VI. Assuming that the Reciprocal Compensation Provisions of the Act are not applicable to ISP Traffic, Do Other Factors Justify Reciprocal Compensation for ISP Traffic?

A. Effects of Reciprocal Compensation Policies on Incentives to Promote Competition and Economic Efficiency

1. Parties' Positions

The ILECs argue that reciprocal compensation for ISP calls is detrimental to competition because it results in asymmetrical windfall profits to CLECs, providing the CLECs with an unfair competitive advantage. Pacific's and Verizon's customers originate several times more traffic destined for ISPs served by CLECs compared with the volume of originating CLEC calls that are destined for ISPs served by Pacific and Verizon. As a result, the ILECs claim they pay out considerably more reciprocal compensation to CLECs than they receive in return for ISP traffic. During 1999, Pacific claims that it sent 833 minutes of ISP-bound traffic to CLECs for every one minute of ISP-bound traffic sent by a CLEC to Pacific. Moreover, Pacific's measurements indicate that 73% of all CLEC traffic during 1999 was attributable to ISP-bound calls. Similarly, Verizon reports that CLECs have billed it for $32 million in ISP-related reciprocal compensation over the most recent 18 month period while Verizon has billed CLECs for only $0.4 million.

The ILECs claim that instead of increasing competitive alternatives to customers, ISP reciprocal compensation actually reduces CLECs' incentive to serve residential customers. The ILECs claim the CLECs instead have simply rushed to serve ISP customers who generate one-way traffic that ensures a steady stream of reciprocal compensation payments, and an opportunity for arbitrage due to the unintended consequences of regulation. Since ISPs originate very little traffic, CLECs pay only very small streams of ISP-related reciprocal compensation payments in the direction of the ILECs. By contrast, regular voice traffic tends to flow more evenly in both directions, creating a more balanced exchange of reciprocal compensation payments between carriers.

Pacific claims that paying reciprocal compensation to CLECs for this asymmetrical ISP traffic flow runs contrary to the goal of promoting competition, particularly in the residential market. Pacific claims the current system actually disincents CLECs from serving residential customers, because the CLECs would have to pay reciprocal compensation to other carriers. Pacific claims that residential customers that call the Internet become huge liabilities to originating carriers, retarding the growth of residential competition. Pacific also claims that CLECs have little incentive to develop new technologies for offering Internet access since doing so would reduce the flow of reciprocal compensation CLECs currently enjoy.

The CLECs do not dispute that a disproportionate share of ISP traffic is terminated by CLECs in contrast to the share terminated by ILECs. The CLECs, however, do not attribute this fact to anticompetitive arbitrage or to improper incentives. Rather, the CLECs view this outcome as a result of positive competitive forces. The CLECs argue that applying reciprocal compensation payments to ISP-bound traffic is conducive to competition, creating a strong incentive for ILECs to become more cost efficient and creating a basis for CLECs to build their business. Conversely, the CLECs argue that eliminating reciprocal compensation would harm local competition.

Focal argues that withholding reciprocal compensation for ISP-bound traffic would penalize CLECs for successfully competing for ISPs by precluding them from recovering the cost of terminating calls to those customers, thus discouraging CLECs from serving ISPs and limiting the competitive choices available to ISPs. Withholding reciprocal compensation for ISP-bound calls could also harm competitive LECs because it would limit their ability to rely on the high call volumes received by ISPs to reduce their per-unit costs and develop the economies of scale and scope currently enjoyed by incumbent LECs and needed to effectively compete with incumbent LECs. Finally, the Commission should take special note of the possibility that moving to bill-and-keep could have adverse consequences for the Internet, which would have significant consequences for the California economy.

CISPA argues that eliminating ISP reciprocal compensation will only bolster ILEC efforts to assert control over California's ISP market. CISPA argues that ILECs and their internet affiliates have specific designs on the internet services market in California. Patterns of discrimination against independent ISPs have already developed in Pacific Bell's service territory. CISPA claims ISPs in California have experienced service quality or other problems as reported in a national ISP survey, demonstrating serious problems with Pacific Bell's ability to serve ISPs and their end users. The evidence demonstrates, at a minimum, the value which independent ISPs place on competitive choice.

CISPA argues that ISPs do not have sufficient safeguards protecting them from ILEC discrimination or misconduct. As end users of telecommunications services, ISPs do not have the benefit of telecommunications laws and regulations developed to ensure competition among telecommunications carriers. ISPs lack recourse for an ILEC's decision to delay network capacity upgrades. The absence of relief for ISPs means that Pacific (or its affiliate SBC Advanced Solutions, Inc.) can delay installation of facilities such as a Digital Subscriber Line Access Multiplexer ("DSLAM") in a central office until its affiliated ISP has secured a customer base to fill the available ports in that DSLAM. Meanwhile, Pacific (or SBC Advanced Solutions, Inc.) remains free to ignore the pending orders of independent ISPs for DSLAM ports. Additionally, independent ISPs do not know how Pacific shares an ISP's customer proprietary network information ("CPNI") with its affiliates; however, it appears that this information is exchanged with Pacific's internet affiliate for purposes of marketing.

2. Discussion

We find no evidence that the continuation of the existing policy calling for the payment of reciprocal compensation for ISP traffic will impair incentives for LECs to compete in an economically efficient manner. Under the present policy there has been a growth in the choice of telecommunications service providers among ISPs. The availability of greater choice in the availability of service providers is good for competition. We find no convincing evidence that our present reciprocal compensation policies are to blame for the fact that there hasn't been greater progress in the development of competition among residential customers. The fact that the customers of the ILECs originate the overwhelming majority of calls to ISPs is to be expected given that the vast majority of the residential customer base continues to be served by ILECs. As noted by the CLECs, there are a number of constraints that have been identified as contributing to the CLECs' failure to garner a larger share of the local residential market. Many of these constraints are being examined by the Commission in connection with the Commission's "271 Proceeding," which involves review of a checklist of factors affecting the competitiveness of the local market.²⁰ We find no basis to conclude that the CLECs would become more active in the residential market if bill-and-keep was substituted for reciprocal compensation for ISP calls.

We find no basis to conclude that our reciprocal compensation policy merely creates an incentive for CLECs to sign up ISPs for the purpose of arbitraging "windfall" profits. To the extent that certain CLECs have a financial incentive to sign up ISPs, the CLECs are not simply arbitraging profits, but provide a legitimate service to the customers of ILECs by delivering their calls to ISPs. We examine in more detail below the allegations that reciprocal compensation results in a "windfall" to CLECs. The fact that certain CLECs have focused a much greater share of their target market on serving ISPs in comparison with ILECs is not, in itself, an anticompetitive result. Although niche markets may develop with certain carriers specializing in serving certain market segments such as ISPs, this process can actually promote a more diverse and dynamic competitive market.

By contrast, we are concerned that the elimination of reciprocal compensation could trigger undesirable consequences that would not be conducive to competition. CLECs would be faced with the choice of either raising their rates to ISPs to make up for the lost reciprocal compensation, or else curtailing service to ISPs if that segment of the market became unprofitable. ISPs, in turn, would face reduced competitive choices for their local exchange service or paying higher local telephone rates. ISPs may become more dependent on the ILECs for their service. Yet, the ILECs may choose to give priority to their own ISP affiliates. In the event of inferior service from ILECs, ISPs would have less recourse to seek competitive alternatives. The ILECs thus would have less incentive to improve the quality of their service to ISPs in order to avoid losing their business to CLECs. The CLECs' loss of reciprocal compensation revenues could also lead to higher telephone charges to ISPs to make up the shortfall. Subscribers of ISPs would face the prospect of potentially higher ISP subscription fees, or per-minute charges, to the extent ISPs sought to pass through any local telephone service rate increases to their own subscribers.

B. Can ISP Traffic Be Accurately Identified and Segregated from other Traffic ?

1. Parties' Positions

Parties dispute whether ISP-bound traffic can be accurately measured and readily segregated from other local traffic on an ongoing basis for purposes of applying a different compensation method from other traffic. If ISP calls were to be excluded from reciprocal compensation payments, some method would be needed to properly identify and segregate ISP calls from other calls subject to the payment of reciprocal compensation.

The CLECs argue that the lack of any reliable system for accurately segregating ISP-bound traffic from other traffic points up the impracticality of imposing a different compensation method for ISP versus other local traffic. The CLECs argue that any attempts to ascertain from customers whether they are using a particular line for ISP purposes would intrude on the privacy of callers. The CLECs also argue that denial of reciprocal compensation would be discriminatory and impractical to implement since calls to ISP are functionally identical to voice-grade calls and cannot be separately identified for billing purposes.

Pacific believes that ISP traffic can be reasonably identified, and is currently making efforts to do so. The CLECs have already been ordered by the Commission to keep track of this ISP-bound traffic. Pacific's intent is merely to track ISP-bound calls in the aggregate so that no customer's privacy is compromised. Pacific also notes that in their filings with the Securities & Exchange Commission, various CLECs have been able to specifically identify the number of their ISP customers. For example, Pac-West states that it is "a leading supplier of Internet access and other Internet infrastructure services in California serving 78 Internet service providers."²¹ ICG states that at the end of 1999 it had "approximately 550 ISP customers."²² Moreover, these CLECs specifically direct their marketing activities at ISPs.²³ Thus, Pacific argues that with the exception of a few minor "grey areas," CLECs are readily able to identify ISPs.

