AN EXAMINATION OF
OUTSIDE PLANT
COSTS
6.1 The BCM should be corrected to more accurately reflect the economy of scale and scope inherent in incumbent LECs' networks
The BCM only partially recognizes the presence of business lines
All else being equal, the unit cost of providing basic local exchange service within a given area declines as the quantity of lines served increases. There are several reasons for this condition. Among other things, fixed common costs (e.g., the common switch costs) are allocated over a relatively greater number of lines, thus driving down the cost assigned to any individual line, and relatively higher numbers of lines means that larger cable sizes are deployed, which result in lower per-line costs (see Appendix 3D).
The BCM relies upon household data that is available in the census, and because the model is census-based, it does not generally account for business lines that actually exist in each of the CBGs. Although the creators of the model have adjusted in some respects for the presence of business lines, the adjustments to reflect the cost impact of business lines have not been made on a sufficiently comprehensive basis to even remotely justify the low fill factors for outside plant that are assumed in the BCM.
One area where the BCM does, at least implicitly, recognize business lines is in the allocation of switch costs, by applying a gross-up factor of 1.75 to the quantity of residential lines before calculating the per-line share of common costs. Thus, for any given CBG, the model multiplies the number of households by 1.75 and then divides that result by the model's assumed switch fill factor of .8 to derive a total "Switched Lines" count for each CBG. A fixed cost per line is then calculated at the wire center level by multiplying the BCM's default switch cost of $647,526 by the percent of the switch cost that is allocated to the line and then dividing that product by the total number of switched lines for the wire center.[1]
However, the BCM does not recognize the presence of business lines in sizing cable. It thus understates cable size and thereby overstates the associated per-line cost. In the Hatfield Cost Model (HCM) submitted in the California Universal Service proceeding,[2] the HCM multiplies the number of household lines by business line multipliers which range from a low of 1.20 in the lowest density to a high of 1.83 in the highest density.[3] The consequence of incorporating this multiplier is to more accurately reflect the presence of business lines.[4]
By way of background, we note that, unlike the BCM, an earlier cost proxy study appears to account for business lines:
We obtained residential telephone penetration levels from the FCC's Statistics of Communications Common Carriers to estimate the total number of residential lines per density class, and we applied single- and multiple-line business access line percentages from the FCC document to the residential figures to obtain total business lines.[5]
Also, more recently, testimony submitted on behalf of MCI and AT&T in a state universal service proceeding states the following regarding the Hatfield Model's modification to the BCM:
We have modified the input data to account for business lines and two-line residences by density range. We select numbers per range to make the final access line totals approximately equal to those shown in the Common Carrier Statistics. This allows the Hatfield Model to size the line plant to accommodate these lines; basic BCM does not have this capacity built into the calculations.[6]
The low fill factors that are incorporated in the BCM likely reflect the volatility associated with providing telecommunications services other than first line, basic residence local exchange service, e.g., business lines and additional residence lines. However, although the BCM reflects the volatility of business lines, among other things, through the Joint Sponsors' assumption of low fill factors, the business lines themselves are not reflected in the BCM. Thus the BCM is internally inconsistent. We tested the implications of leaving the BCM's low fill factors intact, but increasing the lines served in the model by a gross-up factor of 1.44[7] in order to make the BCM more consistent internally. In this "high-volatile, including businesses" run, we multiplied the number of households in each CBG by 1.44; used the corrected switch cost of $134; used the BCM's default (i.e., low) fill factors; and made no other corrections. We compared the results of this run to a "non-volatile, single line residence only" run of the BCM assuming the default values for the quantity of households in each CBG, the corrected switch cost of $134, and the corrected fill factors of 95%.[8]
The low-fill factor run (with the business gross-up) yielded an average monthly cost of $13.20 (which is approximately $3.50 less than the BCM's uncorrected result) and the high-fill factor run (reflecting single-line residence) yielded an average cost of $11.88.[9] For the purpose of modelling universal service costs, the result yielded by the low-fill factor should be rejected because under no circumstances should universal service support requirements be increased as a result of the costs of serving business lines. On the other hand, the high-fill factor represents an upper bound cost because (1) it does not yet reflect other necessary corrections to the BCM, which are discussed in more detail below and in Chapter 7 and Chapter 8 and (2) some of the benefits of the economies of scale and scope associated with the fact that LECs serve residence and businesses should flow back to households.[10]
The use of CBGs as the relevant unit for high-cost analysis does not require that CBGs be used as the basis for determining support requirements
CBGs are relevant in the BCM for several reasons:
* CBGs define the area within which average per-line costs are computed and evaluated for the purpose of defining the need for and size of USF support.
* Each CBG is assigned to one of six density zones to determine certain cost characteristics (e.g., the mixture of underground and aerial plant).[11]
* CBGs define the boundaries of the areas that wire centers serve.
Although each CBG is assigned to one of six density zones for the purpose of determining certain cost characteristics, the presence of other factors (such as the total number of lines, rock hardness, etc.) means that the calculated cost per line is unique for each CBG.[12]
Because certain topographic and demographic information that would help to ensure an objective basis for specifying certain cost inputs to the model is available by CBG, the BCM uses that geographic unit for purposes of modelling costs. Although there are criticisms that can fairly be leveled at the choice of costing unit, the advantages of having these objective inputs is considerable. However, this decision to use CBGs to identify discrete areas that have a higher cost to serve can and should be de-linked from the decision as to the most appropriate unit of aggregation for purposes of determining support requirements.
Very small geographic areas, such as CBGs, are not economic service areas for the provision of telephone service. Standing alone, they would not be served by a switching entity and distribution architecture confined to that particular geographic unit. Individual wire centers and central office switch entities can most efficiently serve thousands, or even tens of thousands, of households and businesses, whereas CBGs embrace at most several hundred households (and no nonresidential customers), and would never be adopted as the foundation of an efficient network architecture.
