide universal service proceeding. OCA Main Brief at p. 22. Bell, in fact, does not advocate the use of its model on a statewide basis. Indeed, Bell's evidentiary presentations in this proceeding through the use of its TSLRIC cost study have been exclusively confined to exchanges and calling areas within its own service territory. Thus, we have no concrete information before us in order to judge whether Bell's TSLRIC study and the underlying Bellcore cost models are readily adaptable for BUS cost calculations for other ILECs throughout Pennsylvania.

The BCM 2, Hatfield model and Johnson model can all be applied on a statewide basis. Indeed, the data submitted for each was done on a statewide basis. Witness Mercer testified that unlike other models put forth in this proceeding, the BCM and the Hatfield Model which incorporate the BCM, use non-proprietary methods and data inputs and they are applicable to all LECs in Pennsylvania. AT&T/MCI Stmt. 1.0 (Mercer) at p. 2.

Additionally, Dr. Johnson testified that the Johnson Model can successfully develop cost estimates for both incumbents and new entrants. Id. at p. 37. It appears that the Johnson Model is different from the Hatfield and BCM models in modeling the BUS costs of more than one provider in a particular market. This difference appears to be founded on the Johnson Model approach in simultaneously estimating the BUS costs of more than one provider in a given geographic area and market, based on the market shares of the respective providers. Certainly, different BUS providers in a specific geographic area and market, and with different market shares, may have different underlying estimates of BUS costs produced by the Johnson Model. Consequently, the ultimate levels of USF support for each of the respective providers' end-users may also be different. This is not necessarily inconsistent with Sprint/United's position that one of the advantages of its BCM model is that is would calculate the same support levels based upon the same forward-looking network for all carriers in the serving area.

Nonetheless, while all sponsors, with the exception of Bell, urge that their models are suitable for use on a statewide basis, we note that PTA and others had concerns with each of the models with regard to their utility for both large and small telephone companies. For instance, as discussed later in this Opinion and Order, PTA urges that the BCM misrepresents the annual carrying charges and discount levels which are generally found with small LEC operations. On the other hand, GTEN and others claim that the Hatfield model does not accurately depict the switching costs of small LECs and the distribution of households in small LEC service territories. Some of these concerns are more related to input variables than a fundamental flaw in design and thus would be more susceptible to correction.

Upon consideration of the above-listed factors, we find that the BCM and the Hatfield model most closely meet our expectations for statewide application and the ability to accurately reflect the costs of large and small carriers alike.



c. Does the model enable analysis on a highly disaggregated basis?

Generally, it costs less to provide telephone service to urban areas where end users are confined in a relatively small geographic area. High cost areas, areas where it costs more to provide telephone service, typically are those areas of the Commonwealth which are sparsely populated or are located in rough terrain. In order to identify those high cost service areas, we must disaggregate, i.e., divide the present service or study areas of ILECs into smaller sections or areas. The difficulty lies in choosing which level of disaggregation will produce the most accurate representation of BUS costs for universal service funding purposes.

Many of the experts in this proceeding support disaggregation down to either the wire center level or to the CBG level. See GTEN Stmt. 1.4 (Williams) pp. 5, 6; MCI Stmt. 1.0 (Bryant) Attachment B. The PTA maintains that disaggregation should be set at the wire center level. However, PTA asserts that Pennsylvania's small telephone companies have very few multi-wire center exchanges and thus the exchange level and the wire center level are virtually the same. PTA argues that because local rates are set for each exchange, BUS support also should be set for each exchange. PTA further argues that disaggregating down to the CBG level would be very complex and difficult to verify and would likely be expensive to administer.

In contrast, Dr. Johnson testified that while the Johnson model disaggregates down to approximately 375 Pennsylvania wire centers, in some instances this would represent too large a geographic area since circuit densities within a single wire center can vary substantially. Consequently, outside of the urban areas, Dr. Johnson states that it would be reasonable to disaggregate below the wire center level. Specifically, according to Dr. Johnson, use of entire wire centers in rural areas as the basis for subsidy payments could permit excessive payments to carriers that would only serve the town centers, while not providing sufficient funding to those carriers that serve the more sparsely populated outlying areas. OCA Stmt. 1.2 (Johnson) at p. 42.

The AT&T/MCI Hatfield model disaggregates down to the wire center level. The model results submitted by AT&T/MCI also aggregated the costs into the following six density zones: 1) the same six population density zones for Bell as they modeled in the July study; 2) the same six density zones for GTEN and United Telephone Company; and 3) the cities of Philadelphia, Pittsburgh, and Punxsutawney. AT&T/MCI Stmt. 1.0 (Mercer) at p. 23. Bell witness Sanford criticizes the use of only six density bands since it improperly masks the extent and nature of the existing subsidy flows. Bell Stmt. 1.2 (Sanford) p. 6. AT&T argues that "...the inclusion of the BCM database, which assigns each CBG in the United States to a specific LEC wire center, permits the Hatfield Model to be exercised at any level of disaggregation that the Commission may deem appropriate." AT&T Main Brief at 38 & n. 80, (citing AT&T/MCI Stmt. 1.1 (Mercer) at 8; Tr. 1402-1403.

On the other hand, Sprint/United, GTEN and PCTA advocate disaggregating to the CBG level. The CBG level actually disaggregates down to the units used by the United States Census Bureau. There are more than 11,000 of these CBGs in Pennsylvania. As noted earlier in this Opinion and Order, the CBG is defined by the Census Bureau in such a way as to on average contain 400 households. However, in actuality, a CBG may contain as few as a mere handful of households or as many as 10,000. GTEN Stmt. 8.1 (Cole) at p. 2.

Previously, we had stated that the service areas should be disaggregated at least down to the wire centers. While disaggregating down to the wire center gives a far clearer picture of the cost of service than existing standards which base costs upon an examination of the entire territory of the LEC, it is not without problems. Within wire centers, there is still a significant cost differential.

To determine the cost for the entire wire center, all the costs are averaged. Once averaged, this differential understates the cost associated with providing service to the high cost customers within the wire center. This "averaging" is not a problem so long as the same LEC serves everyone within the wire center. Although the LEC will receive less support from the Universal Service fund, theoretically, it would recoup the difference from the low cost consumers within the wire center. Once competition enters the local level, however, the guaranteed revenue stream may fade. Competitive LECs that will not be burdened by serving high cost areas may attract low cost customers within the wire centers by offering lower rates. As the low cost customers switch to the competitor, the incumbent LEC remains the carrier of last resort and is left with the high cost customers. While the incumbent LEC is eligible for support from the universal service fund, that support is artificially low due to the averaging within the wire center as a whole.

The clearest example of this averaging is provided in the following from GTEN's Reply Brief:

The least expensive CBG in the Berlin wire center has a monthly cost of providing universal service to a residence of $17.23; the most expensive CBG has a monthly cost of $73.79. Id. Within an individual exchange, such as Titusville, the monthly residential universal service cost ranges between $15.04 and $58.13. Id.
GTEN Reply Brief at p. 26.