Pacific developed independent estimates of the volume of ISP traffic that has been terminated by CLECs for this proceeding. The source for Pacific's figures for an ISP-bound traffic was a study identified as the "Barry Lear Study." The Lear study used a four-step method to identify ISP traffic terminated by CLECs. The four steps are as follows:

1. Pacific developed a list of ISP telephone numbers by searching the Internet for ISP advertisements or Web sites that identified a telephone number of its service;

2. For identifying additional Internet traffic, Pacific applied the selection criteria that calls to a ISP telephone number would be those which receive more than 200 calls per day, or average more than 25 minutes of conversation time per call;

3. Pacific verified the suspected ISP numbers by calling the number to determine that a machine tone was received on the line;

4. Pacific next compared the list of ISP numbers to match with the codes or prefixes for each CLEC.

Pac-West disputes the reliability of Pacific's figures measuring the volumes of ISP-bound traffic it has sent to CLECs. Pac-West claims each of the four steps creates significant opportunities for errors and misclassification of traffic, both in terms of false positives and false negatives. Pac-West argues that step (1) does not capture all of the telephone numbers that ISPs use to terminate calls. Pacific may miss certain advertised numbers, and new ISP dial-up telephone numbers are being introduced all the time. Moreover, some ISP dial-up telephone numbers may not be publicly advertised in mass-market sources, or the numbers may be grandfathered to existing subscribers and thus no longer advertised. In addition, Pac-West argues that many ISPs employ shared modem pools in which the same telephone numbers are used for ISP and non-ISP purposes, so that attempting to classify such a number as terminating either only ISP-bound traffic or only non-ISP bound traffic will necessarily fail. Pacific witness Jacobsen was unable to validate the legitimacy or accuracy of this step, and admitted that he did not know what qualifies somebody to be an ISP.

Pac-West claims step (2) in the study only creates further problems. By assuming that ISP dial-up numbers will have average call durations exceeding 25 minutes or will receive more than 200 calls per day, Pacific excludes all dial-up calls to ISPs below these thresholds that were not already detected in step (1). Pac-West argues that by filtering in this arbitrary fashion, Pacific guarantees that the sample of ISP calls are non-random and biased toward higher volumes and longer durations. Pacific also includes non-ISP calls that meet the thresholds described in step (2).

Pac-West also criticizes step (3), in which Pacific assumes that hearing a machine tone on a called line means that the line terminates to a modem that will always provide a connection to the Internet. Pac-West states this is clearly not the case.

Verizon has not historically tracked originated or terminated calls that were specifically identified as ISP-related. Verizon witness, Beauvais, claims, however, that if the telephone numbers assigned to ISPs are known, that CLECs should be able to track precisely the amount of delivered traffic that it ISP-bound. In any event, Beauvais believes that useful estimates of ISP-bound traffic by carrier can be developed based upon an algebraic formula utilizing call duration as a defining variable.²⁴ Based on Verizon data from North Carolina and Michigan, observed duration for Verizon to CLEC calls ranged from 15 to 45 minutes while the duration for calls from CLEC to Verizon ranged from only 3 to 4 minutes. Beauvais observed that the available California data yields consistent results with a range from 3.5 minutes to 8.7 minutes for traffic inbound to Verizon customers whereas the duration for outbound traffic to CLECs ranged from 8.5 minutes to 23.2 minutes. Beauvais views the duration differences as being largely attributable to the disproportionate ISP-related business of the CLECs.

Pac-West disputes the reliability of Verizon's claims concerning the ability to accurately measure ISP-bound traffic, arguing that its study relies on two critical assumptions that are not correct. First, the study assumes that duration of two categories of calls--voice and ISP-bound traffic--are known with sufficient precision. The second assumption is that there are only two categories of calls to be distinguished. Pac-West further argues that range of potential outcomes resulting from Beauvais' algebraic formula is too broad to be used to produce a meaningful measure of ISP-bound calls or minutes. Pac-West notes that the range of possible variation in the percentage of ISP-bound minutes in Beauvais' formula is over 20%. The percentage of ISP-bound calls derived from Beauvais' formula could thus range between 39.8% and 60.9% of total minutes at a 99% confidence interval. Applying the Verizon methodology to the total quantity of minutes handled by Pacific, Pac-West computes that the range of possible outcomes for ISP-bound minutes could vary by 5.2 billion minutes, as noted in the table below:

TABLE 1:

VARIABILITY OF MINUTE CATEGORIES USING VERIZON FORMULA

Pacific Bell estimate of total local traffic (in calls): 51 billion
(1 billion ISP-bound, 50 billion all other)²⁵

Voice Calls:

ISP-bound Calls:

Verizon argues that although the use of statistical techniques would result in certain individual voice calls being classified as ISP-bound calls and vice versa, that is not in itself a justifiable reason to refrain from using these techniques. The Commission and the CLECs have been willing to use estimation techniques in a variety of circumstances, notwithstanding that the process does not identify each call. For example, existing interconnection agreements between Verizon and CLECs employ a statistical estimation technique to separate local traffic from toll.²⁶

Verizon also argues that parties could conduct traffic studies that sample traffic flowing between ILEC and CLEC to determine a reasonable figure for ISP traffic. Verizon believes parties could readily identify their own ISP customers.

2. Discussion

We address the question concerning the extent to which ISP traffic can be accurately measured for three purposes. First, we consider the question from the standpoint of the accuracy of the ILECs' representations concerning the extent of the imbalance in ISP traffic between the ILECs versus the CLECs. Second, we consider the question from the standpoint of the implications of those measures in terms of the potential financial effects, both on the CLEC and the ILEC. Third, we consider the question from the standpoint of whether a practical method exists to segregate ISP traffic from other traffic for the purpose of applying a bill-and-keep treatment, or some other different treatment in comparison with other types of calls.

With respect to the first question, we acknowledge that there are statistical limitations in the estimation techniques used by Pacific and Verizon in seeking to quantify the magnitude of minutes terminated by CLECs. Yet, the purpose for which these estimates were presented were merely to provide some order of magnitude of the huge asymmetry between ISP calls terminated by CLECs versus ILECs. The ILECs could only infer through indirect means how many terminating minutes of the CLECs involved access to the Internet.

We find the measurement of ISP traffic derived from Pacific's "Barry Lear Study" to be unreliable. Due to its absence from the record, the specifics of the study are not known, such as precise dates of traffic data used, amount of sampling and sampling techniques used, confidence level of the resulting data, etc. Moreover, on cross-examination, Mr. Scholl revealed that certain data proffered by Pacific, such as call completion rate for ISP calls, were derived from a separate engineering study. This study, like the Lear Study, also was never submitted for the record. Pacific witness Scholl, while admitting that he did not participate in or review the engineering study, or use it for any of the Pacific cost studies he performed, stated that it involved only 34 (unspecified and not randomly selected) out of 900 end offices.²⁷

In the case of their own ISP customers, the ILECs had access to more direct and specific record of call termination. Yet, even here, some questionable assumptions and approximations were involved in reporting the results. Even if we take into account the potential measurement bias and statistical limitations pointed out by the CLECs, the ILECs estimates still provide rough approximations of the differences between CLEC and ILEC terminations of ISP traffic. We are still left with the conclusion that there is a very large asymmetry between CLEC and ILEC terminations, even if it cannot be quantified precisely.

With respect to the question of financial impacts of the asymmetrical flow of traffic, we address that question separately in Section VI.D.

With respect to the question of whether ISP traffic could be separately measured and segregated from other traffic for intercarrier billing purposes, we conclude that such results as presented by the ILECs are too imprecise to be useful. As noted above, the range of possible outcomes using Verizon's estimating methodology produces a range of 130 billion minutes for Pacific and 5.2 billion for Verizon. As we noted above, the ILECs have demonstrated that it may be possible to achieve some approximation of the amount of ISP traffic flows only on a broad level. For example, the ILECs have provided examples of published financial reports of the CLECs in which specific numbers of ISP customers served are identified. By inference, the CLECs must have some means of identifying those customers acting as ISPs in order to identify them in their published financial reports. The question is whether any approximation that could be measured would be reasonably accurate enough to be used for intercarrier billing purposes. We conclude that the range of variability in the estimates presented in this proceeding is too broad to serve as a basis for billing purposes.²⁸

Beavais' methodology addressed only the proportion of calls that have longer durations, not the proportion of calls that are ISP-bound calls. Such a methodology based solely on call duration to determine the proportion of ISP-bound calls is inherently unreliable because it fails to exclude classes of long-duration calls other than ISP-bound calls (e.g., telecommuting and other calls to corporate LANs, business conference calls, calls to airline reservations offices, etc.). Witness Beauvais appeared to acknowledge, however, that under his methodology, calls other than those bound for ISPs would be treated as part of the ISP-bound aggregate based on their holding times, and he offered no reliable solution for the problem.²⁹

Aside from the difficulties in accurately measuring calls delivered to ISPs, an additional measurement difficulty involves distinguishing calls to ISPs which actually involve transmissions over the Internet. As Pac-West witness Goldstein testified, of the calls that are actually made for the purpose of using the Internet, many of them are carried out with no actual connection to the Internet, only a temporary Internet connection, or intermittent connections. When not connected to the Internet, the end user may be connected only to the local server of the ISP or to the ISP modem. As testified by witness Terkeurst, various ISP services utilized by a subscriber would not entail connection to the Internet. For example, retrieving e-mail typically only involves accessing the ISP's local e-mail server. Another example could entail viewing web pages that have been locally stored (i.e., cached) on the server of the ISP. No party has proposed a means by which the minutes of usage for ISP communications can be delineated between those that actually involve connection to the Internet versus those that remain locally with the ISP. Without some means of segregating such minutes of use, measurement process used for billing purposes would yield inaccurate results.

C. Does the payment of Reciprocal Compensation to CLECs for Terminating ISP Traffic Constitute a "Windfall"?

1. Overview of Parties' Positions

As one of the reasons supporting their opposition to the payment of reciprocal compensation for ISP traffic, the ILECs claim that it results in windfall profits and subsidies to the CLECs. The ILECs claim that the reciprocal compensation rate paid for local traffic significantly exceeds the actual costs incurred by the CLEC to deliver a call to an ISP, resulting in "windfall" profits. The ILECs claim that since the reciprocal compensation rates are predominantly based on the termination costs for local voice calls, the application of the same rate to ISP calls significantly overcompensates the CLECs. Pacific claims that at least 50% of reciprocal compensation revenues paid to CLECs constitute pure profit, and possibly even more. (Scholl Exh. 106, pp. 19-23.)