In fact, in rural areas, the economic choice may well be to deploy wireless technology rather than to construct long runs of copper or fiber cables with extremely low utilization. Therefore, the BCM should incorporate a cost threshold above which wireless service is deployed rather than wireline. Wireless technologies, by their very nature, are not constrained to CBG boundaries.
By determining the need for USF support on a CBG basis rather than a wire center basis, the BCM significantly exaggerates USF requirements
The BCM overstates the USF requirement because it determines USF need based upon an examination of the cost per line separately for each of the approximate 220,000 CBGs. Instead, the determination of the need for universal service funding should be made over a geographic area at least the size of the wire center, to recognize that a LEC enjoys considerable economies of scale and scope that are only partially operative at the unduly granular CBG level.[13] Moreover, by its adoption of a "scorched node" model design in which all existing wire center locations are held constant, even the CBG-specific costs are misstated to the extent that a more efficient overall network architecture would have been indicated had the number and locations of wire centers been allowed to change. Accordingly, while the average costs of serving households may be calculated separately for each of the 220,000 CBGs (as the BCM does), the results of those individual calculations should then be aggregated at least to the level of the wire center before they are compared with the desired price to be supported.[14]
A clear demonstration of the need for wire center aggregation can be seen in the BCM's treatment of feeder costs. In general, there are typically a number of individual CBGs within the geographic area served by a LEC wire center; hence, under this construct, costs of serving individual customers within a given wire center serving area could vary based, among other things, upon distance from the CO, type of feeder/distribution plant construction, density, and terrain.
Because the BCM develops costs on a CBG by CBG basis, where certain costs are shared by two or more CBGs, the BCM must necessarily allocate these shared costs among the several CBGs in some manner. Where the use of plant by several CBGs within a wire center serving area is less costly, overall, than if each CBG were configured on a stand-alone basis, the savings is the result of some economy of scale. The particular means by which the shared costs and the scale economies are allocated among the several CBGs will materially affect the overall conclusions of the BCM or of any other modelling process that is based upon smaller-than-wire-center serving areas. There are, in fact, several alternative methods by which such costs and scale economies can be apportioned among a group of individual CBGs. To see why, consider the following example:
Suppose that a particular feeder cable consisting of 24 strands of fiber optic cable leaves the CO building heading due north. The cable connects to four Service Area Interfaces (SAIs), each one of which serves a CBG.[15] The four SAIs (A, B, C and D) are located at distances of 10,000, 15,000, 20,000 and 25,000 feet, respectively, from the CO, and each requires four (4) strands (one working pair and one "hot" spare pair). Assume further that 24 strands is the smallest practical size for feeder cable.[16]
The installed cost of this cable is $200,000, which works out to $8 per sheath-foot. How is this $200,000 to be apportioned among the four CBGs?
Method 1: Under this method, each CBG is charged a proportionate share of the segment(s) of the cable that it uses. Segment (1), which connects the CO with CBG (A), costs $80,000 (10,000 feet x $8/foot) and is shared by all four CBGs. Therefore, each CBG is assigned $20,000 for its share of this segment. Segment (2) costs $40,000 (5,000 feet x $8) and is shared by three CBGs (B, C and D), so each of these three CBGs is assigned $13,333. Segment (3) is shared by two CBGs, (C and D), so its $40,000 cost is split $20,000 to each. Finally, Segment (4) serves only CBG (D), so the entire $40,000 cost for this segment is charged to D. The result of these segment-by-segment allocations produces the following costs for each of the four CBGs:
A $ 20,000
B 33,333
C 53,333
D 93,333
Method 2: Each CBG bears the cost of the immediately preceding segment. Under this approach, each CBG bears the cost of getting to it from the previous CBG (or CO, in the case of A) in the string. The result is the following set of allocations:
A $ 80,000
B 40,000
C 40,000
D 40,000
Both of these approaches have certain validity, but neither is entirely satisfactory nor economically correct. Method 1, which is essentially the approach adopted in the BCM, imposes on the most distant CBGs costs that are not avoided if that CBG is not served at all. In other words, if (in our example) CBG D were simply not served, the LEC would avoid the last $40,000 of construction cost for the segment of feeder cable from C to D, but would not avoid any of the costs of providing cable to A, B or C. Indeed, if D were not served, Method 1 would assign most costs to each of A, B and C:
A $ 26,667
B 46,667
C 86,667
In deciding whether to serve D, a new entrant would thus not consider Method 1 costs at all, but would instead look to Method 2. That is, the firm would consider the incremental costs of serving each CBG given that it had already decided to serve other CBGs. If the new entrant had decided to serve CBGs A, B and C, then the incremental cost of serving D is $40,000, the cost of the one additional segment.
Although Method 2 more accurately captures the economic decision process by which a new local exchange service competitor would consider serving a given market area, to be useful the analysis must closely track the decision process itself. For example, it might not be appropriate to utilize Method 2 to determine separately the costs of A, B and C if the CLEC had determined that serving all three of these CBGs was the minimum market size that was economically efficient to serve. In other words, the correct application of Method 2 requires that a "baseline" be established and that increments be examined relative to that baseline. There is, however, no clear algorithm for establishing the baseline except, perhaps, for a rule that would require, as a threshold matter, that the baseline be no lower than the minimum market size that the CLEC can efficiently serve.