Given the extreme cost variances within wire centers as just illustrated, we believe that disaggregation down to the CBG level will provide the most accurate representation of BUS costs. CGBs provide a much clearer, more exact estimate of what it costs to provide service. GTEN witness Williams provides a good summary of yet other important benefits of disaggregation down to the CBG level:

The cost variations within both large and small wire centers is most significant as I pointed out in my direct testimony. [footnote omitted]. The financial risk to COLRs is further increased, if the designated areas are too large, since these cost variations are normally present in the sections of a wire center where an experienced person would expect them to exist. As previously discussed, non-COLRs will be able to carefully select the areas they serve, by one means or another. There are additional reasons not to adopt existing ILEC wire center areas as the COLR geographic unit. First, existing ILEC wire centers will not necessarily be the same geographic areas that ALECs [CLECs] will select to serve, nor will all ALECs serve the same area. This results in CBGs having a number of advantages over ILEC wire center areas. First, readily available data bases already exist for CBGs. For example, information on demographics and terrain is maintained in such a way that it is readily available for any carrier to use. Such information is not readily available on an ILEC wire center basis. Second, CBGs are not directly reflective of any preexisting service areas, which makes them competitively neutral for all carriers. CBGs have the advantage of not requiring ALECs to "fit" the ILECs serving areas. Third, CBGs vary significantly in size, primarily according to density, and are generally much smaller than existing ILEC wire centers. These smaller areas would lower the entry barriers that would be created by requiring ALECs to serve larger areas in order to qualify for COLR support. Fourth, such areas lend themselves to "building blocks" which carriers may use in developing their service areas.
GTEN Stmt. 1.3 (Williams) at p. 18.

The primary criticism of disaggregating down to the CBG level is the administrative burdens of analyzing and working with 11,000 units rather than a much smaller number if wire centers or exchanges are used. For instance MCI witness Mercer stated that he didn't want to use CBGs mainly due to the administrative burden. GTEN Main Brief at p. 17 (citing Tr. pp. 1422, 1425). Similarly OCA contends that the "use of a CBG model allows for much more precise cost analysis, but would be overly burdensome and complex considering the large number of CBGs in Pennsylvania. Having utilized the BCM model ourselves, PUC staff does not find this criticism to be legitimate. Given the advanced computer capabilities available today, we do not believe that disaggregation down to the CBG level will present the administrative burdens alleged by some parties.

The other primary argument advanced against using the CBG level of disaggregation is contained in the following excerpt from the testimony of MCI witness Mercer:

It is that the per-CBG cost estimates are based on a great deal of equipment sharing between CBGs. CBGs share feeder routes in the BCM. Switching costs are determined for a switch sized to serve an entire wire center, not an individual CBG. Also, interoffice facilities are sized to serve an entire wire center, not a single CBG. Each CBG therefore benefits from the efficiencies that result from sharing transport facilities with the other CBGs in the wire center. Thus, I would argue, many aspects of an individual CBG's cost estimate are determined by the CBG's wire center context; the cost is certainly not a stand-alone cost. That being the case, I would argue that while universal subsidies might be set at a wire center level, or for a population density range to which a given CBG belongs, it does not appear appropriate to set a subsidy based on the cost estimate for any individual CBG.
AT&T/MCI Stmt. 1.1 (Mercer) at p. 10.

We do not believe that this concern is compelling enough to overcome all of the other benefits associated with CBG use for determining BUS costs. Additionally, we believe that given the continued significant advancement in model designs witnessed over just the last several months, the problems in measuring economies resulting from facilities sharing are probably not insurmountable. We also are of the understanding that some of the models already take this sharing into account such as the BCM which allows for the sharing of feeder cable between CBGs that are served by the same ILEC wire center.

Based upon this assessment criteria, the BCM's disaggregation down to the CBG level will produce the most accurate representation of BUS costs, and therefore, we find this model superior to the other models presented when viewed from this perspective.

d. Does the model use the best available technology and information? Does it utilize forward-looking technology and information?

In our September 5, 1996, Order, we specifically stated that:

As to the application of a TS-LRIC cost methodology, we accept the view that a well designed, forward looking model is a valid, straightforward approach to identifying the cost of basic universal service. More specifically, the Hatfield model appears to contain many of the elements necessary for utilization in the universal service context. We welcome all parties to develop and apply a forward-looking TS-LRIC methodology in the next phase of this universal service investigation.
Furthermore, we encourage record development of cost studies utilizing facilities of telephone systems using the wire center approach advocated by Bell, the proxy estimation approach advocated by GTE or any other reasonable methodology. In fact, we welcome broad record development of this issue and intend to closely evaluate the application of each alternative.
* * *
Furthermore, all cost studies and the justification for such studies shall clearly delineate between the following types of costs and should include specific, stand-alone justification for each category: 1) the TS-LRIC of basic universal service (without inclusion of joint, shared or common costs including loop costs); 2) joint or shared costs, if any, including loop costs; 3) common costs if any, including overhead; 4) carrier of last resort or community service costs, if any and 5) other categories of costs, if any. Participants who do not adequately disclose necessary information and justify each component of their study will run the risk that the cost study sponsored by the participant will be considered inadequately supported to warrant any consideration by the Commission.
Docket No. I-00940035, Order entered September 5, 1996, at 11-12, 22, emphasis added.

After careful review of the record, we reaffirm our earlier tentative findings in our September 5, 1996 Order that a well defined forward-looking TSLRIC proxy model provides the best basis to compute the cost of providing BUS in Pennsylvania. The proxy model generates a benchmark cost range for a defined set of basic residential services, assuming efficient engineering and design criteria, the deployment of state-of-the-art loop and switching technology, using current local exchange network topology.

Proxy models do not base BUS costs on the actual or embedded costs of any one provider, which is an important feature in a competitive marketplace. Proxies avoid the complex, company-specific calculation of embedded costs. AT&T Reply Brief at p. 37. Further, the amount of subsidy should not be different for different service providers, even if some service providers have lower costs than others, because the subsidy amount should be the same for those who are competing for the same customer. As Sprint/United points out, matching subsidy amounts to individual service providers' costs would favor inefficient providers. Sprint/United Stmt. 1.0 (Jamison) p. 9. Another benefit of the proxy model is that costs can be calculated for all companies using the same assumptions, algorithms, etc. Sprint/United Stmt. 1.0 (Jamison) p. 13.

Universal service subsidies should supplant uneconomic, non-competitive embedded subsidies, be specifically targeted, be competitively neutral, and be portable so that the company receiving the subsidy is also the company that incurs the cost of providing the service. Sprint/United Main Brief, p. 16. The proxy model accomplishes all of these goals.

A TSLRIC cost model that builds a proxy telecommunications network for the purpose of studying the prospective TSLRIC economic costs is intended to maximize the looking forward efficiencies of a newly constructed telecommunications network, assuming existing wire centers and efficient technology in the distribution and feeder network. Thus, this new network in a proxy TSLRIC cost model is intended to minimize the associated TSLRIC costs and may not rely on existing network configurations and architectures or on the embedded network plant parameters and associated embedded costs. Instead, such a proxy TSLRIC model will incorporate looking forward engineering simulations that will maximize the efficiencies derived from the deployment of new telecommunications technologies and network configurations and topologies for the provision of service(s), within a specific geographic region, and meeting a particular level of demand. TSLRIC proxy cost models offer the advantage of estimating the future economic costs for providing BUS, rather than being dependent on past embedded network costs. Thus, a competitive carrier, whether a CLEC or an IXC desiring to enter the local exchange telecommunications services marketplace at a particular locale, may be able to ascertain the relevant economic costs of such market entry and assess whether its provision of BUS will be in need of support from a Universal Service funding mechanism in a competitively neutral manner.

Sprint/United also notes that proxy costs are consistent with price regulation and are consistent with competition while company-specific costs are not. Sprint/United Stmt. 1.0 (Jamison) p. 13.

While most of the sponsors in the proceeding profess that their models are based upon forward-looking TSLRIC studies, one of the primary criticisms of the Bell model, as already discussed, is its use of "embedded" costs to determine BUS costs. OCA maintains that:

If one incumbent LEC is able to achieve full cost recovery from all other carriers based upon its actual costs, such a universal service fund system would not be competitively neutral. (Footnote omitted.) The system that Bell proposes would require other carriers to support Bell's high costs even if other carriers are able to offer service at a much lower cost. Thus, Bell's use of actual costs reflects a disadvantage of the Bell approach.
OCA Main Brief at p. 7.