The ILECs attribute the lower costs of delivering ISP traffic to differences in the type of facilities and processes used in comparison with those used by the ILECs that are used for delivering voice traffic, as well as to differences in the characteristics of ISP calls, themselves. The ILECs claim that ISP calls (1) are longer, on average, than voice calls; (2) exhibit a higher call completion ratio than voice calls; (3) are made to called parties that are likely to be collocated with the CLEC; (4) require more ILEC tandem switching and transport than voice calls; (5) represent traffic that is aggregated by the ILEC before being delivered to the CLEC; and (6) can be switched by the CLEC at a lower cost than voice traffic. By being required to pay reciprocal compensation rates based upon the higher costs of terminating voice traffic, the ILECs argue, the resulting payments constitute a "subsidy" to CLECs and result in "windfall" profits.

The CLECs argue that the ILECs' "windfall"/"subsidy" argument is inconsistent and cannot be reconciled with the FCC's determination that symmetrical compensation should be applied to all local traffic.

The FCC First Report and Order provided for the payment of reciprocal compensation for local traffic based on "symmetrical rates based on the incumbent LEC's costs for transport and termination of traffic . . . " Id. at ¶ 1089; see, 47 C.F.R. § 51.711(a).

The CLECs also deny that their ISP termination costs are lower than costs for other traffic termination and claim the ILECs misconstrue the manner in which CLEC switches have been deployed. ICG claims the alleged differences in call completion ratios, digital to analog conversions and other purported differences between ISP-bound calls and non-ISP-bound calls, even if they could be accurately determined, are irrelevant to a proper determination of terminating switching costs for such traffic. ICG attributes each of the individual arguments made by the ILECs in attempting to distinguish costs associated with ISP-bound traffic from other types of traffic as either (a) factually inaccurate, (b) irrelevant to the derivation of traffic sensitive costs, or (c) already accommodated by the rate structure included in interconnection agreements."

2. Discussion

The Act prescribes an overall framework by which carriers are to be compensated for their costs of providing competitive local exchange telecommunications services. There are three general categories of service that a LEC provides. These are (1) connecting its own customers to the telecommunications network; (2) permitting its own customers to originate traffic destined for customers of its own (or other LECs') networks; and (3) terminating traffic destined for its own customers that was originated by customers of its own (or those of other LECs). (Starkey/pg.17-18).

Under the Act, reciprocal compensation only applies to the third category of service, namely the termination of calls. Specifically, reciprocal compensation is intended to cover the "traffic-sensitive" costs incurred for transport and termination of local traffic, that is, those costs that vary directly as a function of the traffic usage involved with the call. As prescribed in its Local Competition Order (§ 1057), the costs of terminating traffic that are not traffic sensitive (e.g., local loops and line ports), are not to be included in the reciprocal compensation allowance. Instead, these costs must be recovered from each carrier's own end-use customers. The FCC has determined that such reciprocal compensation obligations "apply only to traffic that originates and terminates within a local area;" they "do not apply to the transport or termination of interstate or intrastate interexchange traffic." (Local Competition Order 11 FCC Rcd. 15499, 16013, § 1034 (1996).)

The FCC has defined "transport" in this context "as the transmission of terminating traffic that is subject to Section 251(b)(5) from the interconnection point between the two carriers to the terminating carrier's end office switch that directly serves the called party (or equivalent facility provided by a non-incumbent carrier)." (Local Competition Order at § 1039, see 47 CFR § 51.701(c).) "Termination" is defined as "the switching of traffic that is subject to Section 251(b)(5) at the terminating carrier's end office switch (or equivalent facility) and delivery of that traffic from that switch to the called party's premises." (Local Competition Order at § 1040; see 47 CFR § 51.701(d).)

FCC rules implementing the 1996 Act call for the use of the Total Element Long Run Incremental Cost (TELRIC) of the ILEC as a proxy for CLEC costs rather than separately requiring CLEC-specific cost studies. Section 51.711 requires that reciprocal compensation rates be "symmetrical" and defined as: "rates that a carrier other than an incumbent LEC assesses upon an incumbent LEC for the transport and termination of local telecommunications traffic equal to those that the incumbent LEC assesses upon the other carrier for the same services." Pursuant to the existing FCC rules, therefore, ILECs must charge the CLECs a rate for reciprocal compensation based on the ILEC's costs, and the CLEC must likewise charge the ILEC the same rate (based on the ILEC's costs). There is no option under the FCC's rules for an ILEC or state commission to impose asymmetrical rates for traffic deemed to be "local." Because the FCC rules require that reciprocal compensation rates be based on the level of the ILECs costs, ICG argues that in any event, the Commission does not need to know the current level of actual CLEC costs. Therefore, in the scoping memo for this proceeding, we did not ask CLECs to produce separate cost studies, but rather, sought inquiry concerning the cost characteristics of those functions that are involved in the termination of traffic. Our long standing policy as originally adopted in D.96-03-020 has been not to impose separate cost study requirements on CLECs, recognizing the administrative burden such studies would impose, and the lack of market power that CLECs exercise. Therefore merely because the CLECs did not produce their own comprehensive cost studies in this proceeding, we should not conclude that the CLECs failed to make a proper evidentiary showing. Thus, the cost principles underlying TELRIC provide a relevant standard in evaluating the costs of terminating ISP traffic by either the ILEC or the CLEC.

Federal TELRIC rules require that the cost of a "particular element must be derived by dividing the total cost associated with the element by a reasonable projection of the actual total usage of the element."³⁰ Thus, it is consistent with the TELRIC methodology to apply one uniform TELRIC-based rate for all calls that are subject to termination using the same facilities. The CLECs have shown that they terminate ISP calls using the same facilities as are used to terminate other local calls. Therefore, there is no basis to disaggregate one particular customer class, such as ISPs, and treat them as having a different cost since to do so would contradict the TELRIC principles of costing based upon the total cost of a discrete network element.

The reciprocal compensation rates currently in place for interconnection agreements between Pacific and various CLECs are based on the TELRIC as adopted in the OANAD proceeding applicable to Pacific's unbundled network elements (UNEs) for terminating switching and transport costs. The rates are reciprocal in that each LEC pays that rate to the other LEC for any local traffic that is terminated. Thus, no separate cost studies are performed for the CLEC, but the ILEC's TELRICs are deemed to be acceptable proxies of CLEC costs for purposes of paying reciprocal compensation. The UNE rates for Pacific are disaggregated into two components, as follows:

$.007 per call

$.00187 per minute

The per-call rate is fixed irrespective of the duration of a particular call. The per-minute rate reflects those costs that vary in relation to the volume of traffic terminated. Thus, costs that are not traffic sensitive would not be relevant in evaluating a carrier's actual cost of terminating local traffic subject to reciprocal compensation.

The Commission has not yet established TELRICs for Verizon in OANAD. The rates that the Commission approved in GTE California's (now Verizon's) interconnection arbitration with AT&T in January of 1997 have effectively served as default rates for UNEs and for reciprocal compensation insofar as parties have been able to opt into those rates. The reciprocal compensation charge established in that proceeding and set forth in the interconnection agreement with AT&T is a per-minute charge of $.003629/minute. (See Tr. 29:1-30:10 & Exh. 5 (GTE/AT&T Interconnection Agreement Excerpt) at Attachment 14, App. 1, p. 4.) Although Verizon has, in some instances, been able to negotiate different rates, this ability is limited by the availability of the AT&T rate.

Verizon's reciprocal compensation rate is typically set equivalent to its end-office switching UNE.³¹ Unlike Pacific, Verizon only one blended UNE rate for end-office switching,³² rather than separate rates for "terminating" and "originating" switching. As result, the rate includes the cost of using certain origination-related switch equipment - such as dual tone multi-frequency (DTMF) receivers and tone generators³³ - that a CLEC does not use when it terminates calls to ISPs. Verizon argues that having origination functions in the reciprocal compensation charge overstates the termination cost regardless of the type of traffic at issue. We believe that if origination functions are included in Verizon's reciprocal compensation charge for ISP-bound traffic, those costs should be stripped out.

We find no necessity that CLEC costs must exactly equal the ILEC costs in every respect to justify the payment of reciprocal compensation for ISP calls. By virtue of being an aggregate of total costs, the TELRIC measure may well deviate from the specific cost for any one particular type of call termination. If the CLEC is able to terminate ISP-bound calls at a cost below the TELRIC rate, that, in itself, is no basis to conclude that the CLEC is earning a "windfall" or is not entitled to be compensated by the ILEC at all. The use of TELRIC as a standard for compensation on a reciprocal basis provides a benchmark against which all carriers must manage to provide terminating services at the lowest cost possible.

To the extent that ISP calls may have certain characteristics that distinguish them from voice calls, we conclude that difference, in itself, doesn't justify excluding ISP calls from reciprocal compensation. The ILECs repeatedly compare ISP calls to voice calls, but fail to definitively compare ISP calls with other data-related or other specialized business-related calls. If ISP calls were to be earmarked for disparate treatment from all other local calls, we would also need to consider whether such treatment constituted a form of unfair discrimination. We would need to consider whether certain types of calls other than voice calls that may exhibit similar characteristics to ISP calls such as longer duration or higher volume such that they should also be exempted from reciprocal compensation, or at least compensated at a different rate.