Method 2 is far closer to a "competitive outcome" than Method 1, because it reflects the true cost increment required to serve the highest-cost, most distant subscribers with all of the joint gains from scale and scope economies being assigned to the rural increment. Such an assignment of the economies of scale resulting from serving D in addition to A, B and C is appropriate and economically valid because those economies of scale would not be available at all, unless D were in fact being served. To see why, let us posit one additional Method (call it Method 3) for assigning joint costs among these four CBGs. Under Method 3, each CBG is assigned the entire stand-alone cost of serving it and it alone, which costs are then reduced by an allocation of the economies of scale. The stand-alone and joint costs, and the resulting economies of scale, associated with serving each of the four CBGs, are as follows:
Stand-alone Total Joint Savings
A $ 80,000 80,000 80,000 0
B 120,000 200,000 120,000 80,000
C 160,000 360,000 160,000 200,000
D 200,000 560,000 200,000 360,000
There are, of course, a number of means by which the savings could be allocated. In our present example, however, if A, B and C are the "baseline," then the only way in which the additional $160,000 of savings would become available to the carrier would be for that carrier to offer service to area D. Hence, strictly from the standpoint of the decision to serve, as long as area D generates sufficient revenue to offset the $40,000 of net additional construction cost, the carrier should choose to serve D.[17]
Costs should clearly be examined on a wire center, rather than on a CBG, basis in assessing whether a particular area exhibits "high cost" and is deserving of "high cost support." We have noted that there are substantial economies of scale and scope in jointly serving all of the individual CBGs within the wire center's serving area, and that the incremental cost of serving the ostensibly "high-cost" regions within the wire center service area must be determined in light of such economies of scale. In the examples provided here, it is apparent that the costs of serving the most outlying portions of a wire center district are also the ones that may benefit most heavily from the existence of the balance of the feeder and distribution infrastructure within that district, and it is economically wrong to assign to the most outlying area any costs that are not specifically avoided if the subject outlying area is not served at all. In the above example, the BCM would attribute to D a capital investment cost of $ 93,333. While still far below the $200,000 stand-alone cost of serving D, this method of assigning shared costs substantially overstates the actual incremental cost of extending service into the outlying area.
Method 2 best reflects competitive entry decisionmaking. When improving land for a new subdivision, a real estate developer, for example, is required only to construct the roads, water mains, and other infrastructure elements to the nearest point of connection to the existing infrastructure; he is not required to build a new road all the way to the center of town. The process for assigning shared costs to individual CBGs must similarly consider only those additional costs that would not be required but for the incremental addition of the new, outlying area. This is the approach that should be used if costs are to be determined at a CBG level. However, it is precisely because of the impact of scale economies on the shared costs of serving multiple CBGs that costs should not be determined at such a granular level, but should instead be determined and aggregated at a level no smaller than an entire wire center serving area.
Current high-cost support works off an average cost basis: LECs receive support when their costs in a given study area are more than 115% of the national average. Using this approach, the LEC only receives support if, on average, across its entire study area, its costs to serve are higher than the norm. Thus, a LEC with a combination of high-cost, low-cost, and average-cost exchanges might get little or no support, if the higher cost exchanges are cancelled out by lower cost ones. The design of the BCM prevents this sort of levelling effect from occurring. Under the BCM, any CBG that is modelled as having high-cost characteristics automatically generates an entitlement to support, regardless of the larger cost profile at the wire center (or some larger economic serving unit). The following example illustrates this point.
Consider the Colorado wire center AVONCOMA, which consists of 5 CBGs with per-line monthly costs ranging between $15.52 and $50.15. For sake of this example, assume Cost Factor #2 and a price of $30 per month being supported. In Table 6.1, the third column displays the BCM's calculations, given these assumptions. The fourth column is merely the calculated cost minus the support threshold of $30. The final column is the total CBG support, which is the support requirement per household multiplied by the number of households. In this example, three of the five individual CBGs would require support, amounting in the aggregate to $101,765.
Line Costs Compared Separately for Each CBG:
5
USF = _ (Support)CBGsi = $101,765
i=1
Table 6.1
Costs Are Too
High if Based
upon CBG
Averages
Illustrative
Analysis of
Wire Center
AVONCOMA,
Colorado
Assumes Cost
Support at $30
Per Line
Level1818See
Appendix 6.
CBG Households Monthly Support/Line CBG Support
(H.H.) Cost/Line
80379534002 144 $ 50.15 $ 20.15 $ 34,813
80379534004 177 $ 15.52 - -
80379534003 595 $ 22.76 - -
80379534005 266 $ 36.65 $ 6.65 $ 21,220
80379534006 771 $ 34.94 $ 4.94 $ 45,732
Total 1953 $ 101,765
Now take the alternative approach in which support is calculated at the wire center level. Utilizing the same five CBGs, the total cost in the final column is the monthly cost per household multiplied by the number of households, then multiplied by 12 to annualize the costs. Using this approach, the weighted average of the monthly cost per line is $30.82, an amount that barely qualifies the wire center for support at the $30 threshold, and the total support requirement across the wire center drops to $21,548. The major difference between this approach and the former is that high cost CBGs are counterbalanced by the low cost CBGs. Thus, by looking at high cost support requirements over the broader unit of a wire center, the overall support is significantly lower.[19] As this example shows, for those wire centers that straddle inexpensive and expensive CBGs, the difference between determining support strictly on a CBG-specific basis and determining it by rolling up results to the wire center level can be very large (in our example, $80,217 a year for a single wire center).
CBG-Based Line Costs Aggregated and Compared for the Wire Center:
5 5
USF=[ _ (Weighted Cost)CBGi divided by _ (Households)CBGi]
i=1 i=1
minus $30 * 12 months = $ 21,548
Table 6.2 Costs Should
Be Aggregated to the
Wire Center
Illustrative Analysis
of Wire Center
AVONCOMA, Colorado
Assumes Cost Support
at $30 Per Line Level
CBG Monthly Cost/Line Households Total Cost
80379534002 $ 50.15 144 $ 86,664
80379534004 $ 15.52 177 $ 32,964
80379534003 $ 22.76 595 $ 162,504
80379534005 $ 36.65 266 $ 116,988
80379534006 $ 34.94 771 $ 325,508
Total $ 30.82 1953 $ 724,628
Less Threshold $ 703,080
Total Support per ETI $ 21,548
corrections
This illustrative analysis shows an 80% reduction in the USF requirement that is actually needed. In order to implement ETI's correction on a statewide (or national basis), the Joint Sponsors would need to revise the model to allow for aggregation of cost results at the wire center level.