We agree with AT&T/MCI witness Mercer that for purposes of universal funding determinations, the Bell approach, while capturing the flavor of a TSLRIC, may omit the key aspect of a forward-looking approach, in which the best available technology is applied to the provision of the services under study, rather than an embedded approach. AT&T/MCI Stmt. 1.0 (Mercer) at p. 3. For instance, Bell uses loop investments based on the existing investment in loop plant using the Vintage Retirement Unit Cost system. Bell uses outputs from the Trunk Inventory Record Keeping System to obtain information on the amount of interoffice facilities in use. AT&T/MCI Stmt. 1.0 (Mercer) at p. 3.

Although Bell's looking forward TSLRIC approach may attain certain optimization efficiencies, e.g., the incorporation of new telecommunications technologies in a given network will almost always render the network more efficient. On the other hand, the extensive reliance on existing network configurations and embedded plant parameters may retard the level of all possible operational efficiencies that could be attained under Bell's TSLRIC modeling approach. Such an approach may tend to produce cost estimates for BUS that are closer to a telephone utility's existing and embedded costs. In turn, such embedded cost figures can be more easily associated with the utility's historical and existing revenue streams from its various services. Thus, the telephone utility in question will possibly be in a better position to identify the net revenue effects from the receipts and/or disbursements to and from a Universal Service funding mechanism.

Other models are also criticized for not being pure TSLRIC models in certain respects. The BCM is criticized for producing average costs, rather than incremental costs. See, OCA Main Brief at p. 7. According to OCA, regardless of the claimed advantages of the BCM, the model produces cost results for providing universal service to residential customers that are overstated. OCA Main Brief at p. 7. The Hatfield Model is also criticized for not being "forward-looking" because it uses historical expense factors and embedded cost factors, which is discussed in more detail below.

Nonetheless, when considered under this criterion, we find that while all of the models are premised on being a forward looking TSLRIC approach, the Hatfield, BCM 2 and Johnson models most closely meet our expectations with regard to being a forward looking TSLRIC proxy methodology at this time.

e. Is the model readily adaptable to changes in the definition of basic universal service?

It appears that all of the models introduced in this proceeding can estimate the cost of providing basic service and subsequent changes to the definition. Nonetheless, we note GTE's concern that the BCM 1 results did not include several services included in the Commission's definition of universal service including EAS and a white pages directory listing. It would appear that this is more of an "input" problem than a "design" problem and is correctable.

We also note that these concerns appear to have been rectified in the BCM 2 model.

f. Does the model design accurately reflect costs of putting in place the loop and switch necessary to provide BUS?

All of the models presented in this proceeding can be improved upon or refined in certain respects. However, based upon the following discussion, we believe that the BCM 2 most accurately measures the cost of putting in place the loop and switch necessary to provide universal service.

Bell's model does not reflect forward-looking costs to the degree contemplated and requested by this Commission in its September 5, 1996 Order. Rather, it appears to rely to a great extent on historical costs in determining BUS cost within its service territory. For instance, as noted by MCI Witness Mercer, both Bell's loop and interoffice facilities reflect historical costs, rather than the forward-looking costs of providing these services.

The revisions to the Hatfield Model resulted in some improvements to the Model, however, the Model still contains the flaws associated with BCM 1 which it incorporates and could use improvement in that regard. All in all, we have identified the following areas of improvement for the Hatfield model:

1) The Hatfield Model as updated in the record of this proceeding incorporates the BCM 1 and thus contains all of its flaws, some of which have now been addressed in BCM 2.
2) Like the BCM which it incorporates, the Hatfield Model may overstate the price of cable resulting in a near doubling of the cost of installation. GTEN witness Cole argues also that while incorporation of the BCM greatly improves the utility of the Model, several changes are a step backwards. For instance, the Hatfield treatment of the switching module is a step backward from the BCM's and is primitive by comparison with RAND's switch module. GTEN Stmt. 8.2 (Cole) at p. 6.
3) The Hatfield Model also utilizes the BCM's low line fill factors -- 25% in areas of low population density -- which leads to higher costs due to the excess amount of outside plant required to serve a given number of customers. (The fill factor is a measure of how much of the plant is being used to provide service to customers.)
4) The Hatfield Model appears to incorrectly estimate switching costs by grossly underestimating the cost of small central offices. GTEN witness Cole also states that the Hatfield Model large office estimate is unverifiable. GTEN Stmt. 8.2 (Cole) p. 9.
5) There is a question as to whether certain engineering and cable splicing costs are included. The Model still appears to exclude costs for, inter alia, tandem switching, signalling, interoffice transmission, the protector, underground conduit in dense areas and a directory. See GTEN Stmt. 8.1 (Cole) at p. 4. The Model still appears to underestimate the material and installation costs associated with the distribution network in urban areas.
6) The Model relies on historical data for the development of certain expense factors. For instance, the sponsors concede that they did not have studies or data to estimate operating expenses and therefore, they relied primarily on historical data for Bell as recorded in the 1994 ARMIS 43-03 Reports. Historical information was used to determine cable and wire expenses, network operations, etc. Bell witness Emmerson notes that the Model changes the BCM by extracting intermediate results from the BCM to break out categories of loop investments for purposes of applying expense factors based on Armis reports. This is based on historical accounting records. Bell Stmt. 3.2 (Emmerson) at p. 14.
7) The Model does not provide estimates sufficiently accurate for use by small companies because of the wide variation from actual experienced material costs. GTEN witness Cole testified that Dr. Mercer's significantly lower DLC costs appear to be deep discounts from the list prices, discounts that are based on conversations with experts and are nonverifiable. GTEN Stmt. 8.1 (Cole) at p. 10. Witness Cole goes on to state that a by-product of using the deeply discounted material prices for cable and for DLC is an equally deep reduction in structure costs, given the way the BCM works. GTEN Cole claims that because structure costs don't vary at all with materials prices, this is a built-in flaw in the BCM which helps produce the low numbers in Dr. Mercer's tables. GTEN Stmt. 8.1 (Cole) at p. 10.
8) The Model disaggregates down to the wire center level which is not as fine a delineation of high cost areas as CBGs provide.
9) The Model's overhead loading factor of six percent is based on what may be an inappropriate comparison with firms in the auto, manufacturing and airline industries. We are inclined to agree with GTEN Witness Cole who claims that the six percent has absolutely no relevance to the overhead costs of a LEC, at least at this time. GTEN Stmt. 8.1 (Cole) at 13.
10) The sponsors concede that the BCM does not provide the information necessary to determine the location of tandem switches or the way in which the various wire centers are interconnected. Witness Mercer stated: "While the prior version of the Hatfield Model attempted to locate a tandem in the center of a given aggregated study area, and assumed all switches were interconnected via a tandem switch, this is not possible in the BCM-provided wire center disaggregation." AT&T/MCI Stmt. 1.0 (Mercer) at p. 15. Thus Mercer conceded that a new model of interoffice facilities is required.
11) AT&T also concedes that refinements are necessary to the switch algorithm to permit separate identification of fixed and usage-sensitive switching costs. Witness Mercer stated that while they believe the switch cost algorithm they use reflects usage costs, it does so essentially by considering the total cost of the processors that are necessary to support the normal usage levels associated with the number of lines served by the switch in question, rather than by specifically identifying fixed and usage-sensitive cost components. Mercer testified that the sponsors were considering modifications to its switch cost algorithm that would make such a separate identification. AT&T/MCI Stmt. 1.1 (Mercer) at p. 14.