The ILECs however, claim, to use the term "windfall" and "subsidy" to characterize the difference between the TELRIC rates paid and the true costs incurred by the CLECs to deliver ISP traffic. The ILECs do not precisely quantify a standard as to how much of a profit margin may constitute a "windfall." The ILECs' primary argument appears to be, however, that the marginal CLEC profit is so huge, however, that a precise threshold need not be drawn.

Based on our review of the parties' allegations, we do not find evidence of huge differences in the cost of termination of ISP-bound traffic compared with that of all other local traffic. Whatever differences may exist between specific calls on an individual basis, however, do not rise to the level of "windfall" profits. In the context in which it is used here, the term "windfall" implies an unearned profit advantage unfairly gained by the CLECs at the expense of the ILECs. Yet, in order to conclude that such reciprocal compensation was unearned, we would have to find that the CLEC collects the funds without performing a commensurate function or service of benefit to the ILEC or its customer. Yet, on the contrary, the CLEC does perform a necessary function. If the CLEC did not terminate the ISP call, the call originator would be unable to access the ISP or to utilize its services. In the alternative, the call originator would have to find an ISP served by the ILEC and the ILEC would have to terminate the call, itself, incurring its own TELRIC in the process.

We consider below and respond to the specific arguments presented that termination cost of ISP calls is significantly less than other local calls.

3. Specific Factors Claimed to Result in Lower CLEC Costs for Terminating ISP Calls

a) Differences in Network Configurations and Facilities Used

(1) Parties' Positions

Pacific argues that while the ILEC is required to maintain a network that serves all types of customers over a wide geographic area, CLECs may pick and choose which types of customers to serve, such as ISPs. As a result, Pacific claims that CLECs can limit the number of facilities they build, and deploy lower-cost networks with less functionality than Pacific's. Pacific argues that ISPs are frequently collocated in the CLEC central offices. In those cases, argues Pacific, no CLEC loop plant is involved in transporting traffic to the ISP. Thus, Pacific believes that it is placed at an unfair competitive disadvantage by having to compensate CLECs at a rate that significantly exceeds the CLECs' true costs.

Pacific also claims that CLECs are able to lower their termination costs for ISP traffic through the use of new generation routing products that do not use a traditional voice circuit switch to deliver ISP traffic. Witness Hamilton describes this new generation equipment as an "Internet Gateway." Hamilton testified that this new technology enables CLECs to replace several pieces of equipment used in traditional switching as well as the Class 5 switch. Because this equipment is designed specifically for ISP calls, Hamilton states that it obviates the need for many of the traditional voice switching features. This technology generally cannot be used to originate traffic, but merely receives and routes traffic to an ISP. Because of the reduced functionality, Hamilton argues that it follows that these Internet Gateways will have lower costs.

Moreover, Pacific claims that the configuration of the CLECs' facilities forces Pacific to incur additional transport and switching costs in delivering ISP traffic to CLECs' points of interconnection, rather than directly to ISPs.³⁴ Pacific claims it incurs the additional costs because CLECs have generally chosen not to establish a point of interconnection in each of the local calling areas where ISPs originate calls. Pacific argues that CLECs often design their networks to have only a few points of interconnection per LATA, thus causing Pacific significant transport costs to haul traffic from the originating point to these locations. Pacific argues that the typical configuration of CLEC networks is actually adding costs to Pacific. Moreover, Pacific claims it is not equitable that when Pacific serves an ISP, it has to fund termination costs from the services ISPs buy or from other customers, while CLECs may look to Pacific to cover their costs.

Various CLECs actively participating in this proceeding provided testimony and written comments on the configuration of their facilities used to transport and terminate ISP traffic. Level 3 states that the principal architectural differences between ILEC and CLEC networks arise largely in the relative mix of the switching and transport components. ILECs generally have a hierarchical network, so that within a given geographical area, multiple end offices subtend on tandem offices. These tandem offices aggregate traffic and network management functions associated with the area served by each of the end offices subtending it. Because the ILECs have millions of subscribers statewide, they can afford to deploy relatively efficient, large-scale switching systems in close geographic proximity to their customers.

Level 3 argues that while many CLEC networks are physically configured differently than ILEC networks, they provide the same functionality for all local communications traffic, including ISP bound traffic. Pac-West witness Selwyn explained that CLEC and ILEC networks are generally comprised of three principal components: subscriber loops, end office switches, and interoffice network, which are trunking and switching facilities that provide interconnections among end offices and between end offices and other carriers. In contrast, a CLEC's customer base is only a small fraction of the size of the ILEC's customer base. As such, in lieu of using tandems and multiple end offices, CLECs typically deploy a small number of large switches which perform both tandem and end office functionalities to serve a comparable geographic area to that of the ILEC. CLECs transport their customers' traffic over relatively large distances. Because transport costs have become far less distance-sensitive with the use of high-capacity fiber optics, enormous amounts of capacity can be deployed at little more than the cost of more conventional transport capacity sizes.

ICG witness Wood acknowledges that ISPs may use a variety of facilities to connect with the serving LECs switch, but denies that the choice of facilities or methods of connection has any impact on the usage sensitive costs which are the only relevant costs recoverable through reciprocal compensation. ICG claims that that the characteristics of the particular facility used by a LEC to deliver traffic to its own customers is irrelevant to the rate for reciprocal compensation because the costs of these facilities are non-traffic sensitive, and are recoverable from end users. Moreover, regardless of what type of facilities are used to provide service for a specific type of customer, i.e., ISPs, when a CLEC uses a fully functional switch, it is purchasing the ability to service all line types. At the switch matrix level, which is the basis for costing out reciprocal compensation rate, a call path is assigned at the individual channel level, without reference to the type or capacity of the physical facility connecting the switch to the end user.

ICG's network is built upon an architecture that can generically be referred to as a SONET Ring architecture. These SONET rings are comprised of fiber optic facilities and multiplexing equipment that provides for aggregating, connecting and dispersing an individual customer's traffic to a larger SONET data stream. Witness Starkey testified that ICG employs a common network that is used to service its entire customer base. Both general business customers as well as data customers (primarily ISPs) use the same switches, fiber optic backbone and SONET rings for accessing the network, as well as for originating and terminating calls. Starkey testified that ICG employs fully functional Class 4/5 circuit-based switches (i.e., Lucent Technologies' 5ESS) that are shared by all of its local exchange customers. Focal denies that it uses the new technologies referenced by Pacific in terminating ISP traffic in California. Focal states that it utilizes Nortel DMS-500 switches in California that provide all of same call origination functionalities offered by the ILECs.³⁵

ICG argues that because of the lumpiness inherent in switching investments and the fact that CLECs began to compete without an embedded customer base, any given CLEC may experience per-minute switching costs that are actually higher than those of the ILEC, including the cost of calls delivered to ISPs. ICG denies that any relevant cost differences exist between ILECs and CLECs that would justify paying an asymmetrical rate for reciprocal compensation.

(2) Discussion

It is an uncontested fact that CLECs networks tend to be configured differently from those of the ILECs in the manner described above. The relevant inquiry, however, is whether the CLECs' network differences cause significantly lower traffic-sensitive terminating switching and transport costs of the type that are recoverable through reciprocal compensation. We conclude that while the differences in network configurations between ILECs and CLECs may result in various differences in costs, those differences generally do not relate the traffic-sensitive terminating transport and switching costs that are the subject of reciprocal compensation. Rather, they relate to the non-traffic sensitive costs that are already recovered from end users.

Based on the testimony of witnesses noted above, we find that the CLECs generally use fully functional switches that offer both originating and terminating functions, and that are used to serve all of their customers, not just ISPs. ICG witness Starkey testified that ICG is a full service provider that uses a fully functional Class 4/5 circuit-based switches that are shared by all of its local exchange customers. (Rebuttal of Starkey; Exh. 2 at page 2.) Thus, whatever cost savings the CLECs may be able to achieve in successfully managing their switching resources, there is no basis to conclude that they fail to provide complete functionality on par with that offered by the ILECs.

Likewise, the question of whether the ILECs incur higher originating transport costs as a result of differences in network configuration is separate and distinct from the question of what are CLEC's termination costs. We have designated a later phase of the proceeding for consideration of issues relating to intercarrier compensation for transport charges incurred by originating carriers based upon differences between the rating and routing points of calls. We make no final determination in this decision concerning the level of transport costs that Pacific incurs in originating and delivering local traffic to CLECs points of interconnection or what forms of intercarrier compensation may be warranted for such originating costs. We do note, however, that Pacific's originating transport cost estimates assumed the point of interconnection was always located at or near the CLEC's switch so that Pacific would be responsible for providing interconnection trunks from its switch to the CLEC's switch. (Tr. At 1534; Hamilton.) On cross-examination by Pac-West, however, Pacific witness Hamilton testified that the length of interconnection trunks assumed by Pacific was based on a data base which erroneously measured the distance from the Pacific switch to the CLEC switch, rather than to the point of interconnection. (Tr. at 1593-94; Hamilton) Pac-West witness, Mills, however testified that numerous points of interconnection in the Pacific/Pac-West interconnection agreement are not at the Pac-West switch. (Tr. At 1593-94; Mills)

In any event, Pacific's arguments concerning its costs of originating transport charges are not unique to ISP traffic, but apply equally to any and all calls transported to another carrier for termination. Every call that originates on Pacific's network that is handed off to another carrier for termination will necessarily require Pacific to transport the call to the relevant point of interconnection with the other carrier's network. In certain circumstances, Pacific may incur higher originating transport costs to hand off calls to another carrier rather than to terminate the call over its own system with no interoffice transport required and no intermediary switching operations. Yet, the distinction being drawn is not between termination of ISP traffic versus other local traffic. Rather, the distinction that Pacific draws is between the ILEC exclusively handling the origination and termination of any type of call versus a CLEC handling both the termination portion of any call. In other words, it is actually the introduction of a competitive element into the process that gives rise to Pacific's claim of higher transport costs. The question before us here, however, is not to second-guess the merits of competition, itself, not to probe how carriers choose to establish points of interconnection in a competitive setting. Instead, the focus of our inquiry here is on the functions and cost characteristics relating to the terminating end of ISP calls.

b) Longer Call Duration of ISP-Bound Traffic

(1) Parties' Positions

Pacific Witness Scholl conclude that "ISP-bound calls delivered to CLECs" are "typically much longer in duration" than a "traditional" voice call. They state that "ISP-bound traffic on Pacific's network during 1999 averaged 29 minutes in duration." By comparison, Pacific reports that the average duration of a local voice call originated by flat-rate residential service in 1994 (prior to the growth of ISP traffic) was 3.78 minutes.³⁶

In 1994, Pacific reported average local usage per flat-rate residential line of 19 minutes per day. By comparison, America Online (AOL), a major ISP, reported 52 minutes of usage per day by its customers for Internet access. Pacific argues that these comparisons highlight the difference between ISP and other types of calls.