Competitors will seek out opportunities to develop their own economies of scale and scope, and will not find it efficient to limit service to discrete CBGs
Some have suggested that the geographic unit for determining both cost and universal service support must be very small, in order to prevent competitors from "cream-skimming." According to this argument, allowing higher- and lower-cost CBGs within a wire center to cancel each other out is fine so long as the LEC is serving the whole area, but not appropriate if a competitor can selectively "pick off" customers in the lower-cost CBG and leave the incumbent with the unsupported higher costs in the remainder.
As a general matter, the history of other telecommunications markets that have been opened to competition shows that concerns about "cream-skimming" and niche marketing as an entry strategy are typically misplaced and overstated. For example, initial niche competition in the long distance market matured into robust competition in the small business and residential markets. That robust competition itself has had the effect of lowering prices and making service more affordable for consumers, and is a far more efficient means of promoting affordable service than the indirect and distorting approach of requiring all providers to contribute to "social pricing." Moreover, although interexchange carriers confront lower average costs for serving customers in urban and metropolitan areas than in remote communities that are distant for the interexchange carriers' (IXC) "points of presence," IXCs offer service ubiquitously throughout the US and do not engage in geographic rate deaveraging or offer better prices to their urban customers.
There are several reasons why it is implausible that competitive providers of local exchange service will selectively pick off individual CBGs to serve, shunning selective high-cost CBGs within a wire center:
* It would be inefficient for a new entrant to design a network to serve such small segments because they would not be likely to be able to recover fixed costs if they were allocated over only units of 400 lines.
* Facilities-based competitors will seek to exploit their own economies of scale and scope (e.g., cable companies will serve, at a minimum, their franchise areas, not smaller niches).
* Distribution systems are not typically engineered to "hop around" to noncontiguous 400-customer serving areas, but rather are designed to radiate out from the center.
* The desirability of serving any particular area varies not only by the cost to serve but also by the anticipated revenue stream (i.e., not every high-cost area is undesirable to serve).
* Advertising for customers through mass media (e.g, newspapers and television) is not conducive to targeting such minuscule markets.
The question of how to define the appropriate area within which to assess the need, if any, for universal service support is an entirely different question from the question of how to define the areas that local exchange carriers should be required to serve.
6.2 The failure to eliminate costs incurred to enable the provision of second (and additional) lines would grossly distort universal service support requirements
As discussed previously, just as it is indisputable that a residential customer's first exchange access line meets the Act's criteria defining universal service, it is just as plain that a customer's second or additional lines do not fall within the definition. Additional residential access lines are not necessary for public health or safety, are not subscribed to by a significant portion of households (let alone a significant majority), and are, by and large, still used by more affluent customers for discretionary purposes (e.g., children's phone).
The inclusion of additional residential access lines within the scope of universal service would not be an issue if the aggregate capital investment (and associated revenue requirement) for the provision of residential service were unaffected by the potential demand for additional lines. Were that the case, any revenues derived from additional lines would be pure "gravy" from the perspective of the LEC. In fact, however, in order to provide capacity to satisfy demand for additional lines, LECs have designed and constructed far more extensive feeder and distribution infrastructures than would have been required for a "one line per household" service objective.
The costs of additional lines must be identified and removed from the aggregate cost of distribution plant to arrive at the economic cost of providing universal service
For context, it is helpful to start with a brief overview of the architecture of LEC outside subscriber plant. LEC subscriber outside plant is configured in a tree-and-branch type of architecture beginning at the serving central office building (the wire center). Feeder cable leaves the central office building in several different directions. At various points along its route, the feeder passes through a Service Area Interface (SAI) in which a portion of the total feeder cable capacity is cross-connected to a network of distribution cables that serve up to several hundred customers, usually located within 1,000 to 3,000 feet of the SAI.[20] Distribution cables run down individual streets and roads in the area subtending the SAI, where individual twisted copper pairs are connected to drop wires serving individual houses along the route. In the case of (relatively large) multiple-unit buildings, the SAI might be physically located within the building itself, with the building cable serving to distribute the service to individual apartments.
There is an inextricable relationship between the design of the outside plant and the demands for service that it is expected to satisfy. Also, because of the relatively high fixed costs of initial construction, the quantity of plant that is placed must be sufficient to satisfy demand for a reasonable period of time. Thus, in specifying an outside plant construction job, the LEC confronts an economic trade-off between incurring higher initial costs (to provide additional capacity for growth) vs. higher future costs if more frequent relief jobs are required because insufficient capacity was installed at the outset.
The three principal drivers of the investment cost of subscriber outside plant are (1) the initial base capacity, (2) the anticipated growth in capacity demand over the life of the plant, and (3) the variability (volatility) of demand for capacity over time. If demand were fixed (i.e., not growing) and stable (not variable) over the anticipated life of the installation, there would be no need to construct additional spare capacity at the outset; all that would be required is capacity minimally necessary to satisfy the known demand, with allowances for maintenance/administrative spare and for "breakage" (i.e., the need to order the next highest capacity cable, which is available in discrete capacity units). But subscriber outside plant supports a number of different services and demands, including the initial residential access line, additional residential access lines, business exchange access lines and trunks, private lines, and other special services. To the extent that these services may exhibit significantly different growth and variability characteristics, they impose capacity demands on the outside plant that are not in direct proportion to actual utilization.
It is also important to note that, with respect to outside plant, the facilities requirements that are imposed by the first residential access line in each dwelling unit are extremely stable -- indeed, probably the most stable and predictable of any of the categories of service that are provided by a LEC. If demand for residential access were limited to one line per household, there would be no need for the LEC to provide distribution capacity for either growth or to accommodate variability and churn.