Finally, all-in-all, we are concerned with the results of the Hatfield Model which produce some of the lowest BUS costs. This leads GTEN to believe that the Hatfield Model "should not be considered a minor modification to the BCM, but rather a wholesale alteration with the effect of producing low universal service funding." GTEN Main Brief at 18, footnotes omitted. For instance, GTEN states that the deficiencies in the Hatfield Model produce drastically different results than does the GTEN version of the BCM. GTEN's results, without overheads, average $90.96 while the Hatfield Model, with overheads, averages $39.43. GTEN Initial Brief at p. 18. GTE also states that loop investment differences are dramatic -- in the lowest density areas, the Hatfield loop investment was approximately 41 percent of the BCM's figures. Id.

Although the most widely endorsed of the models presented for our review, the BCM is also the subject of some criticism. GTEN regards the BCM as the leading proxy model because it uses a forward-looking cost methodology for allocating universal service funding and focuses on the CBG. GTEN Stmt. 8.1 (Cole) at p. 5. GTEN, however, raises concerns regarding the BCM's current treatment of switching cost. GTEN states that only one switch architecture is examined, the DMS-100. Additionally, the switch data are from 1986 at the latest and the raw data on switch costs may have been transformed by a consultant for New England Telephone Company in ways that cannot be reconstructed at this point. GTEN Stmt. 8.1 (Cole) at p. 9.

Additionally there appear to be costs omitted entirely from the BCM which are pertinent to the provision of universal service, including white pages directory listings. GTEN Stmt. 8.1 (Cole) at p. 4. GTEN identifies the following additional omitted costs: usage, EAS, and a directory. GTEN does not consider these omissions to constitute a deficiency in the model itself because the model was not built to estimate precisely the cost of providing universal service. Rather, the model was built to identify high cost serving areas and the factors that cause them to be so. GTEN Stmt. 8.1 (Cole) at p. 8. Additionally, we note that the BCM-2 appears to already correct for some of these omissions.

The drop, protector and main distribution frame investments were omitted, and growth rates in access line demand are not taken into account which makes it a little less forward looking. GTEN indicates that the sponsors appear to have now added these costs. The latter factor affects the cable size choice and as a result the long-run costs of providing service. GTEN Stmt. 8.1 (Cole) at p. 8.

Bell also noted that the BCM failed to use any actual data for engineering and planning feeder routes or distribution facilities and ignored the investment and expenses associated with certain loop components such as serving area interfaces, the terminal and the drop. By excluding these, loop costs are understated. Bell Stmt. 3.2 (Emmerson) p. 8.

Initially, the uniform population distribution assumption of the Model was problematical for territorially large and population sparse rural CBGs. The Model built too much loop plant in rural areas and likely overestimated the cost of providing service there. GTEN Stmt. 8.1 (Cole) at p. 11. At present the rural serving areas are being bounded by the model developers to be within a half mile on either side of a road, which should considerably reduce rural loop cost estimates. Those cost reductions, however, could be offset in some rural CBGs when and if the Model builders succeed in incorporating terrain slope information into the Model to explain high costs in hilly/mountainous areas. GTEN Stmt. 8.1 (Cole) at p. 8.

GTEN notes that with all of the changes planned to be made to the BCM 1, there is one significant omission, the de-linkage of structure placement costs from cable materials prices. GTEN claims that what a LEC pays a vendor for copper cable has no bearing on how much it costs a construction crew to put it in the air or in the ground. GTEN Stmt. 8.2 (Cole) at p. 5.

Bell also notes that the BCM misrepresents annual expenses since the factors make operating and annual capital expenses dependent on the wrong cost drivers and invariable with respect to the type of plant used. Bell Stmt. 3.2 (Emmerson) p. 9.

Nonetheless, the BCM 1 appears to produce results that are neither unreasonably high or unreasonably low. Additionally, many of the problems discussed above have now been corrected in the updated BCM 2.

Based upon the testimony of the various parties to this proceeding, the Johnson Model could also be improved upon in certain respects. For instance Sprint/United witness Dunbar notes that several of Dr. Johnson's inputs impute similar characteristics on all wire centers in the state. Sprint/United Stmt. 2.1 (Dunbar) at p. 8. Witness Dunbar testified in this regard:

For example, he has a single input that assigns the same percentage of wire center loops to zone 1 for all wire centers. A single input assigns the same percentage of business loops to zone 1 in all wire centers. In his model, all business loops in a wire center have the same length. All residence loops in a wire center have the same length. These assumptions are not at all realistic. Neither are the assumptions that a wire center area is square and has only two zones.
Id. at p. 8 (emphasis in original).

It also appears that the Johnson Model's reliance upon a square wire center of two zones is not reasonable when matched to actual customer location data. For example, Sprint/United notes that Pennsylvania has 11,688 CBGs spread across 800+ wire centers. There are as few as one CBG in a single feeder route and as many as 54. The loop lengths across the CBGs vary greatly. The Johnson Model does not sufficiently recognize these differences. Sprint/United Stmt. 2.1 (Dunbar) p. 9.

Moreover, the Model does not appear to distinguish according to terrain difficulty per wire center. Given that this is a significant factor influencing the cost of the loop, the Model could produce unreasonable results. Witness Dunbar notes that there are 20,025 unique terrain areas in Pennsylvania that are spread across the 800+ wire centers with each terrain area having its own cost of plant. Sprint/United Stmt. 2.1 (Dunbar) p. 9.

Bell points out that its tests of the Model using residence specific loop lengths produced illogical results. Bell witness Sanford testified that 70% of Bell's wire centers have longer residence loops than the average loop lengths that Dr. Johnson computed in his Model. Bell Stmt. 1.2 (Sanford) p. 8.

Overall, the results of the Johnson Model appear somewhat high compared to the other models studied. As revised, the BCM 2 appears to capture the cost of putting in place the loop and switch necessary to provide BUS better than the other models at this time.

g. Has the model been considered at the Federal level for Federal funding purposes? Is there evidence of wider use of the model?

Neither the Bell Model or the Johnson Model were the subject of scrutiny at the Federal level. However, it should be noted that like the Bell Model, the Cost Proxy Model submitted by PacTel in the Federal universal service proceeding, also relied upon proprietary information to a significant degree. This was identified by the Joint Board as a major concern or disadvantage with the model's use in the future.

Both the Hatfield, the Hatfield/BCM Model and the BCM 1 and BCM 2 Models were the subject of in-depth scrutiny at the Federal level. While the Joint Board stopped short of endorsing any specific proxy model, it did determine that a properly crafted proxy model should be used to calculate the forward-looking economic costs for specific geographic areas. The Joint Board recommended that the FCC continue to work with state commissions to examine the various models and develop an adequate proxy model that could be used to determine the cost of providing supported services. Additionally, the Joint Board recommended that a proxy model be developed for adoption by the FCC by May 8, 1997, the deadline for action under the Federal Act.