Verizon witness Beauvais testified that ISP-bound calls average between 20 to 30 minutes per call.³⁷ Some of the data supporting this duration were collected from trunks devoted solely to ISPs and thus, include no traffic delivered to local plain old telephone services (POTS) customers.³⁸ Other data come from a study that involved specifically identifying ISP telephone numbers and verifying the traffic as modem traffic by calling the identified numbers.³⁹ Similar studies performed by Pacific and Roseville Telephone Company (Roseville) show average call duration times for ISP-bound calls of 29 and 25 minutes per call.⁴⁰

Beauvais testified that there are a variety of other sources that reflect average holding times for ISP-bound traffic that are 30 minutes or greater.⁴¹ An independent entity - Nielsen/NetRatings -calculated an average ISP-bound call duration of 30 minutes and 27 seconds.⁴² Traffic data that a CLEC provided to Verizon in Michigan - and that the relevant CLEC confirmed was purely ISP-bound traffic - displayed an average holding time of 42 minutes per call.⁴³

Various CLECs have argued longer hold times are not unique to ISP calls, in that particular local voice calls or types of call traffic - for example, traffic in households with adolescents who make long calls to their friends - potentially overlap with the hold times of ISP-bound calls.⁴⁴ The ILECs witnesses consistently use the term "voice calls" to describe calls that are not delivered to ISPs. The CLECs claim this is a false dichotomy. A significant volume of non-voice (data) calls exists that is unrelated to, and does not involve, ISPs, (i.e., some calls that are not voice calls are also not ISP-bound calls.) Conversely, not all calls to ISPs are data calls, some are voice calls. As a result, while it may be meaningful to refer to "voice" vs. "data" calls, it is not accurate or appropriate to place all calls to ISPs into either classification. Verizon responds, however, that the arbitrage opportunity presented to CLECs under the existing regulatory regime arises in part from the difference in the average duration of ISP-bound calls in the aggregate as compared to the average duration of voice calls. The longer average duration for ISP-bound calls reduces the per-minute cost of the "call set-up" i.e., the costs that occur on a per-call basis, but do not vary with the length of the call. Since Verizon's reciprocal compensation rate does not separate out this fixed component, Verizon must pay CLECs for every additional minute even though there is no corresponding increase in per-minute costs. Because of the longer duration of ISP calls, Verizon claims that CLECs receive at least five times more for the call set up allowance than the fixed cost of performing the service would merit. Verizon understands that Pacific's default reciprocal compensation rates separate out the costs between a flat per-call set-up fee and a per-minute charge, and thus, apparently the overpayment due to call duration is an issue unique to Verizon.

The CLECs claim that Pacific and Verizon do not know the average duration of an ISP-bound call (whether delivered by a CLEC or Pacific). As described in response to ICG's Data request No. 18, Pacific has attempted to estimate the number and characteristics of calls to ISPs (delivered on its own network or by a CLEC). ICG argues that the process used has been thoroughly discredited in other states as being over-broad and inaccurate.

The CLECs note that Verizon's Beauvais cites to data that "is rather dated and is not California-specific,"⁴⁵ yet he concludes that the average holding time for voice calls is approximately 4.8 to 4.9 minutes in Illinois. He then compares this to data from a single CLEC in Michigan, and concludes that the average holding time for calls to ISPs is approximately 42 minutes, though "limited data" that Verizon has collected for California suggests an average duration for JSP calls of 20 to 30 minutes.

(2) Discussion

We find that the data studies presented by the ILECs concerning call duration estimates produce rather wide variations, and fail to provide any estimates that would be useful for quantifying a specific difference in the cost of an average ISP call versus all other local calls. We have already discussed various defects in the ILECs' studies of call duration previously in Section VI.B. While we acknowledge that the ILECs' call duration studies do not provide precise measures of ISP call duration in relation to that of all other local calls, we do find that on average, ISP call duration appears to measurably exceed the average duration for other local voice calls. Yet, by limiting the comparison only to voice calls, the ILECs fail to take into account the effects of other categories of non-ISP calls that also may have longer-than-average durations compared with voice calls. It would be arbitrary to single out ISP calls as having a longer duration without noting that various other categories of local calls also can have similarly long durations in comparison to all voice calls.

In any case to the extent that the per-minute duration of ISP calls exceeds that of voice calls, we find that the extra duration of ISP calls does not overcompensate CLECs, at least with respect to Pacific's payments. Pacific's reciprocal compensation payments incorporate its adopted TELRICs for switching which separate out a "per-call" charge that does not increase based on the duration of a call. Thus, CLECs only receive per-minute compensation for the TELRIC component that varies with minutes of use. Pacific witness Scholl acknowledged that once an ISP call is established, the unit cost per additional minute is typically going to be a constant amount. (Tr. 1074-1075). Therefore, any increase in reciprocal compensation revenues for longer duration calls would be offset by a corresponding increase in variable costs incurred by the CLEC for each additional minute.

In the case of Verizon, however, no fixed cost set up component is segregated out of its UNE rates. Therefore, we agree with Verizon that CLECs recover additional reciprocal compensation revenues related to longer-than-average duration ISP calls that exceeds the additional costs incurred by the CLEC. We conclude, however, that the proper remedy to this disparity is not to eliminate reciprocal compensation, but rather to properly design Verizon's reciprocal compensation rates to separate out the non-variable UNE component, as already reflected in Pacific's rates. The OANAD proceeding is the proper forum to implement this rate realignment.

c) Higher Call Completion Ratio

(1) Parties' Positions

Pacific also identifies a higher call completion ratio for ISP as opposed to other calls. Pacific reports that normal voice calls are answered about 75% of the time whereas Internet calls are answered 95% of the time since they are answered by machines. Pacific claims the higher completion ratio reduced the TELRIC of the terminating switch set-up per completed call. Although TELRICs for terminating switching set-up costs are incurred on a per-attempt basis, they are billed only on completed calls. Thus, Pacific calculated a conversion factor to adjust its TELRIC based upon the higher call completion ratio for ISP calls. Using the Commission-adopted TELRIC for terminating switching set up of $0.007 per call, Pacific calculated a 14% minimum rate reduction comparing local voice calls versus ISP-bound calls.

Focal's witness argues that high call completion rates are not necessarily limited to ISP calls, but would apply to any business where a prompt answer of the call is important. Focal argues that because ISP call completion rates are not unique in comparison to completion rates for various other service-oriented businesses, there is no basis to conclude that the ISP call termination costs are unique in this respect.

(2) Discussion

The relative rate of call completion for ISP calls only has meaning in comparison to other specified categories of calls. While the ILECs limit the comparison to residential voice calls only, there are various other categories of local calls besides residential voice calls where higher-than-average call completion ratios similar in nature to ISP calls exist. Businesses that are not service-oriented and residences on the other hand, could reasonably be expected to have lower call completion rates than for ISP calls. This dispute essentially gets back to the basic question of whether it is appropriate to single out ISP calls for separate cost measurement without doing the same for other types of non-residential local calls that may deviate from residential voice call characteristics. As we concluded above, it would produce arbitrary and discriminatory results to single out ISP calls for disaggregating measurement of call completion ratios while ignoring other calls with similar call completion characteristics. Thus, while Pacific's mathematical calculation appears correct indicating a 14% reduction in call set-up costs as a result of the difference in call completion ratios, we still find that this cost differential is not unique to ISP calls. It could apply to various other types of local calls with high call completion ratios.

d) Lower CLEC Switching Costs Due to Use of Trunk-to-Trunk Switching for ISP Calls

(1) Parties' Positions

CLECs typically use high volume Integrated Services Digital Network-Primary Rate Interface (ISDN-PRI) technology to deliver ISP traffic. ISDN-PRI is a digital technology that provides 24 channels of capacity to an end-user customer. An ISDN-PRI line is typically configured with 23 bearer channels that are used to transmit traffic, and one data channel that is used for signaling. The technology is designed to serve the needs of high-volume customers, such as ISPs.

Pacific claims that CLECs' use of these digital facilities to terminate ISP calls require trunk-to-trunk switching that is different from both a technical and routing perspective, and is less costly, than the trunk-to-line terminating end-office switching used for terminating normal local voice traffic. Equipment vendors have developed switches designed for this type of termination. Pacific claims these switches do not perform all the same functions as a traditional voice switch. Witnesses Hamilton (Exh. 123; pp. 7, 21-23), Scholl (Exh. 106; pp. 16, 19-22), and Harris (Exh. 146; pp. 23-24) on behalf of Pacific, and Jones on behalf of Verizon argue that CLEC's experience lower call termination costs as a direct result of the fact that CLECs are delivering a high volume of traffic to ISPs.

