DEVELOPMENT OF
CORRECTED SWITCH
COSTS
5.1 Outdated, unrealistically high switch costs overstate line costs
How the BCM reflects switch costs
The BCM "deploys" switches to serve all CBGs associated with a given wire center. The cost allocation of this theoretical deployment is accomplished by taking the common fixed cost of a digital switch (which, in the BCM, is $647,526) and allocating 79%[1] of these common fixed costs to all lines (the total of actual households grossed up for (1) business lines and (2) an 80% fill factor) served by the wire center. Common costs are the central processor frames, billing and data recording equipment, power equipment and backup power, main distribution frame, testing frames, and basic software.[2] In addition, a variable switch cost of $238.87 is also assigned to each line. Thus, the total switch-related costs assigned to a line consist of two parts, the line's share of the common cost (which varies among the wire centers depending upon the total number of lines served) and a per-line cost (which is identical for each line).
Because of the wide range of total numbers of lines served by a wire center, the final switch-related investment cost per line varies from a low of, for example, $4.88 per month[3] to a high of, for example, $254.82.[4] As Figure 5.1 indicates, on average, approximately one-third of the monthly line cost (according to the BCM) is associated with the switch. [5]
The major factors that affect the BCM's computations of the per-line switch-related costs are: (1) the common switch costs; (2) the per-line switch costs; (3) the fill factor; (4) the business gross-up factor; and (5) the percentage of switch costs allocated to lines.
The BCM relies upon a relatively old (1990) secondary source for its switching costs.[6] Switching costs have decreased significantly since 1990, in part to the rapid pace of technological advances in the computer and electronics field, and in part due to the intense competition that has emerged in the switch market.[7] These same factors will ensure that prices continue to decline; in fact, competition in the switch market is likely to intensify if the RBOCs reenter this market now that the MFJ manufacturing restriction has been lifted. Moreover, not only has there been a sharp drop in the list prices of digital switching equipment, manufacturers have been offering progressively larger discounts to LECs as part of volume purchase contract negotiations.
Not only does the BCM rely upon an outdated secondary source, but the cost data used in that publication is even older than the 1990 copyright date: The source is a New England Telephone Company submission in an investigation by the Massachusetts Department of Public Utilities begun in 1986, some four years before the cited text was published.[9] Therefore, the switch cost data employed in the BCM model is at least ten years old. Considering the fact that digital switching costs were steadily declining over the last decade, this category of costs is overstated. Indeed the Bolter book cites an industry observation that "the equipment local carriers buy is going to fall in price."
Examining these cost data and then applying BCM's assumptions about the distribution of traffic-sensitive and non-traffic-sensitive costs illustrates the implausibility of the switch cost data upon which the model relies. If the cost of a digital state-of-the-art switch with a capacity of 100,000 lines indeed included $647,000 in fixed costs and $238.87 in per-line costs, and one used the BCM's algorithms, the total non-traffic-sensitive switch costs that would be allocated to lines would approach $24-million,[10] and only approximately $200,000 would be allocated to traffic-sensitive use.
How the BCM should be corrected
ETI developed a corrected switch cost figure by analyzing additions to plant and to digital access line capacity in California. In 1994, Pacific Bell added approximately $541.4-million in digital switch gross plant (Account 2211) and retired approximately $104.8-million in the same account.[11] This account includes Class 5 and Class 4 switches, as well as current and older generation digital switches. Assuming that approximately 80% of the plant additions in this account were for end office digital switches (the balance being for tandems, SCPs and other network switching facilities), approximately $433.1-million is attributable to subscriber lines. The BCM assumes that 70% of digital switch costs are line-sensitive and that the remaining 30% are traffic-sensitive. Of the latter, 30% are assigned to residential access line service, representing the "base" level of local use included within the basic residential service package. This results in a total assignment of $342.1-million in switch-related costs to lines added in 1994. In 1994, the net gain in capacity for digital access lines was 02,344,303.[12] However, the actual gross additions were greater than the 2.34-million net line gain, because approximately $104.8-million of digital switch plant was retired during that same year. Assuming that the digital switches that were retired were likely of a mid-1980s vintage (generally fairly small machines used in rural exchanges to replace step-by-step equipment), the average original cost per line for that equipment was probably in the $500 range. Therefore, the $104.8-million in retirements represents a decrease of approximately 209,600 digital lines. On this basis, the gross additions (corresponding to the 80% of the $541.4-million) represented about 2,553,903. Dividing the gross plant additions for end-office switches (80% of $541.4-million, or $433.1-million) by the 2,553,903 digital lines added, we develop an estimate of the per-line incremental cost of $170. However, only 70% of this cost, or $119, is line-sensitive. Of the remaining $51 of traffic-sensitive cost, only 30%, or about $15, is assignable to a primary residential access line. Hence, based upon this analysis, the corrected incremental switch-related investment that should be assigned to the primary residential access line is approximately $134.[13]
However, erring on the side of overstating the switch costs, ETI ran the BCM assuming that, for the sake of exaggeration, 100% of the Pacific Bell plant additions were for end office switches (an implausible situation, conducted as a type of "upper bound" analysis), which meant that we ran the BCM using a per-line switch cost of $167. Implementing the ETI corrected switch cost produces a dramatic reduction in the amount of universal service support requirement. The effect of making this sole correction (i.e., to $167 per line) is to reduce the statewide average monthly cost to $14.64, i.e., by approximately 14%, and, to reduce the USF requirement that the BCM computes by between 19% and 21%, depending upon the price threshold.[14]
We also conducted a sensitivity analysis of this assumption by analyzing the impact of using the cost of $134 (which reflects the assumption that 80% of Pacific Bell's additions to plant were for end office switches). The result of this sensitivity analysis -- relative to the result of using a cost of $167 -- was a reduction of sixty-six cents.[15]
Table 5.1
Implausibly High
Switch Costs
Exaggerate USF
Requirements
BCM ETI Correction Percent Difference
USF if $20 $77,846,835 $61,393,675 (21%)
USF if $30 $50,692,630 $41,171,091 (19%)
USF if $40 $37,662,589 $30,674,620 (19%)
Average monthly cost $16.941616The $14.64 (14%)
figure of $16.94
is eight cents
less than the BCM
results for
Washington State
reflected in the
December 1 Joint
Submission due to
minor differences
between the model
that was submitted
on September 12,
1995 to the FCC
and the model that
was used by the
Joint Sponsors for
the December 1,
1995 submission.