While the Joint Board recommended the use of embedded costs for rural telephone companies as defined by the TA-96 for a three-year period with a phase-in to the proxy model methodology over a corresponding timeframe, the Joint Board's decision in this regard was for purposes of the Federal funding mechanism. We do not believe that the Joint Board decision in any way precludes the use of proxies by state commissions for rural telephone companies for the purpose of determining state funding levels. Additionally, with the extra precautions and considerations given to small telephone companies herein, we believe that any potential adverse effects in using proxies for small telephone companies will be mitigated. In the event small LECs believe that they are adversely affected by the use of proxies, we make available today a small LEC waiver process described in more detail in Section XI. infra.

h. Overall Assessment

Overall, at this time we tentatively find that the BCM 2 most closely meets our expectations with regard to a suitable methodology for determining BUS costs in Pennsylvania. Nonetheless, like the Joint Board, we stop short of giving our complete approval to this methodology absent consideration in the following months of further improvements to all models and of the FCC's ultimate decision at the Federal level. It is our intent to incorporate any additional improvements and/or refinements subsequently made into the BCM 2 or other model we ultimately settle upon. Additionally, since it will be some time yet before the Pennsylvania universal service funding mechanism is in place, we also do not desire to totally preclude consideration of yet other model parameters that may prove to be a more accurate model for measuring costs.

For instance, GTEN witness Cole referred to the Cost Proxy Model ("CPM") sponsored by PacTel, recently endorsed by the California Public Utilities Commission to estimate proxy costs in California. GTEN explains that it is a loop plant model built on the individual customer level of aggregation. GTEN Stmt. 8.2 (Cole) p. 7. GTEN witness Cole also testified that Bellcore was working on a revised model. GTEN Stmt. 8.2 (Cole) p. 8. We also note that the sponsors of the Hatfield model apparently have a newly refined model, the Hatfield 2.2.

We also take note of the position of some parties that the models evaluated in this proceeding be further developed and evaluated before a specific model is adopted by the Commission. These parties recommend that the Commission not immediately adopt any particular model, but instead, allow time for these, and other models which are just now being developed, to be properly evaluated and tested. GTEN Stmt. 1.3 (Williams) at p. 20. Some parties have suggested that "...the Commission should conduct workshops aimed at reaching consensus on a proper model or models for use in determining the cost for providing Basic Universal Service." Bell Main Brief at 36.(16)

In acting today and tentatively endorsing the BCM 2 model, we give parties a clearer indication of the type of model we believe would give the best picture of the costs of providing BUS in Pennsylvania. So that we can take full advantage of further advancements in the models prior to implementation of the state funding mechanism, we direct the Universal Service Task Force we establish today to continue to evaluate the various models as ultimately revised or refined and the ultimate findings of the FCC for federal funding purposes, and to apprise the Commission of these further developments and its recommendations. The Task Force shall report back to the Commission with its findings on this issue within six months from the entry date of this Opinion and Order.

E. Revisions to BCM

1. Annual Carrying Charges

The Hatfield Model used an annual carrying charge factor of 22.9%(17). This compares to the annual carrying charge factor of 31.67% utilizing ARMIS data. The BCM Model's 31.67% annual charge factor was based upon ARMIS data. Since we believe that the use of non-ARMIS data is more appropriate, we have chosen to use the Hatfield Model's annual charge factor of 22.9% for purposes of computing BUS costs for the larger Pennsylvania telephone operating companies.

PTA argues that the BCM which utilizes a 31.6765% annual carrying charge factor for cost development derived from ARMIS data, is not representative for small companies. According to PTA, the annual carrying charges for PTA companies range from 30.28% to 71.05% with an overall weighted average of 38.25%.

We agree with the PTA that certain inputs, including the annual carrying charge factors for small telephone companies, those with under 50,000 access lines, should be included when calculating BUS costs under the BCM 2 Model. We find that while, in general, the range of annual carrying charge factors given by the PTA appear to be excessive, we believe that it would be appropriate to use a higher weighted average annual factor ratioed down to be consistent with the non-Armis data used for the larger telephone companies to reflect the higher risk factor for the smaller telephone companies, using non-ARMIS data. This change reflects a 28% annual carrying factor for the smaller companies (those with under 50,000 access lines) and has been incorporated into the model.

2. Discount Rates for Small Telephone Companies

PTA also argues that the BCM accepts only one discount rate for equipment purchases which puts smaller companies at a disadvantage since larger companies receive significantly larger discounts. Once again, we agree with the PTA that to the extent the model can differentiate between the discount rates applicable to small and large telephone companies, it should do so.

F. Joint and Common Costs

1. Allocation of the Local Loop

a. Position of the Parties

We now address one of the most contentious issues in this proceeding, whether the local loop is a "joint" cost and should be allocated among the services that use it, or whether the entire loop is a direct cost of providing BUS, and should be allocated to BUS in its entirety. The principal costs associated with providing BUS are the subscriber loop costs. GTEN Main Brief at pps. 4-5 (citing Panzar Direct at 10). According to TCG witness Gabel, loop costs generally account for from 30 to 50 percent of the total invested plant of a telephone utility. TCG Stmt. 2.0 (Gabel) p. 10. In our September 5, 1995 Order in this proceeding, we tentatively concluded that the loop should be classified as a joint cost subject to allocation among the services which use it.

About the only matter the parties agreed upon was the definition of joint and common costs. A joint cost is:

a specific type of common cost--one incurred when production processes yield two or more outputs in fixed proportion.

Bell St. 2.0 at p. 12. Likewise, Bell witness Emmerson defined a shared cost as:

...a cost which is incurred by two or more services. As such, "shared cost" is an incremental cost to the group of (two or more) services but is not an incremental cost of any one service in the group. Since shared costs do not vary as the output of individual services vary, they are not incremental to any one service.
Bell Stmt. 3.0 (Emmerson) at p. 8.

The ILECs, including PTA, Bell, GTEN, and Sprint/United, along with AT&T and OSBA argue that the local loop is a direct cost of BUS and should not be allocated among different services. These parties rely primarily upon the testimony of Bell witness Kahn and GTEN witness Panzar both of whom argue that it is the act of subscribing to the network which causes loop costs to be incurred. Dr. Kahn states that the only economically correct method of measuring cost is to follow the rule of cost causation. Bell St. No. 2.1 (Kahn) at 2.

Dr. Kahn testified that:

Consumers impose the cost of the loop on a telephone company and on society by the act of subscribing to telephone service. The causation principle therefore requires that the cost of providing the loop be fully incorporated in the cost of that basic service--which the Commission Order defines as "single party, voice grade, incoming and outgoing access to the public switched network." (August 31 Order at 4). Conversely, if, as I understand to be essentially the case, actual use of the loop for local or long distance calling or for other services imposes no loop costs on the supplier and if subscribers were to refrain from placing those calls or using any of those other services it would not save any of those costs, there is no sense in which usage or other services can be held causally responsible for loop costs.
Bell Stmt. 2.0 (Kahn) p. 3.

Dr. Kahn testified that the critical substantive economic criteria for determining whether the loop is a shared or joint cost are:

- Does the dialtone or access to the telephone company switch that the loop provides a subscriber have its own costs, defined causally, separate from the costs imposed by usage and the other services? The answer is unquestionably yes.
- Do the other services have their own separable costs, separable from one another and separable from the costs of the loop? The way to answer that is to ask what would happen to the total costs of the company if subscribers used a finite additional amount of each of those services or if they ceased using each or all of them entirely (these two measures corresponding to LRIC and TSLRIC of those several services). If the costs of the loop would remain unchanged in either of those two instances, the LRIC or TSLRIC of the other services would include none of the costs of the loop.
- Finally, are subscriber access and each of the other services demanded or supplied in fixed proportions? [footnote omitted]. The answer is of course no. Some subscribers to telephone service obtain the loop merely in order to receive calls, along with the option of placing calls--including 911--when and if they wish to do so; others to place many local but few long-distance calls; still others to do the reverse; and there are similar variations in the demands for the other services.