Pacific estimated a TELRIC-based CLEC trunk-to-trunk terminating switching set-up price for ISP-bound traffic by applying the ISP-bound traffic completion rate to the price for a tandem switching setup attempt. The tandem switching function is a trunk-to-trunk switching function that Pacific claims is a reasonable surrogate because it reflects a similar terminating function as that performed by CLECs for ISP-bound traffic.

Verizon claims that trunk-to-trunk switching involves the use of different hardware and software to complete the call, as compared with trunk-to-line switching. For example, a switch used to terminate a trunk-to-line call to a POTS ("plain old telephone service") customer has one line card for each POTS customer served. By contrast, Verizon claims a switch delivering a trunk-to-trunk call to an ISP would not use line cards at all, but would use trunk cards carrying much higher traffic volumes per card.

While Verizon recognizes that the line card/trunk card distinction does not directly affect traffic-sensitive costs, Verizon claims an indirect effect exists insofar as other equipment-based differences are triggered. Specifically, witness Jones testified that the number of switch modules varies directly with the number of line cards or trunk cards, and that the switch modules have some usage-sensitive characteristics that impact reciprocal compensation costs.

The CLECs dispute the ILECs' claims concerning the lower costs of trunk-to-trunk switching. Pac-West argues that the claims of the lower cost of trunk-to-trunk switching are tenuous at best, and in any event, aren't relevant to reciprocal compensation since they are non-traffic sensitive. ICG claims the switching costs of call termination incurred by CLECs is not a function of the type of the customers served. ICG claims it incurs a cost of end office switching that does not vary depending on the identity of the called party whether an ISP or not. ICG disputes the ILECs' claims that trunk-to-trunk switching is less costly than trunk-to-line switching. ICG witness Starkey claims that: the ILECs' "trunk-to-trunk" switching arguments are fundamentally flawed because they depend on cost concepts that are not consistent with a proper TELRIC study. Starkey testified that: "[A]ttempting to derive disparate per-minute-of-use rates for different types of traffic originated by or delivered to a specific subset of customers has no causal validity."⁴⁶

(2) Discussion

We find that the use of trunk-to-trunk switching is not uniquely linked to ISP-bound traffic, but may be used for other forms of local traffic as well. Therefore, any cost savings inferences that could be drawn about the use of trunk-to-trunk switching would not be unique to ISPs, but could also apply to other kinds of local traffic. Moreover, the ILECs failed to establish that the use of ISDN-PRI facilities necessarily entails trunk-to-trunk switching.

Pacific witness Hamilton testified that it is impossible to determine whether a CLEC is performing trunk-to-trunk or trunk-to-line switching based merely on the fact that ISDN PRI facilities are used to connect an ISP or other customer to the central office.⁴⁷ Hamilton testified that trunk-to-line switching occurs "[w]henever the trunk needs to identify a particular line in order to activate the set of steps it needs to take to connect that path."⁴⁸ In contrast, trunk-to-trunk switching occurs "[w]hen [the switch] can ignore the digits beyond the prefix . . . because it's sending [traffic] out in bulk."⁴⁹ Hamilton further acknowledges that an ISDN-PRI customer may choose to have a particular telephone number assigned to a particular PRI channel, in which case trunk-to-line switching would occur.⁵⁰ Alternatively, the customer may choose not to have a particular telephone number assigned to a particular PRI channel, in which case trunk-to-trunk switching would occur.⁵¹ Hamilton also concedes that the same switch could perform trunk-to-line as well as trunk-to-trunk switching.⁵² Consequently, it is impossible to determine whether a CLEC switch is performing trunk-to-line switching or trunk-to-trunk switching or both unless one has examined each particular switch and identified how ISDN-PRI facilities are configured.⁵³

We also find the testimony of Pacific's witness Hamilton unpersuasive in claiming that CLECs incur lower traffic-sensitive switching costs as a result of trunk-to-trunk switching. On cross-examination, Hamilton admitted he did not know what a traffic-sensitive versus non-traffic sensitive cost is.⁵⁴ Pacific's counsel stipulated that Hamilton's testimony only addressed differences in network functions, but made no representation concerning traffic-sensitive versus non-traffic sensitive costs.⁵⁵ Yet, since only traffic-sensitive cost reductions are relevant to reciprocal compensation, we can draw no inferences from Hamilton's testimony concerning lower CLEC switching costs as they relate to reciprocal compensation.

As ICG witness Starkey testified, traffic-sensitive switch processor and switching fabric costs are incurred for inbound calls terminated to ISDN-PRI trunks. Whether traffic is provided over a trunk facility or a line facility, the job of the switch in terms of mapping calls to their predetermined destination points remains the same. While certain specific switch components may differ between trunk and line switching, the two primary traffic-sensitive cost drivers within a switch (i.e., capacity--switch fabric costs measured in time slot availability, and processing time, measured in milliseconds) remains the same.⁵⁶

Verizon witness conceded that the per-call set up cost for the PRI configuration used for ISP traffic is actually slightly higher because more processing time is required, but argues that the higher cost is outweighed by the longer holding times associated with ISP traffic.⁵⁷ Yet, as we have noted above, the proper way to correct for the longer holding times is to disaggregate the reciprocal compensation rate to allow for separate fixed charges for set up costs. By making this correction in the rate, there would be no overstatement of costs due to long holding times. On this basis, Verizon's testimony supports a finding that traffic-sensitive set-up costs are actually higher for ISP traffic. In any event, the ISP call set-up charge is no less than for voice traffic.

We find the testimony of ICG witness Starkey to be persuasive that ICG purchases and deploys fully functional Lucent 5ESS switches that support multiple switching architectures including line-to-line, line-to-trunk, and trunk-to-trunk. Starkey's testimony is based on discussions with actual ICG switch engineering personnel.⁵⁸ We likewise have no basis to conclude that Pac-West's switches serving ISPs are limited only to trunk-to-trunk capabilities. Verizon witness Jones made no inquiries and did not know either the costs or configuration of Pac-West's facilities.⁵⁹

Witness Wood testified that, in any event, trunk-to-trunk switching costs are non-traffic sensitive and thus, are not part of the costs subject to reciprocal compensation recovery. Witness Starkey also claims the ILECs confuse traffic-sensitive with nontraffic-sensitive costs.

In conclusion, we find no basis to conclude that ISP traffic-sensitive termination costs are uniquely lower due to the use of trunk-to-trunk switching.

e) Lack of Line Concentration using ISDN-PRI

(1) Parties Positions

Verizon argues that CLECs' use of ISDN-PRI technology results in a difference in line concentration accommodated by the switch for ISP calls compared to voice-grade calls. For local traffic terminated to a customer over a standard voice-grade line port, there is typically a line concentration ratio of six to one. This means that the number of standard local POTS lines coming into the switch will be six times greater than the available number of paths through the switch for such traffic. Rather than having a dedicated amount of capacity through the switch, the lines share the switch path capacity at a ratio of six lines to one path.

This six-to-one line concentration configuration works well for standard POTS traffic because each POTS line is generally only used for short periods of time, and all lines are not typically in use at the same time. Because the volume of traffic over each POTS line is relatively low, the lines can efficiently share paths through the switch without substantial amounts of call blocking (a call is blocked when it does not make it through terminating switch because there is no available path). For calls to POTS customers, the switch module - a piece of peripheral equipment that is part of the switch - performs the line concentration function. This function allows the larger number of end-user lines to share the smaller number of paths through the switch. Because the paths through the switch are shared among multiple lines, the use of the switch during the peak hour imposes congestion costs on other line-concentrated users in the form of call blocking or rationing. Call blocking or rationing occurs as a result of the available path being in use. These congestion costs are the busy hour line costs, measured in cenum call seconds (CCS).

As explained by Verizon witness Collins, the busy hour line CCS provides a measure of costs that are caused by the line concentration accommodated in the switch. Because these busy hour line CCS congestion costs vary based on the volume of the traffic flowing through the available shared switch pathways at a given time, costing models treat these termination costs as traffic-sensitive.⁶⁰ Therefore, where reciprocal compensation applies, such costs are included in any reciprocal compensation charge that the originating carrier must pay to the terminating carrier on a per-minute or per-call basis.

For calls to ISPs over ISDN-PRI connections, Verizon argues however, the situation is very different. When ISP-bound traffic is carried over higher-volume ISDN-PRI trunks, the ratio of incoming to outgoing traffic through the switch is one-to-one. That is, each incoming line (or trunk in the case of traffic that has already gone through a separate originating switch) has dedicated capacity (i.e., a guaranteed path) through the switch. Unlike local POTS traffic, the incoming line does not share that capacity with other traffic. This arrangement is used for ISP-bound traffic because such traffic tends to be higher in volume.

Verizon argues that because the switch reserves dedicated capacity for the traffic that flows over that connection, there is no line concentration and no competition with other non-dedicated traffic for available pathways through the switch. Verizon claims because the ISP that is receiving the call over a PRI connection does not have to compete with other customers for switch capacity, the ISP imposes no congestion costs on the switch as a result of the amount of traffic that is carried to it over the connection.⁶¹ From the perspective of the terminating carrier, it does not matter how frequently the ISP is constantly using its dedicated capacity. In any case, the number of pathways through the switch available to other customers remains the same.

Verizon claims that this difference in the manner in which the switch paths are allocated reduces the traffic-sensitive costs incurred by the terminating carrier. That is because the level at which the ISP uses its dedicated capacity - i.e., the amount of traffic received by the ISP through the switch - does not affect the congestion in the switch. As a result, Verizon claims the CLEC does not incur traffic-sensitive busy hour line CCS costs when it terminates ISP-bound traffic.⁶² Since only traffic-sensitive termination costs are eligible for recovery by the terminating carrier, the line CCS costs that have been included in the reciprocal compensation rate for line-concentration, Verizon claims that standard voice traffic must be removed from the rate when ISP-bound traffic is at issue.⁶³

Verizon argues that the lack of line concentration performed by the switch on ISP-bound traffic delivered using ISDN-PRI technology results in significantly lower traffic-sensitive switching costs being incurred by the CLEC for termination of traffic.