An illustration may be helpful. Assume a street that has 80 households, each with a single residential access line. The distribution cable necessary to meet the requirements of this street would have 80 working lines, plus approximately 10% spare capacity (for maintenance requirements). Using standard cable sizes, the distribution plant would be provided over a 100-pair cable, with a ratio of working lines to total lines of 80%. Now, assume that the LEC knows that, on average, 20% of residential customers order a second line, but does not know which customers will request an additional line (or possibly more than one additional line). Thus, the LEC decides to install (on average) two pairs per household, increasing the number of pairs on the street to 160. With maintenance and administrative spare (again, 10%), the number of lines required totals 176, and the cable size deployed will need to be 200-pair cable. But, since the average demand for second lines is only 20%, only 16 more access lines in addition to the original 80 (first) access lines are actually utilized. In this example, including the capability to serve additional residential access lines results in a doubling of cable capacity and causes the utilization to drop from 80% to an extremely low 48%.
The foregoing discussion clearly illustrates that the aggregate cost of the distribution plant would be lower if it were designed solely to support a one-line-per-household demand. Building distribution plant piecemeal is not efficient, however. Thus, there is no argument that the distribution infrastructure should be built to accommodate more than the core one-line level of demand, recognizing the demand for additional services and because the incremental cost of building these additional units of capacity is significantly less at the time of initial construction. Nonetheless, to arrive at the economic cost of meeting the universal service objective, the costs for additional lines must be stripped away from the aggregate cost of constructing distribution plant. The maximum cost level that can legitimately be ascribed to universal service would be the stand-alone cost of constructing plant just sufficient to satisfy the non-growing, non-variable first line demand.
Universal service customers must obtain a reasonable share of the benefits arising from scale and scope economies
However, the costs that should actually be recovered through a universal service mechanism are considerably less than this upper limit. If the entire stand-alone cost is assigned to universal service, then all of the benefits associated with the economies of scale and scope that result from the ubiquitous deployment of plant would flow to services other than the first residential access line. Such a result would stand universal service policy on its head, and violates both the Act's definitional guidelines and its prohibition against cross-subsidization.
Again, the previous example is helpful to illustrate our point. Suppose that the stand-alone cost of the 100-pair installation (the one-line-per-household case) is $14,000, and that the cost of installing a 200-pair cable (the size that will accommodate first lines as well as additional lines) is $20,000. Under the stand-alone cost method, the entire $14,000 would be assigned to the 80 first line pairs, representing an investment cost of $175 per working primary access line. The cost for the additional 100 pairs placed to accommodate demand for additional residential lines would require (if placed at the same time as the first 100 pairs) is an additional $6,000. If seen in this light, the additional cost of the second 100 pairs is only 30% as much as the base cost of the first 100 pairs. In this example, this significant cost reduction in the construction of the additional lines is directly attributable to the deployment of the first 100 pairs: in other words, it represents a benefit from an economy of scale that arises directly from the LEC's fulfillment of its universal service obligation.
If the entire stand-alone cost of the first 100 pairs ($14,000) is assigned to universal service, then 100% of the gains from the economy of scale will flow to additional lines and none will inure to the benefit of universal service customers. The benefit at issue in this example is $8,000 -- i.e., the difference between the $14,000 stand-alone cost of the second 100 pairs and the $6,000 additional cost that the LEC actually incurs. There is no defensible policy reason why 100% of that $8,000 benefit should flow to discretionary services or to LEC shareholders, with none of it flowing back to defray any portion of the cost of furnishing the universal service primary residential access line.
To address this issue, an economically reasonable method must be developed to apportion such benefits to the two distinct uses of the shared distribution cable. There are several alternative methods for achieving such an apportionment:
* Apportion the benefit on the basis of capacity committed to each service category: using the previous example, since 100 pairs are deployed to serve "first line" demand, and an additional 100 pairs are placed in order for the LEC to satisfy "additional line" demand, the $8,000 of economy of scale would be spread according to the ratio of 100:100, or $4,000 to each category. The $14,000 stand-alone cost (applicable to either of the two groups of 100 lines) would be reduced by $4,000, so that both the first and the additional lines would each be assigned a cost of $10,000. Expressed on a per working line basis, the cost per first line would be $125 (i.e., $10,000 / 80), whereas the cost per working additional line (20% of 80 households, or 16 lines) would be $625.
* Apportion the benefit on the basis of working loops associated with each service category: In our example, there are a total of 96 working lines (80 first lines plus 16 additional lines). Spreading the $8,000 of economy of scale across these 96 working lines results in a per-line benefit of $83.33. Thus, $6,667 of benefits would be assigned to the first line service (80 x $83.33) and $1,333 of benefit would flow to the additional line service (16 x $83.33). This would result in a per-line cost for first lines of $91.67 [($14,000 / 80) - $83.33] and for additional lines of $791.67 [($14,000 / 16) - $83.33].
By contrast, the costing process employed by the LECs to allocate the shared costs of outside plant between first and additional lines puts a disproportionate cost burden on first (universal service) lines. The LEC methodology makes no distinction among services that share outside plant with respect to their relative growth and variability. Rather, it incorrectly calculates a per-working-pair unitized cost through a process that implicitly assigns all unused capacity in direct proportion to in-service capacity. This is accomplished by simply dividing the total cost of the outside plant by the number of working pairs. In our previous example, the LEC process calculates a per-working-pair cost of $208.33, by dividing the total $20,000 cost by the 96 working pairs. This methodology produces a total assignment of cost to the first line, universal service category of $16,667 which, in our example, is substantially in excess of the $14,000 stand-alone cost of providing for first line demand. The method also assigns only $3,333 of cost to the additional line demand. Thus, not only does the entire benefit of the scale economy flow to the additional line capacity, the first line service is actually made worse off than it would have been if the plant had been designed to accommodate first line demand only, i.e., at a cost of only $14,000.