Using Dr. Kahn's causation principle, since actual use of the loop for local or long distance calling or for other services imposes no loop costs on the supplier and the supplier does not save any of the loop costs if its subscribers do not use any long distance or other services, usage or other services cannot be held causally responsible for loop costs. Bell Main Brief at p. 23. Bell states that there is no sound economic basis for splitting the cost of the loop between BUS and other services, but, in order to comply with our September 5, 1996, Order directives, Bell carried out its TSLRIC calculations with 90% of the loop costs allocated to BUS. Id. at 12.

Similarly, GTEN also urges that the making of toll calls does not impose greater loop costs on the supplier of loops, so it makes no sense to allocate loop costs to toll calling. GTEN Main Brief at p. 7. Loop costs should be attributed to network access since this is the cost causing factor. GTEN Main Brief at p. 3. GTEN witness Panzar argues that those parties advocating an arbitrary allocation of loop costs to other services are mistaken because they ignore the fact that other services, such as toll calling, do not cause loop investment. GTEN Initial Brief at p. 7.

GTEN witness Panzar also testified that the argument that loop costs are not part of the incremental cost of universal service, because they must be incurred in order to provide interexchange access and intraLATA toll even if basic service is eliminated, confuses "cost measurement" with "cost recovery." Witness Panzar states that the correct measure of LEC costs of implementing the Pennsylvania universal service plan requires use of the entire incremental loop costs in high cost areas. How those costs are recovered is a pricing issue. GTEN Stmt. 3.1 (Panzar) at p. 11. Dr. Panzar also testified that by confusing the resource costs associated with serving various groups of high cost customers with the full cost of implementing a universal service plan, witness Johnson seriously understates the costs of providing universal service. The foregone revenue resulting from below-cost pricing of basic service in high cost areas is also part of the LEC's cost of implementing a universal service plan. GTEN Stmt. 3.2 (Panzar) at p. 2 Finally, Dr. Panzar testified that "[a] Universal Service costing methodology for the coming competitive era should be designed to be compatible with the maximum possible amount of competition. It must not offset costs by the amount of implicit or explicit subsidies and contributions flowing from other services, or from low cost regions to high cost regions." GTEN Stmt. 3.2 (Panzar) at p. 6.

GTEN witness Williams argued that in the event the Commission were to adopt a loop allocation methodology, there is no assurance that sufficient intraLATA toll revenues would be generated from within the geographic area in question to cover its allocated share of loop costs. Id. at p. 29. Further, GTEN states that arguments that costs should be allocated among various services is flatly inconsistent with the current market in which CLECs are requesting the sale of unbundled loops. GTEN states that the fact that CLECs are purchasing unbundled loops contradicts the claim that network access is not a service in its own right. Purchase of an unbundled loop provides precisely the service -- access to the network -- which is the cost causative factor for local loops. GTEN Initial Brief at p 6.

Finally, while GTEN supports 100% assignment of the loop to BUS, it states that if an allocation is to be made, it would be most appropriate to base that allocation on SLU.(18) While GTEN witness Williams believed it to be impossible practically speaking to perform a loop allocation based on SLU for all the numerous services that use the loop, he did perform a loop allocation based on SLU for local exchange service, intraLATA toll, and both intrastate and interstate, interLATA access. GTEN Reply Brief at p. 13. Using SLU for allocating the loop produces an allocation to intrastate service of approximately 87% in GTEN's service area.

Sprint/United agrees with Bell and GTEN that loop costs are a direct cost of providing BUS, and therefore, should not be allocated to other services as a shared cost. Sprint/United Main Brief at p. 41. Sprint/United argues that it is wrong to equate "joint use" with "joint or shared cost". It argues that since an increase in the demand for basic network access directly increases the necessary investment in the local loop, it is obviously a direct cost to basic network access; it cannot be a shared cost. Sprint/United Main Brief, p. 42. Sprint/United argues that the OCA's 50% allocation is arbitrary, not supported by the evidence, and seriously understates the cost of universal service. Sprint/United Main Brief at p. 45.

Ultimately, PTA believes that the "issue of loop allocation is a non-issue, the argument over which is very unconstructive and wasteful." PTA Main Brief at p. 6. PTA states that the issue becomes largely irrelevant when local service rates are capped at the affordability benchmark. Id. at p. 6. PTA states that even if the entire loop is allocated to universal service, rates are capped at a level less than total cost. Like GTEN, PTA argues that costing is entirely different from cost recovery, and that under the PTA's proposal, LECs will continue to recover the cost of the loop from services other than BUS, to the extent that they can. The rest will be recovered from local service and the universal service fund. PTA Main Brief at p. 7.

AT&T also supports 100% allocation of the loop to BUS. AT&T Stmt. 1.0 (Darrah III) p. 35. Witness Darrah testified that all loop costs are properly assignable to the costs of local exchange services. AT&T does not currently have a specific proposal for the Commission on an alternative method for assigning those costs. Nevertheless, states AT&T, if the Commission does determine to assign some portion of the loop, that assignment should begin by removing the 25 percent of loop costs that are assigned to interstate services by the FCC separations process. If these costs are removed from the cost of the loop, then the Federal SLC and interstate CCLC would not be included in the revenue analysis. AT&T Stmt. 1.0 (Darrah III) at p. 36.

On the other hand, those parties favoring some allocation of the loop include the OCA, MFS, PCTA, OTS and Teleport. OCA witness Johnson testified that the loop is a joint cost required for the provision of at least three different services: local exchange service, intrastate long-distance service and interstate long-distance service. OCA witness Johnson testified that Bell's allocation is "purely arbitrary" and would overcompensate the LECs since they would undoubtedly continue to generate large sums from other services. OCA Stmt. 1.0 (Johnson).

The OCA's expert witness, Dr. Johnson proposed a 50% allocation of loop costs to BUS. OCA Stmt. 1.0 (Johnson) at 17-18; OCA Main Brief at p. 5. OCA argues that universal service should only have allocated to it a 50% portion of loop costs because other services will use the same loop and the serving LEC will be benefited by that amount. OCA Main Brief at pp. 14-15. "This is sensible, since recovery of loop costs must be a shared responsibility of all services using the loop". OCA Stmt. 1.0 (Johnson) at p. 17. Additionally, OCA states that the 50% allocation more closely approximates the share of the revenue burden historically borne by basic services than do any of the other proposed loop allocators, which OCA witness Johnson states amounts to approximately 55% of the total revenues generated by services directly using the local loop. OCA states that the Subscriber Line Charge, Dial-tone Line Rate, local usage and TouchTone is roughly 50% of the revenues produced by Bell's customers for telephone local exchange service. OCA Stmt. 1.0 (Johnson) at 18; OCA Main Brief at 5. OCA Stmt. 1.0 (Johnson) at p. 18. The OCA also argues that the 50% allocator is consistent with recent actions by the FCC and the Washington Utilities and Transportation Commission.(19) OCA notes that the FCC proposed an allocation method which would have allocated 25% of the loop to interstate toll, 25% to intrastate toll, and 50% to local service, and that at least one other state commission has adopted a 50% allocation factor.(20) Like TCG and OCA, OTS supports a 50% allocation of loop costs.

Alternatively, OCA witness Johnson testified that the Commission could use other more sophisticated allocation approaches including revenue-based methods, usage-based methods and direct cost-based methods. Id. at 20. For instance, the Commission could use usage-based allocation factors including SLU discussed above or the Subscriber Plant Factor ("SPF").(21) For most utilities, a SPF allocation to local would be somewhere in the range of 50-70% depending upon the relative toll and local usage levels in the carrier's service area. OCA Stmt. 1.0 (Johnson) at p. 18. OCA witness Johnson states that he chose the 50% allocation factor because it was a relatively simple and stable allocation approach. Id. at 20.