The CLECs dispute Verizon's claims. Focal witness TerKeurst denies that the switching of a dial-up call onto an ISDN-PRI connection is any less expensive than switching a voice call onto a separate voice circuit. TerKeurst testified that many customers utilize ISDN connections for voice traffic without conversion to analog signals, so that this aspect of the switching process is not unique to ISP-bound traffic.⁶⁴

CISPA witness Montgomery testified that, if anything, the ISDN-PRI service used to terminate some ISP-bound traffic is actually more costly, when compared to the termination cost for analog traffic. Montgomery attributes this to two factors: (i) providing ISDN-PRI service requires that additional software be activated in the central switch processor; (ii) that the functionality of ISDN service is more taxing to the central switch processor.⁶⁵

As Focal witness Terkeurst similarly noted the fact that circuits are concentrated within a single ISDN-PRI loop does not reduce the switching requirements. ISDN-PRI connections and the ability to switch digital traffic onto such connections without conversion to analog are not unique to ISP-bound traffic, but are available to any business customer wishing to purchase them. In fact, a number of incumbent LEC business customers purchase such connections.⁶⁶

ICG witness Starkey similarly testified that Verizon's claims of lower cost due to line concentration differences were unfounded and reflected a misperception of the manner in which traffic-sensitive costs are incurred.

(2) Discussion

We find no basis to conclude that CLECs avoid traffic-sensitive switching costs merely because of "dedicated" capacity assigned to ISP incoming calls. While we recognize that ISP call termination may be provisioned over ISDN PRI circuits which utilize higher line concentration than voice traffic, we find that the use of such circuits is not unique to ISPs. Moreover, we find no basis to distinguish the use of switching resources used by ISDN circuits from other circuits.

ICG witness Starkey testified that even though ISDN circuits may be provisioned with 1:1 concentration ratios, they share the same finite switching resources (i.e., internal transport links, the switch fabric and the processor), as do other circuits. ISDN circuits are allocated switching resources as calls are made, regardless of the concentration ratio to which they've been engineered. The only difference between an ISDN circuit engineered with a 1:1 concentration ratio versus a more concentrated circuit is the level of priority in the process of allocating switching resources in "real-time." While this may impact which circuits experience "blocking" (i.e., no time slots available), these switched services still consume usage sensitive resources. Starkey's testimony indicated that the switch's processor actually requires more time to process a call delivered via ISDN compared with other types of more traditional traffic.⁶⁷ Based on this testimony, we thus conclude that ISDN-PRI services, regardless of concentration ratio, use traffic-sensitive switch resources (i.e., internal transport links, timeslot management resources and switch processing time), and incur related costs. Therefore, based on the testimony of ICG, Focal, and CISPA as noted above, we find insufficient basis to accept Verizon's claim that CLECs incur lower traffic-sensitive termination costs as a result of line concentration differences that apply only to ISP traffic.

D. Does the Payment of ISP Reciprocal Compensation Result in Unrecoverable Losses to the ILECs?

1. Parties' Positions

Pacific claims that its current retail rate structure precludes recovery of ISP reciprocal compensation payments from its own end use customers at least for those that are billed a fixed monthly rate with unlimited local calling. Pacific argues that the vast majority of its customers calling the Internet have flat rate (1 FR ) service, and do not generate any additional revenue to cover the per-minute of use charges paid by Pacific for ISP reciprocal compensation. Pacific claims the price for flat rate residential service does not cover the cost of the access line, much less the additional costs generated by usage-sensitive reciprocal compensation payments.

Witness Jacobsen testified that when the average Internet user uses a dial-up connection for an hour a day (just over the average usage reported by AOL), the LEC originating calls for that customer must pay about $3.79 per month in reciprocal compensation payments.⁶⁸ However, Pacific argues, the Commission has set the price of flat-rate residential service below either the direct embedded cost or incremental cost of the line.⁶⁹ In fact, the Commission set the price of residential flat service (1FR) at only one half of the fully allocated cost less the End User Common Line ("EUCL") charge.⁷⁰ Thus, Pacific claims that its 1FR service is priced below its forward-looking costs, even without any usage.⁷¹

While most customers use residential flat-rate service for dial-up access to the Internet,⁷² Pacific claims it has receives no additional revenue from 1FR service if that customer uses the service for Internet traffic. Pacific argues that an increase in basic service prices to cover reciprocal compensation payments would unfairly shift the burden of these payments to all customers, whether or not they access the Internet. Pacific claims a rate increase of $0.60 per month would be required to fund ISP reciprocal compensation payments to CLECs in the year 2000, increasing to $1.80 per month in the year 2002 based on its assumed growth rates.

Pacific's witness Jacobsen reports that Pacific paid $173 million in ISP-related reciprocal compensation to CLECs during 1996 through 1999. In the year 2000, Jacobsen reports a drop in such payments to $135 million. Yet, by the year 2002, Jacobsen projected a growth in payments to $450 million, based on an assumed compounded growth rate of 5% per month.⁷³

Verizon likewise claims that it has been incurring massive net losses as a result of the Commission's reciprocal compensation rules. Based upon on billing records for the period of November 1, 1998 through May 24, 2000, CLECs have billed Verizon approximately $32.4 million for reciprocal compensation while Verizon has billed the same CLECs only about $0.4 million.⁷⁴ Based upon the average hold times for the traffic flowing in each direction, Verizon estimates that approximately $27.1 million of the payments it made to CLECs were for ISP-bound traffic while less than $200,000 of the payments CLECs made to Verizon were for ISP-bound traffic.⁷⁵ As a result Verizon claims a net loss over that time period of approximately $27 million.⁷⁶

Verizon claims it cannot recover its reciprocal compensation costs attributable to ISP-bound traffic from the flat rate it charges to the typical residential end-user. Verizon is presently allowed to charge flat rate, one-party residential customers--the customers most likely to access the Internet--$17.25 per month. Verizon claims the reciprocal compensation for ISP traffic can easily consume half of the total monthly revenue from the end user. After deducting the other costs that must be recovered from the end user revenues, Verizon argues that it cannot recover its payments for ISP reciprocal compensation.

Roseville also claims that it will suffer significant financial hardship from the payment of ISP-related reciprocal compensation. Roseville reports it has 5,400 trunks connected with CLECs of which 99% of the traffic is ISP bound. Roseville estimates that its reciprocal compensation payments to CLECs in 1999 would have been approximately $1.2 million if it had to pay all CLECs with which it is interconnected based on the rate of $.002 per minute. Roseville projects growth in this amount to $2 million in 2000 and over $2.6 million in 2001. For a company with 1999 intrastate revenues of only about $94 million, Roseville argues that these amounts are significant. On the other hand, Roseville projects receipts of reciprocal compensation from CLECs of only $11,000, $19,000, and $25,000 for 1999, 2000, and 2001, respectively. While Roseville's monthly service charge is only $18.90, Roseville argues that it would have to pay reciprocal compensation of $21.60 for a customer that accessed the Internet for six hours per day. Roseville also reports it has spent $6.2 million to upgrade its central offices to accommodate the volume of ISP traffic.

The CLECs, CISPA, and TURN discount ILEC claims that they are losing money as a result of ISP reciprocal compensation payments. The CLECs claim that the ILECs derive substantial additional revenue from end user customers placing calls to ISPs through the offering of services such as Caller ID, Call Waiting, and Call Forwarding. ICG argues that Pacific's argument that it is losing money on residential customers is contradicted by its own behavior in providing customers a monetary incentive not to disconnect additional lines.

ICG also argues that the rate of growth in ISP terminated minutes will substantially abate due to (1) growth in DSL lines which are not subject to reciprocal compensation and which are most likely to be ordered by those customers with the heaviest Internet access. Likewise, ICG points to Pacific's and Verizon's aggressive deployment of other service alternatives to ISP dial-up access (such as dial-to-frame, virtual point of presence, and CyberPOP services) that will reduce the volume of dial-up ISP usage. Further, ICG notes that ILECs which own ISP affiliates have the capability to compete for increasingly larger shares of Internet business.

TURN likewise argues that far from being a financial drain to ILECs, the Internet provides enormous potential for the ILECs to tap vast new sources of revenue. Pacific, Verizon and Roseville, either directly or through their affiliates, are all actively marketing Internet service to ISPs and to end user customers.⁷⁷ Pacific's affiliate, Pacific Bell Internet Service, purchases services from Pacific. From 1996 through 1999, the yearly revenues of Pacific Bell Internet Service have grown almost ninefold and the number of subscribers has grown almost fivefold.⁷⁸ Pacific's parent SBC is engaged in a comprehensive refurbishment of its network to facilitate the provision of Internet and broadband services, and to achieve a substantial share of broadband market penetration.⁷⁹ In addition to pursuing significant broadband market share, SBC/Pacific has "conservatively targeted . . . annual savings of about 1.5 billion -- 850 million in cash operating expenses, and 600 million in capital expenditures by 2004" from its broadband initiative.⁸⁰

The CLECs also dispute the ILECs' cost shortfall claims by arguing that the ILECs would incur the costs of terminating the ISP calls themselves if CLECs did not terminate it. The CLECs argue that the payment of reciprocal compensation is equitable because the ILEC thereby avoids the cost of terminating ISP traffic. Since the TELRIC cost is the same whether the ILEC or the CLEC terminates the call, the CLECs claim the ILEC should be indifferent as to whether termination is done by a CLEC or an ILEC.