There is no economic rationale for merely spreading spare capacity in proportion to in-service capacity, since the need for spare is not driven by the static quantity of in-service demand extant at any given point in time. In our example, the 80 working first line pairs account for 5/6th of the total working pairs in the cable, yet have no requirement for any spare capacity other than for maintenance and administrative purposes. The decision to construct an additional 100 pairs was driven entirely by the potential demand for additional lines, and should be assigned to that category of service. Instead, the LEC "proportionate allocation of spare capacity" method assigns 5/6ths of the total spare to first lines when in fact none of the additional 100 pairs are required for or utilized by first line service demand.
The additional costs of providing capacity for second lines are not insignificant
LECs seek to portray the additional costs of deploying additional capacity at the time of initial construction as being minimal, implying that the policy of providing capacity for additional lines and other (non-universal) service demands imposes little extra investment costs on the LEC. In fact, evidence adduced in the California Universal Service Proceeding suggests that the additional costs for this extra capacity may be quite substantial. In particular, it appears that the sizing and placement of Service Area Interfaces (SAIs) is driven by the total potential capacity demand rather than the number of initially-deployed working loops. Thus, if the "ideal" size of an SAI (for a given neighborhood) is 400 lines, the SAI will be assigned to only 200 dwelling units, on the basis of a potential demand of two lines per unit. The proliferation of SAIs will impose additional capacity costs on the feeder cables that connect the SAIs with the central office. Additional fiber strands and associated electronics (pair gain equipment) will be required for SAIs served by fiber, and additional copper feeder capacity to serve the more fragmented SAI architecture may also be needed. The proliferation of SAIs may also result in inefficient choices being made as between the use of copper vs. fiber optic feeder cable. The "crossover point" between these two technologies is a function of distance and capacity. Deloading of SAIs may make fiber feeder more costly on a per-working-pair basis, leading to more use of copper than might otherwise occur.
Any model used to determine universal service costs must correct for the overstatement of universal service costs incurred to accommodate demand for additional residential lines and other services
In summary, it seems clear that the aggregate cost of providing outside plant has been materially increased by virtue of the LECs' decisions to accommodate the demand for additional residential access lines and other services that go well beyond any universal service obligation. There are several potential methods of addressing and correcting this overstatement of universal service costs:
(1) Calculate service-specific objective working fill (utilization) factors based upon each service's rate of growth and variability of demand.
(2) Calculate the stand-alone cost of an infrastructure sized to support only first residential access line demand.
(3) Calculate the stand-alone cost of an infrastructure sized to support all services other than the initial residential access line.
(4) Calculate the differential between (2) and (3), and calculate a per-loop unit economy of scale benefit by dividing that differential amount by total working loops.
(5) Calculate a unit cost per working pair separately for first line (universal service) and for all other services as follows:
For first line loops: (2)/working first line pairs - (4)
For other loops: (3)/working non-first line pairs - (4)
6.3 The fill factors for the feeder and distribution should be increased
What the model does
The fill factors, which are a significant cost driver in the BCM, are significantly lower than they should be for a network designed to serve one line per household. The low fill factor results in exaggerated cost results because the model "deploys" excess capacity in order to satisfy the user-specified "objective" fill. Objective fill is the fill for which one engineers. Actual fill could actually be less than the objective fill as the following example illustrates. For example, assume the modelling of service to 26 households in the least dense zone. The BCM's objective fill of 25% for distribution means that for every 25 lines in service, there is an assumed need for spare capacity of 75 lines. Using this rule, for the 26 lines to serve the 26 households, the model identifies a need for 104 lines. However, this requirement is then added to a second rule that recognizes that the cable is sized in 100 pair increments (anything over a 100 pair requirement must go to a 200 pair cable). Thus, in the example, the rule would put 26 lines on a 200 pair cable, resulting in an actual fill factor of 26/200 or 13%. This result is not a flaw in the model but is simply an inevitable characteristic of engineering for spare capacity.
However, there is an apparent flaw in the BCM: The extremely low fill factors for distribution suggest that the BCM is modelling the expectation of large amounts of growth in second lines. The inappropriate inclusion of excessive spare capacity not related to the provision of primary basic telephone lines is a major issue that must be addressed. Because of the significance of this issue, it is addressed in detail earlier in the previous section of this chapter.
What the model should do
We have engineered the BCM's network using objective fill factors of 95% for the feeder plant and the distribution plant assuming 100% primary access line penetration. This objective fill allows for a minimum of (and often much more than) 5% spare capacity. Recognizing that the BCM engineers a network to serve all households, yet, the national subscribership is 93.8%,[21] the results of a national run should be divided by this percentage for a final number, or, alternatively, results of state-specific runs should be divided by the penetration rates of the individual states. This calculation reflects the fact that the BCM "deploys" lines to all households yet at any given time approximately 6.2% are not connected.
Using the ETI corrected fill factor of 95%, not only will spare capacity rarely fall below 10%, much of the time it will be greatly in excess of 10% as is illustrated in the simplified analysis reflected in Table 6.3 below.
Table 6.3 Cable Sizing Means that Actual Fill Will Be Significantly Less than Objective Fill Illustrative Analysis (95% Objective Fill) No. of HH Cable Size Actual Fill 47 50 94% 48 100 48% 50 100 50% 60 100 60% 70 100 70% 80 100 80% 95 100 95% 96 200 48%
ETI reran the BCM using the corrected fill factor of 95% for the distribution and feeder plant in all density zones (see Table 6.4).