Teleport, MFS and PCTA support the OCA arguing that the loop is a joint and common plant cost element which is essential not only for local exchange service but for the provision of intraLATA toll, interstate toll and the provision of custom calling services. OCA Stmt. 1.2 (Johnson) at 24; Teleport Main Brief at p. 21. TCG witness Gabel and PCTA supports a 50 percent allocation to basic local exchange service. TCG Stmt. 2.0 (Gabel) p. 10; PCTA Main Brief at p. 15.Teleport Main Brief at p. 21. MFS Witness Montgomery argues that by shifting their "embedded" revenue requirement of universal service, Bell and others "increase their latitude for lower pricing of potentially competitive services in the future." MFS Stmt. 2.0 (Montgomery) p. 5. PCTA agrees that the result of Bell's 90 percent allocation is to "maximize" the size of the fund and cross-subsidize Bell's competitive services. PCTA Main Brief, p. 15

b. Discussion

We reaffirm our findings in our September 5, 1995 Order at Docket No. L-00950105 that the local loop is a "joint cost", not a direct cost of providing only those services included in the definition of BUS. It is used for a variety of services other than BUS and must be allocated among the services which utilize it. For universal service funding purposes, not allocating a portion of the local loop to all the services which utilize it fails to give recognition to the fact that the loop is used to provide many services in addition to BUS.

This finding is consistent with our earlier rulings including Pennsylvania Public Utility Commission v. Breezewood Telephone Company, 74 Pa P.U.C. 431 (1991) wherein we stated:

...[W]e consider the costs associated with the loop from the central office to the customers premises a non-traffic sensitive joint cost.
* * * *
AT&T states that the Recommended Decision is not clear on whether NTS costs are joint costs of providing local and toll services. It asserts that our Final Order should declare that dial-tone line costs are not "joint costs" of various services, but instead are the costs of establishing the physical connection between each customer's premises and the Company's central office.
There is no dispute that both the local customer and AT&T make use of the same local network to complete both local and interLATA calls. If it were not for the existence of the local network, AT&T would be required to construct at considerable expense an alternative means of access to the local customer. We find that CCLC is the cost of compensating BTC for the use of the common line, and as such, CCLC clearly pays for a service received by AT&T. Thus, dial tone line costs are joint costs.

We reject the ILECs' arguments that the local loop is not a joint cost because other services which use the loop do not result in any additional cost. We do not find the arguments of Bell's expert witness Dr. Kahn persuasive on this point. In particular, we do not accept the basis of Dr. Kahn's argument that because the loop is needed for local service and the incremental cost of the loop does not increase to provide other services, that its full cost must be attributed to local service. This same argument could be made with respect to toll service. Since the loop is necessary to provide toll service, it could at the same time be argued that the full cost should be allocated to toll, and in so doing the incremental cost to provide local service would be zero. Moreover, since the installation of an additional subscriber loop increases the capacity available for placing and receiving all three types of calls, the telephone company cannot increase the capacity for local calls without concurrently increasing the capacity for toll calls. OCA Stmt. 1.0 (Johnson) at p. 11. We find persuasive the arguments of Dr. Johnson that the local loop clearly fits within the definition of a joint cost since access capacity is simultaneously expanded for multiple services in fixed proportions (one more line is available in each case). Only if there is congestion at a particular time is there any tradeoff between use of the local loop for different purposes. Id. at 11.

We also reject the argument that because BUS essentially provides for network access, it is appropriate to allocate 100% of the loop costs to BUS. The "access" that is provided is not only to provide BUS but to give subscribers access to a myriad of services other than BUS as well. Without that access, customers would be unable to obtain toll service or enhanced services. It is improper to ascribe the "access" element solely to BUS when that access is used to obtain and use a host of other non-BUS services. Sprint/United arguments that equating "joint use" with "joint or shared cost" is inappropriate are equally unpersuasive. A joint cost is by definition one which is incurred through the provision of multiple services that utilize the facility at issue.

Having decided that the loop is a joint cost, rather than a direct cost, and should be allocated to the services that use it, we must now determine the appropriate percentage of the loop to allocate to BUS. OCA argues that we should allocate 50% of the loop to BUS. OCA bases its 50% estimate in large part upon current revenue breakdowns and administrative simplicity concerns. Bell argues that 90% of the loop should be allocated to BUS, if it is determined to be a joint cost. Others, argue that if an allocation does occur, it would be most reasonable to base the allocation upon SLU. OCA Witness Johnson notes that the Commission could utilize more sophisticated usage-based allocation methodologies. Correspondingly, GTEN stated that if the Commission did allocate loop costs, it should use SLU.

Significantly, there is almost virtual unanimity of opinion among the parties that allocations of this sort are never an exact science. OCA witness Johnson explained that allocating joint or common costs by any allocation percentage is always somewhat arbitrary. "While allocation of fixed costs is an inherently arbitrary process, there are cost allocators that, though arbitrary, are reasonable." OCA Stmt. 1.3 (Johnson) at 5. Dr. Kahn also testified that anything less than 100% is properly characterized as "arbitrary." Id. Nor we note are we bound through relevant case law to base our allocations on only those equations or formulas resulting in great mathematical precision. See Smith v. Illinois Bell Telephone Company, 282 U.S. 133 (1930). As long as there is a reasonable basis for our decision, it meets present legal standards in this area.

While we accept the underlying position of the OCA that the local loop is a "joint cost", we reject its recommended allocation of 50% for several reasons the most important of which is that we do not find revenue breakdowns and administrative simplicity to be the proper basis for determination of this issue. While any allocation process is inherently arbitrary, neither administrative simplicity or revenue breakdowns provides any degree of precision or relation to actual use of the network.

We believe use of SLU would be most appropriate at this time, until we have an opportunity to study the matter further and determine the effect of our policies adopted today on achieving the very important objectives of both the Federal Act and Chapter 30. SLU correlates the actual use of the network to the percent allocated to BUS, which would provide a more reasonable basis for allocation initially. While we declined to adopt a SPF or SLU based allocation in the Breezewood Telephone Company case, we did not do so based upon any perceived deficiency with either allocation methodology. Rather, we suggested that the "industry continue to look at the various cost allocation methodologies for the purpose of facilitating a more detailed analysis of these methodologies under appropriate circumstances." Id. at p. 494. We, therefore, will adopt the use of SLU initially, for use in allocating costs to BUS for the larger ILECs in Pennsylvania including Bell, GTEN, Sprint/United and Commonwealth. Most of these companies, with the exception of Commonwealth, are cost companies and must do traffic studies on a regular basis.

It is our intent to reevaluate the policies adopted herein should we find that companies are not meeting Chapter 30's objectives. One of the primary goals of both Chapter 30 and the Federal Act is network modernization. The Federal Act, in particular, through § 254 contemplates that fund recipients will put the monies received back into the network, to achieve in part, network modernization. Reinvestment works to everyone's benefit including the ILEC's competitors who either purchase unbundled elements from the ILEC or purchase ILEC services for resale. If we find that state universal service funds are not being used to meet the important objectives of both Chapter 30 and the Federal Act, we will take corrective action and modify our policies accordingly.

Because the current SLU figures were not readily available from the record of this proceeding, we obtained estimates from some of the larger ILECs in Pennsylvania, in order to obtain an idea of what the funding levels would be necessary. Based upon data we have been able to obtain for GTEN and Bell, their local exchange SLU percentages approximate 74%. For purposes of estimating fund size and the impact upon individual carriers, we have assumed a 74% SLU for the larger carriers including Bell, GTEN, ALLTEL, Sprint/United and Commonwealth. Actual funding levels, which we do not anticipate will vary significantly from the estimates contained herein, will be determined from the updated SLU information submitted by the companies in Phase II of this proceeding. In Phase II, the larger Pennsylvania LECs including Bell, GTEN, ALLTEL and Sprint/United will be required to provide updated SLU data to us by June 1, 1997. It is our intent to use an average of these as a proxy for the SLU of Commonwealth, an average schedule company.