The ILECs respond by arguing that they incur additional transport expense when delivering traffic to CLECs, as compared to keeping all traffic on their own network. Specifically, Pacific witness Hamilton asserts that Pacific Bell bears the vast majority of the network burden to support ISP-bound calls, since competitive LECs may have only one or two points of interconnection in a LATA and Pacific must transport a call a significant distance before handing it off to the competitive LEC. Pacific witness Scholl testified similarly and contends, as a result, that ISP-bound calls delivered to competitive LECs cost Pacific more, not less, to deliver than it would cost Pacific to deliver ISP-bound local traffic to ISPs on its own network.⁸¹

Focal argues that this argument does not apply to its own network. For example, as indicated by Focal witness Tatak, Focal has at least 45 physical points of interconnection with Pacific in the two LATAs in California where it operates.⁸² However, even in the case of a CLEC which has fewer points of interconnection, the ILECs offer no evidence that there is something peculiar to ISP traffic that causes a disproportionate burden on the ILECs' transport burden.

Focal further argues that the average per minute rate paid out by ILECs has been decreasing since 1996, and will continue to decrease due to the re-negotiation of interconnection agreements. Focal claims Pacific's two-part compensation rate ensures that the originating carrier only pays for service they are receiving from the terminating carrier, with no over-recovery. Pacific responds that the growth of the Internet market has created a growing burden of payments that is unsustainable at any compensation rate.

2. Discussion

We recognize that the ILECs incur significant costs to make reciprocal compensation payments to CLECs. Yet, even to the extent that some losses arguably might accrue to an ILEC as a result of paying reciprocal compensation for ISP traffic, that fact would not, in itself, justify the ILEC withholding reciprocal compensation payments otherwise due for services performed. Financial loss is not a valid basis for any carrier to justify withholding payment for any services performed for its benefit by a third party. Moreover, a fundamental principle underlying the New Regulatory Framework (NRF) established by this Commission was that ILECs were to bear financial responsibility for the business risks that future events would not turn out exactly as expected or as wished. In return, the ILECs gained new opportunities to enhance investor earnings by pursuing new business ventures with profit potential. Pacific's Project Pronto is but one example of such a potential opportunity.

Project Pronto (as described in Exhibit 50, SBC's Investor Briefing) is a $6 billion "Sweeping Broadband Initiative" investment program.⁸³ Project Pronto promises, for SBC as a whole, "annual savings of $1.5 billion by 2004," "capital and expense savings [that] pay for [the] initiative on [a net present value ("NPV")] basis, "$3.5 billion in new revenue by 2004," a "100 basis-point improvement in annual revenue growth," and "significant value creation, well in excess of $10 billion NPV."⁸⁴ The Investor Briefing states, "SBC's new network investments will have a profound impact on its cost structure; in fact, the efficiencies SBC expects to gain will pay for the cost of the deployment on a NPV basis. These efficiencies are conservatively targeted to yield annual savings of about $1.5 billion by 2004 ($850 million in cash operating expense and $600 million in capital expenditures)."⁸⁵ Pacific's witness Mr. Jacobsen stated that he saw "Project Pronto as an attempt to live out th[e] new [NRF] framework."⁸⁶

Pacific argues that Project Pronto is not relevant in determining whether Pacific has sufficient revenues to fund ISP-related reciprocal compensation. Witness Jacobsen testified that, "It would be inappropriate [for Pacific] to make decisions now based on cost savings that we're going to hopefully reap in the future. . . . It would be very premature for the Commission to say, Gee, if these materialize, you might be in a position to fund a windfall to your competitors."⁸⁷ Jacobsen also testified that the statements in the Investor Briefing "are projections based upon a lot of assumptions and a lot of hopes. I don't think you can say for sure that these are going to come to pass."⁸⁸ Pacific witness Hamilton clarified that the voice over ATM (VoATM), also known as "voice trunking over ATM" ("VTOA") portion of Project Pronto was currently on hold for several reasons.⁸⁹

We recognize there are business uncertainties associated with Project Pronto as testified to by witness Jacobsen. Yet, irrespective of any specific benefits that may ultimately be realized from Project Pronto, the relevant issue is that Project Pronto represents an example of the structure of risk and reward incentives under NRF. The presence of reciprocal compensation gives Pacific and SBC an incentive to achieve as many cost savings and efficiencies as possible through ventures such as Project Pronto.

Moreover, Exhibit 164 indicates, in Verizon's response to ICG's Data Request No. 13, that GTEC is currently generating $55.5 million from advanced technology products and services that GTEC has developed or deployed to serve ISPs. See also, Exhibit 167 (showing Verizon's 1999 revenues of $16.8 million from local exchange dial tone access line services ($35.39/line) sold to ISPs). None of this revenue, of course, existed at the time of the IRD decision in 1994, and none of the incremental profits associated with the sale of these Category II services will cause any adjustment of the rates that GTEC charges for its other services. In claiming that they have no sources of revenue to offset reciprocal compensation payments to CLECs (see, e.g., Exh. 15, p. 24, ll. 7-9 (Mr. Jacobsen for Pacific)), the ILECs fail to recognize the potential for such new revenue sources.

The reciprocal compensation obligation thus provides incentives for the ILECs to seek to win over ISP customers and aggressively market alternatives to dial-up access to ISPs. By doing so, the ILECs can minimize their reciprocal obligation burden by migrating customers off dial-up access.

The fact that CLECs have been more aggressive in marketing their services to ISPs and have achieved a much greater than anticipated share of the ISP market does not justify insulating the ILECs against the risk resulting from such an unexpected outcome. Notwithstanding our misgivings with the underlying premise of this particular line of argument, we find no substantive basis in the ILEC's claims of significant financial loss due to payment of reciprocal compensation to CLECs.

The beginning point for Pacific's argument is that 1FR revenues already fail to recover costs even before consideration of reciprocal compensation revenue. We find that Pacific's narrow focus only on 1FR revenues to the exclusion of other revenue sources runs contrary to the stated intent of the Implementation Rate Design (IRD) proceeding which took into account that various revenue sources would provide differing levels of profit contribution. Pacific witness Scholl conceded that IRD intended that toll services priced above cost were to be recognized as an offsetting contribution to cover any shortfalls in 1FR cost recovery.⁹⁰

Moreover, we find no basis to conclude that an incremental call duration of ISP calls results in a significant financial harm to Pacific. Dr. Selwyn testified that the "ILECs' existing retail local exchange tariffs are generally set at sufficiently high levels to compensate for most, if indeed not all, of the ILEC's usage costs associated with local dial-up calls to ISPs."⁹¹ Dr. Selwyn provided an example of a California subscriber to America On-Line's ("AOL") Internet service who connects to AOL through a second residential exchange line obtained from Pacific. Assuming the subscriber's AOL usage is 64 minutes a day (i.e., 32 hours per month) and average per-call duration is 30 minutes, Selwyn calculated the total incremental costs of the associated local telephone usage. Selwyn used Pacific's most recently approved TSLRICs for local usage. Selwyn further compared such costs to the total local usage component implicit in Pacific's residential flat-rate charge and concluded that Pacific's local usage rate component "more than compensates Pacific for the incremental costs of that customer's dial-up ISP calls."⁹²

When the Commission last conducted its comprehensive review of Pacific's local residential rates in its 1994 IRD proceeding, it purposely set such rates at one-half of Pacific's reported fully allocated costs. Such rates, however, consisting of direct embedded, or historical, costs plus an allocation of common overhead costs⁹³ are unlikely to be below Pacific's long-run incremental costs of carrying local traffic based on forward-looking technologies. Thus, Pacific's arguments fail to provide a cost basis against which to evaluate whether current revenues recover its forward looking costs of providing 1FR service today.

None of the ILECs' data showing growth in Internet usage prove that such usage has significantly affected their cost assumptions made when they established their residential rates. As Dr. Selwyn testified, available FCC data "demonstrate[s] that the Internet has had a significant impact upon the demand for additional residential access lines, but has had little impact upon the average volume of local traffic carried over each line."⁹⁴ Beginning in 1990, "the demand for additional residential access lines began to mushroom, and by the end of 1998 . . . over one-fifth of all U.S. households had an additional residence line, representing some 20.4-million such lines nationwide."⁹⁵ "During that same period, the per-line volume of local calling increased by only 19 percent."⁹⁶

Pacific's own data, shows that Internet usage has only resulted in an average duration increase for all of its local traffic from 2.25 minutes to 2.47 minutes in 1999, an increase of only 0.22 minute. The average overall duration of 2.47 minutes for local calls, including ISP calls, is still less than the 3 minutes which Pacific acknowledges is the average local call duration assumption underlying Pacific's local service rates set in the 1994 IRD proceeding.⁹⁷ We also find no basis to support Pacific's estimates of growth in the rate of its payments for ISP-related reciprocal compensation through the year 2002. In fact, through its own marketing efforts to promote alternative Internet access services such as Digital Subscriber Line (DSL), Pacific has the potential to actually reduce the magnitude of such reciprocal compensation payments to CLECs going forward.

Roseville raises the additional issue of its financial burden for new plant upgrades to accommodate interconnection with CLECs. While the $6.2 million investment reported by Roseville may be significant in magnitude, such costs are byproduct of CLEC interconnection generally, and are not uniquely limited only to ISP-bound traffic. Thus, the fact that Roseville may incur network upgrades as part of its general obligation to facilitate competition in its service territory is not a reason to withhold reciprocal compensation payments for the termination of traffic by CLECs, including that of ISPs. Whatever means may be appropriate for Roseville to recover its plant upgrade costs, the elimination of its reciprocal compensation obligations is not the proper remedy.

In any event, any claims of ILECs regarding the need to raise end user retail rates to fund reciprocal compensation payments are beyond the scope of this phase of the proceeding. Thus, the record provides no basis to conclude that the impacts of Internet usage has adversely affected the ILECs' overall financial health.

⁴⁷ Tr. at 1543-44 (Pacific Witness Hamilton).

⁵¹ Id.

⁵² Id. at 1590.

⁵³ Id.