Conversation with
James Dunbar
(Sprint) on March
25, 1996. The
September version
of the model was
the one that was
widely circulated
and that is the
basis of our
report, with the
exception
regarding
modifications for
Surface Types that
we made (see
footnote 72,
above).
Notes: Data reflect
a cost factor of
22.97%; a corrected
per-line switch cost
of $167. The fixed
cost per line of
$647,526 was set to
zero. The table
reflects no other ETI
corrections. Results
are for Washington
State. Source:
Washington BCM files.
The lack of publicly available switch cost data should not excuse an exaggerated calculation of the size of the USF. Given the sensitivity of per-line costs to the costs that are assumed for switches, it is critical that a concerted effort be made to acquire the relevant switch cost data from the best available current source.[17] Clearly, the key reason that the switch cost data in the BCM are flawed is that LECs refuse to disclose this information, claiming that it is proprietary; until such time as LECs provide data that can substantiate a different result, the costs developed by ETI should be used. LECs do not have "clean hands" to challenge the accuracy of switch costs derived in this manner when they persistently decline to produce the allegedly "actual" costs on which they rely.
The model unrealistically deploys Nortel DMS 100 switches in small rural exchanges
In modelling switching, the BCM assumes that all switches are Nortel DMS 100s. The model appears to address those situations where remote serving units would typically be deployed[18] to minimize the costs of serving rural areas in one of the two following ways, either of which would cause the cost results to be exaggerated: Either the model deploys DMS 100 switches in these situations or it does not deploy any switch, but rather assumes distribution facilities all the way to the area served.[19]
For example, as Table 5.2 shows, the BCM "deploys" a switch costing a minimum of $647,526 in a wire center that serves only 23 households and 50 total lines.[20] As a result of deploying a DMS 100 to serve a relatively small number of lines, the total switch-related component of the per-line cost for the wire center in this example is $207.76 per month.[21] By comparison, the Washington state average switch-related cost component is $5.63. It would make no economic sense to deploy a large-scale switch in a small wire center, and numerous lower-cost alternatives are available. The BCM, however, does not consider such alternatives.
The inappropriate deployment of DMS 100 switches is a significant component of the overstated per-line cost for basic local exchange service that is yielded by the uncorrected BCM.[22] ETI's analysis shows that of the households for which the BCM determines a universal service support requirement,[23] 44% are in CBGs where, according to the BCM, the total switch investment is over $500 per line.[24] Clearly, one must question whether the decision to deploy a stand-alone switch under such conditions is prudent or necessary to meet universal service objectives.
Table 5.2 Comparison of
Switch Costs for the BCM
Washington State Average
and for an Illustrative
Rural Wire Center2525As is
the case with all numbers
yielded by the BCM, the
results of the uncorrected
BCM are neither fish nor
fowl because the per-line
switch costs reflect the
presence of business lines
but the outside plant costs
do not. This internal
inconsistency is addressed
in Chapter 6.
Washington State Average Wire Center STPSWAXA
Outside plant cost per line $11.31 $178.94
Switch cost per line $5.63 $207.76
Total $16.94 $386.70
Switch Cost as Percentage 33% 54%
of Overall Cost Per Line
Note: STPSWAXA is Stevens
Pass.
[2]Mark Bryant, MCI, in California PUC Workshops on Universal Service, February 5, 1996, Tr. 374, lines 13-16.
[3]This is for a line served by the RNTNWA01 wire center (Renton, a suburb of Seattle), which serves 42,935 households in 90 CBGs.
[4]This is for a line served by the STPSWAXA wire center (Stevens Pass, a small area in Eastern Washington), which serves only 23 households in a single CBG.