Table 6.4
BCM Fill
Factors for
Cable Should Be
Increased
BCM ETI
Correction
Density (HH/Sq. Feeder Distribution Feeder Distribution
Mile)
0-5 65% 25% 95% 95%
5-200 75% 35% 95% 95%
200-650 80% 45% 95% 95%
650-850 80% 55% 95% 95%
850-2550 80% 65% 95% 95%
>2550 80% 75% 95% 95%
For this computer run we corrected the fill factor and, with the exception of adjusting the related structure costs, we did not incorporate any other ETI corrections. Increasing the fill factor implicitly raises the issue of the structure costs because of the BCM's algorithms for costing outside plant. The BCM includes structure cost multipliers to reflect the cost of deploying copper and fiber, with the multipliers varying depending upon several factors (e.g., whether the terrain is urban or rural and whether the cable is aerial or buried). The way the BCM is designed results in an entirely linear relationship between the size of the cable itself and the structure costs associated with installing the cable.[22] However, it is unlikely that the relationship is entirely linear because, for example, in rural areas, even if the cable size is reduced (thus reducing the cable costs), the structure costs are unlikely to decrease by a corresponding magnitude -- there are certain structure costs associated with deploying cable that simply do not diminish even if the size of the cable is scaled back.[23] Because this characteristic is most pronounced in rural areas, we modified the distribution cable multipliers to correct for this effect.[24]
Correcting the fill factor, and increasing the structure cost multipliers for the distribution cable plant resulted in a revised average cost for the state of Washington of $14.83, approximately $2.00 less than the BCM uncorrected results.[25] However, an examination of the results page of this run, which is included in Appendix 8B, shows that we have in effect over-compensated for the structure multiplier because the ETI results -- despite the higher fill factor -- yield higher costs for the lowest density zone than does the uncorrected BCM. Therefore, we also ran the BCM with a corrected fill factor of 95% without adjusting the distribution cable multiplier, i.e., using the BCM's default multipliers. This run yielded an average cost for the State of Washington of $13.69,[26] a result which does not adequately reflect the structure costs in rural areas, i.e., which under-compensates for the effect. Finally, we increased the structure multipliers (relative to the BCM), but to a level less than the original over-stated adjustments. The effect of using these multipliers and the correct fill factor of 95% was an average cost of $14.37.[27] These adjustments to the multipliers are reflected in our analysis in Chapter 8, below. Furthermore, absent any more compelling information, we recommend that they be used in future runs of the BCM.
6.4 The BCM makes an uneconomic choice between deploying copper and fiber in the feeder plant
One of the critical "decisions" that the BCM must make is when to deploy fiber rather than copper in the feeder plant. The deployment of fiber also requires the use of digital loop carrier equipment,[28] which is a significant component of the total average cost per line. Thus, there are two related aspects of this portion of the BCM that merit particular scrutiny: (1) The algorithm in the BCM which simply "decides" to deploy fiber in all instances where the total distance of the feeder and distribution plant exceeds 12,000 feet regardless of the capacity of cable being deployed and (2) the costs that are assumed for the digital loop carrier equipment.
As explained in Chapter 3, the BCM deploys copper main feeder to CBGs that have a total distribution distance (feeder and distribution) less than 12,000 feet and deploys fiber main feeder to CBGs that have a total distribution distance in excess of 12,000 feet, and selects one of two types of digital loop carrier equipment for these CBGs on the basis of household density.[29]
Implicit in this particular algorithm are important assumptions by the Joint Sponsors regarding the proper engineering of a local exchange network and the cost components that drive network development. First, the copper/fiber tradeoff for each CBG is made in isolation of other CBGs in the same quadrant. For example, the BCM assigns copper feeder to a CBG that has a total distribution distance that is less than 12,000 feet even when CBGs further out along the same main feeder route are served by fiber. In other words, the BCM would, in this case, deploy copper feeder alongside fiber feeder in the same conduit as opposed to using fiber feeder for all CBGs in the quadrant. The latter method would appear to be the more cost effective alternative as the cost of adding additional strands of fiber to serve the closest CBG would most likely be less expensive than the cost of deploying copper plant along the entire main feeder segment required by the first CBG.
One would expect that the copper/fiber crossover point that is "hardwired" into the BCM would reflect the economic crossover point between the deployment of copper and fiber feeder plant, that is, at a total distribution distance of 12,000 feet one would be indifferent as to the selection of copper or fiber main feeder plant from a cost standpoint. This assumption is complicated by the BCM's use of total distribution distance (feeder and distribution) as the point of reference for selecting a main feeder plant type as opposed to basing that decision on the length of the main feeder alone. For example, the BCM would allocate fiber main feeder plant to a CBG with a 2,000 foot main feeder segment and a total distribution distance within the CBG of 11,000 feet.[30] In contrast, a CBG that had a 10,000 foot main feeder segment and only a 1,000 foot distribution plant requirement would be served by copper main feeder plant because the total distribution distance is less than 12,000 feet. Furthermore, even if these two CBGs had the same number of households and the same household density, the first would be assigned a 2,000 foot fiber main feeder segment while the latter would be assigned a 10,000 foot copper main feeder segment.
We tested the implicit assumption that a 12,000 foot total distribution distance represents an economic crossover point for copper and fiber main feeder and determined through two different types of analyses that the BCM's 12,000 foot crossover point as presently constructed does not deploy the most cost effective network configuration. First, we ran the entire BCM using Washington State data and various crossover points for copper and fiber feeder plant.[31] Without altering any of the BCM's other user inputs we decreased the copper/fiber crossover point from 12,000 feet to 9,000 feet. Not surprisingly, this change resulted in an increase in the statewide average monthly cost from the default level of $16.94 to $17.84. We then increased the copper/fiber crossover point to 15,000 feet, again leaving all other user inputs and algorithms unchanged, and found that the statewide average monthly cost for Washington State decreased by $0.72 per month to $16.22. As illustrated in Table 6.5 below, the average monthly cost continued to decline as we increased the copper/fiber crossover point successively from 15,000 feet to 18,000 feet, to 21,000 feet, and finally to 24,000 feet. This analysis proves that on a statewide basis, the BCM's 12,000 foot copper/fiber crossover point, when used with the Joint Sponsors' default per line costs for SLC and AFC electronics of $500 and $550 (with the BCM's assumed discounts) does not lead to the most efficient network possible. Thus, the algorithm and the cost input data are contradictory: Our analysis shows that, if the cost data that the BCM assumes are realistic then the BCM's copper/fiber trade-off decision is uneconomic. Alternatively, if, for the sake of argument, the BCM's algorithm for the copper/fiber trade-off decision is "correct" then clearly the cost data are wrong.