We find, however, that there are compelling reasons to allocate a higher percentage of the loop to BUS for the smaller Pennsylvania LECs, including North Pittsburgh and those ILECs that qualify for streamlined treatment under Chapter 30. For the following reasons, we believe that a 90% allocation of the local loop is appropriate at this time for smaller Pennsylvania LECs. First, many of the smaller Pennsylvania LECs are average schedule companies, and do not have the capability to perform the traffic studies necessary to determine SLU. Second, and perhaps most importantly, small rural Pennsylvania LECs do not currently offer the array of non-BUS services that larger LECs operating in Pennsylvania offer. Most of the smaller LECs do not offer non-BUS services to the extent that the larger LECs do in Pennsylvania. In this regard, we recognize that the consequences of misallocation would be far more disastrous to the smaller LECs since they do not have the array of services to provide contribution to their firm. Finally, we expect that all companies will make a good faith effort to comply with the requirements of Chapter 30 and we will reevaluate our policies established herein to ensure that our policies are effective in this regard.

2. Common Costs

Common costs are costs that are incurred by the company in order to run the business as a whole; but they are not related to any single service or subset of services. Bell Stmt. 1.0 (Sanford) at p. 14. Common costs include costs such as administration and legal expenses. As a prelude to our discussion of common costs, we take note of the testimony of GTEN witness Cole who states that "[p]roper cost allocation is an economic oxymoron." Dr. Johnson, according to witness Cole, misrepresents the economics literature when he states that there is no agreement in that literature on how to allocate common costs. Witness Cole went on to testify that there is agreement in the economics literature that no common cost allocation methods contain any economically meaningful cost-causative information. GTEN Stmt. 8.1 (Cole) at p. 10.

OCA witness Johnson used a 50% allocation factor for joint costs, and added 10% to the direct costs as an allowance for common costs. OCA Stmt. 1.0 (Johnson) at p. 33. Similarly, the sponsors of both the Hatfield and BCM models utilized an add-on of 10% to direct costs as an allowance for common costs. (Hatfield subsequently adjusted its common cost allocation down to 6% based upon the airline and manufacturing industries.)

Bell, on the other hand, states that it has included bill inquiry and account maintenance costs as shared costs which have not been included in the cost of any other service or category. Bell Stmt. 1.0 (Sanford) at p. 15. Witness Sanford also testified that the majority of shared-fixed costs were allocated based on the proportion of direct investment. Common overhead costs were included by applying a common overhead factor to the cost of the total BUS obligation (excluding common overhead). A special study was performed to arrive at the factor for which Bell has claimed proprietary protection.

Sprint/United witness Jamison testified that including a portion of shared costs is appropriate because revenues from services must generally make a contribution to covering shared costs if a company is to stay in business. Sprint/United Stmt. 1.0 (Jamison) at p. 11. Witness Jamison testified that in general the amount that TSLRIC should be marked up should be based on the quantity of shared costs a company has relative to its overall costs:

The markup would be calculated by commingling two quotients: (1) the result of dividing the company's true overhead costs by the sum of its TSLRICs for all major service categories; and (2) the result of dividing the company's shared incremental costs that belong to the services in universal service by the associated TSLRICs. This calculation represents an average amount of mark up above TSLRIC that a company needs to cover its shared costs.
Id. at p. 12.

GTEN witness Williams argues that ordinarily the contribution amount would be based on an analysis of the demand characteristics of the service in question. However, since that is not available for universal service, another basis must be found. GTEN Stmt. 1.2 (Williams) p. 12. Another method of determining contribution is to analyze the aggregate contribution resulting from GTEN's current prices for those services which utilize the local loop but GTEN has no such study available at the present time. Id. The other method according to GTEN witness Williams is to compare the cost data provided by Dr. Cole to embedded cost. GTEN witness Williams, using the embedded loop cost from data submitted to NECA developed such a comparison and based thereon, is recommending that a 40% contribution rate be employed to determine contribution. Id.

We find the 10% allocation recommended by several parties to be reasonable, and allocations at this level shall be presumed reasonable. Nonetheless, we believe that companies should be given the option of submitting studies such as that suggested by Sprint/United above. If the results of those studies indicate that a different allocation of common costs should be made, this should be considered in determining the appropriate allocation percentage.

F. Asset Impairment

GTEN asks the Commission to recognize and give effect to what it calls a "regulatory compact" between GTEN and this Commission. The nature of this compact according to the Company is that, while not a formal, signed agreement, it has governed the relationship between the incumbent LECs and regulatory commissions for the last 60 years. GTEN Main Brief at p. 28. According to GTEN the regulatory compact obligated the incumbent LEC to provide universal service in exchange for the opportunity to earn a fair return on its investments in an exclusive franchise area. GTEN Main Brief at p. 28.

GTEN argues that pursuant to this regulatory compact, the Commission set depreciation rates for the Company which "permitted recovery over a longer period of time than that which would have been appropriate in a competitive marketplace." GTEN Main Brief at p. 29. The purpose of this was to keep local exchange rates affordable. Id. GTEN goes on to charge that now in a competitive environment it is unlikely to have the ability to recover any under-depreciated asset investments which results in "asset impairment".

GTEN witness Sovereign determined the level of GTEN's asset impairment by using the remaining lives of investments to provide local exchange service. Witness Sovereign testified that the prescribed regulatory remaining lives had a composite average of 11.4 years while the economic remaining lives had a composite average of 5.9 years. All together, GTEN witness Sovereign calculated that its local exchange asset impairment resulting from the new competitive marketplace would total $241 million of which $187 million is intrastate. GTEN Stmt. 2.1 (Sovereign), p. 2. GTEN requests that it be permitted to recover the $187 million asset impairment from the funding mechanism. GTEN Main Brief at p. 29.

GTEN proposes to recover approximately $187 million through the Universal Service Fund for "asset impairment." Impaired assets would be given a shorter remaining life in order to assure full asset recovery in the new world of competition. GTEN Stmt. 1.2 (Williams) p. 77; Tr. 94-96. GTEN proposes to recover the remaining under-depreciated portion of these assets over six years, at the rate of $23.6 million per year. GTEN Stmt. 1.2 (Williams) p. 77; Tr. 94-96. GTEN argues that this is appropriate given the emergence of competition in the local exchange market and the Company's claim that it will not fully recover the under-depreciated portion of its investments. GTEN Stmt. 2.1 (Sovereign) at 4. GTEN argues that under the traditional regulatory process, the Commission had previously 'assured' full recovery of capital investments.

We agree with those parties whose position is that this is not the appropriate forum to determine whether GTEN or other ILECs should be permitted to recover allegedly "impaired assets" on an accelerated basis. It is wrong to guarantee recovery from all Pennsylvania telephone customers outside the context of a rate proceeding for GTEN's alleged under-depreciated assets. Accord, OCA Main Brief at p. 49. Nor is GTEN entitled to full recovery or guaranteed recovery of capital investments. Duquesne Light Co. v. Barasch, 488 U.S. 299, 316, 109 S. Ct. 609, 620 (1989). Simply put, the purpose of the state universal service funding mechanism is not to guarantee the revenue requirement of any ILEC. Such a request must be submitted in the context of a Section 1308 proceeding.

Having determined how BUS costs will be determined, the next step in the equation is the determination of the BUS rate.