[5]This conclusion is based on an analysis of the approximate 5000 CBGs in Washington State.
[6]See Joint Submission, at IV-4 which lists as the source for switching costs, Telecommunications Policy for the 1990s and beyond, Walter G. Bolter, James W. McConnaughey,and Fred J. Kelsey (M.E. Sharpe, Inc., Armonk, New York, 1990), at 167, Table V-2.
[7]Brand, T.L. et al. "An Updated Study of AT&T's Competitors' Capacity to Absorb Rapid Demand Growth," AT&T Bell Laboratories, April 19, 1995 submitted as Attachment C to Pacific Bell's Second Set of Data Request in R.95-01-020, February 28, 1996.
[8]The presence of such discounting is not denied by LECs, and in fact they regularly decline to divulge any details of such terms and conditions, claiming that they are subject to confidentiality agreements with the switch manufacturers.
[9]Massachusetts D.P.U. 86-33, Investigation by the Department into the propriety of the cost studies filed by New England Telephone and Telegraph Company on April 18, 1986, pursuant to the Department's Orders in D.P.U., March 21, 1989.
[10][(100,000 lines * $238.87/line) + ($647,000) * (70)%]. The non-traffic-sensitive costs would "approach" this number because. clearly, the switch would not be deployed to full capacity.
[11]Pacific Bell, Report on the Results of Operations of Pacific Bell, Tracking Code PD-XX-14, 1994, at 13-9.
[12]This is the difference between actual 1994 digital access line capacity of 10,964,743 and actual 1993 digital access line capacity of 8,620,440. Pacific Bell Central Office Equipment Annual View Deployment and Utilization Forecasts.
[13]Note that the costs basis here is for 1994. Digital switch costs are continuing to fall, so in a forward-looking cost model a further downward adjustment could be made. According to Telephone Plant Index (TPI) data provided by the USTA in CC Docket 94-1, central office costs decreased by approximately 10% from 1984 to 1992, i.e., at about 1.3% per year over the eight-year period. Hence, extrapolating to 1997, the switch costs could be further reduced by approximately 4%.
[14]Mechanically, we corrected the switch cost as follows: we changed the amount in cell $H$5 of the Tables and Inputs sheet of the OTPT60~1.XLS file from the default $238.87 to the corrected $167. Also, we changed the fixed cost per switch located in cell $H$7 of the same file, from the default of $647,526 to $0. The practical result of this latter adjustment is to change the fixed cost per line calculations in column C of the Tables and Inputs sheet to $0, which, in turn, nullifies the effect of the non-traffic sensitive, and traffic sensitive percentages located in cells $H$9 and $H$12 respectively. (The reason is that these BCM calculations only affect the allocation of the fixed cost per switch which we have assigned a zero factor.) The effect of the business gross-up factor in cell $H$14 is similarly nullified by assigning a zero value to the fixed cost per switch.
[15]See Appendix 8B. Indeed, even the $134 cost per line may well be too high. In 1993, Pacific Bell announced a $1-billion plan to deploy modern switching technology statewide, and indicated that the new switches would serve 9.1-million lines -- thus its deployment plan would cost approximately $110 per line. Pacific Bell Press Release, January 25, 1993.
[17]California PUC, Universal Service Proceeding, A Discussion of Input Assumptions Used in the Hatfield Proxy Model, op. cit., footnote 27, at 2 which states, "Even if they were readily available, the discounts large LEC purchasers obtain are necessary to determine the effective prices the LECs paid. Hatfield Associates Inc. ("HAI") understands that such discounts can be impressive: e.g., in the vicinity of 50% or more for switching equipment." See id., at 7, citing U.S. Central Office Equipment Market -- 1994, McGraw Hill in support of HAI's estimates per line for digital switching in 1995 of $105 for BOCs, $114 for GTE, and $241 for other independents. Appendix 4B shows that Tier I LECs serve approximately 93% of lines.
[18]See, e.g., New England Telephone (Vermont) Depreciation Study, December 4, 1992, Account Narrative for Account 2212.
[19]This problem has been recognized by the Joint Sponsors and they have indicated their intent to address this issue. Ex parte submission, February 21, 1996, op. cit., footnote 75.
[20]The STPSWAXA wire center. This number reflects a business gross-up factor of 1.75 and a fill factor of 80%. This results in 50.3 lines ((23*1.75)/0.8) which we rounded to 50 for this analysis.
[21]This number is based upon a cost factor of 22.97%.
[22]In their Ex parte submission of February 21, 1996, the Joint Sponsors indicate that they plan to develop a matrix that will allow for the design of host and remote switches, and that will identify fixed and per-line costs for various switch sizes. The Joint Sponsors indicate further that different fixed and per-line cost factors will be used depending upon the switch size to allow for optimal switch selection. These improvements are appropriate and worthy of development. See, Ex parte Submission, op. cit., footnote 75.
[23]This assumes Cost Factor 2 and a price threshold of $30.
[24]Expressed in another way, according to the unrealistic results of the BCM, more than 60,000 households in Washington would require per-line switch costs of over $500 to serve. Results sheet, otpt60*1.xls.