Table 6.5
Average
Monthly Cost
by Household
Density Class
Using Various
Copper/Fiber
Crossover
Points*
Generated by
the BCM
Total
Distrib
ution
Distanc
e
(Feeder
and
Distrib
ution)
Measure
d in
Feet
Density Class 9,000 12,000 15,000 18,000 21,000 24,000 Percentage
**
<= 5 $99.07 $99.08 $99.08 $98.99 $99.01 $99.02 3.3%
5 to 200 $26.25 $25.96 $25.53 $24.98 $24.55 $24.05 20.0%
200 to 650 $14.92 $13.74 $13.03 $12.22 $11.68 $11.35 14.6%
650 to 850 $12.77 $11.56 $10.44 $9.68 $9.29 $9.04 5.8%
850 to 2550 $12.51 $11.29 $10.43 $9.93 $9.66 $9.57 36.8%
>2550 $10.19 $9.17 $8.65 $8.43 $8.44 $8.43 19.5%
Statewide $17.85 $16.94 $16.22 $15.69 $15.37 $15.15 100.0%
Average Cost
* Assumes
Annual Cost
Factor 2;
results are
for
Washington
State. **
The BCM's
default
copper/fiber
crossover
point is
12,000 feet.
As a second approach for testing the robustness of the copper/fiber algorithm, we tested the crossover point at the individual CBG level. Specifically, we duplicated the BCM's calculation of the "Total Loop Cost" for CBGs that were alone in a quadrant and which did not require sub-feeder segments.[32] We first identified CBGs in the Washington State input data that met the above criteria and which also had a total distribution distance (feeder and distribution) of slightly over 12,000 feet. The BCM had deployed SLC fiber feeder for all of these CBGs and the default "Grand Total Loop Cost" for these CBGs is shown below in Table 6.6. We then generated the "Total Loop Cost" for these CBGs using copper feeder and determined that it was significantly less costly to serve these CBGs with copper feeder plant. Again, this analysis was conducted using the BCM's default SLC per line cost of $500 and the default SLC discount of 20%.[33] Appendix 8C summarizes results for the same analysis for CBGs which had a total distribution distance of approximately 15,000 feet, 18,000 feet, 21,000 feet, 24,000 feet and 27,000 feet.
Table 6.6
Analysis of BCM
Main Feeder
Selection and
Alternative
"Total Loop
Cost" Assuming
Copper Main
Feeder
CBG # Total Feeder BCM "Grand Copper Savings Percent
and Total Loop "Grand
Distribution Cost" Total Loop
Distance Cost"
530419717004 12,194 $179,329 $101,895 $77,434 43%
53031950003 12,397 $131,017 $86,374 $44,643 34%
530419715002 12,420 $317,565 $179,482 $138,083 43%
530210208002 13,168 $314,366 $201,236 $113,130 36%
530050108021 13,585 $250,767 $169,915 $80,852 32%
Note: All CBGs
are in
Washington
State and are
"single CBG"
quadrants with
no sub-feeder
segment.
(Washington
State Input
Source Rows:
1790, 2224,
4366, 585,
1372).
The lesson from both the statewide runs and the individual CBGs analysis is clear -- the BCM default inputs and assumptions do not yield a true economic crossover point for copper and fiber feeder plant at 12,000 feet. In fact the BCM's default inputs produce a total distribution distance copper/fiber crossover point that is at least more than double that length given that the average cost for Washington State is still declining at a crossover point of 24,000 feet. Furthermore, the BCM's algorithm for choosing the copper/fiber crossover point still includes the ambiguity injected by the use of total distribution distance (feeder and distribution) as the point of reference for selecting a main feeder technology as opposed to basing that selection on the main feeder distance alone. The BCM does calculate the main feeder distance for each CBG and so the copper/fiber crossover algorithm could easily be changed to reference the main feeder distance as opposed to the total distribution distance. Furthermore, capacity plays an important role in the selection of an appropriate main feeder technology. The BCM's only recognition of the significance of capacity in the selection of feeder technology consists of the allocation of "AFC" to CBGs that are to be served by fiber and which are in the lowest household density class. However, a CBG's capacity requirement should figure into the initial choice between copper and fiber feeder plant and not just to the selection of various fiber electronics technologies as in the BCM.
As a final means of testing the BCM's 12,000 foot crossover point, we conducted further analysis of the cost comparison that is reflected in Table 6.6 in order to approximate the SLC electronics cost necessary to make 12,000 feet a true economic crossover point for copper and fiber main feeder plant. We used the same five CBGs that are shown in Table 6.6. In Table 6.7 below, we present the aggregate BCM "Grand Total Loop Cost" for these five CBGs and the aggregate Copper "Grand Total Loop Cost" as well as the total distribution cost. The distribution cost remains constant regardless of the main feeder technology. Therefore we subtracted the total distribution cost for the five CBGs so as to isolate the total main feeder expense using fiber and using copper. These total feeder costs for the five CBGs are shown in columns five and six of Table 6.7 We then recalculated the cost of fiber main feeder for the five CBGs assuming various per-line SLC electronics costs and a constant discount level of 20%. As shown in Table 6-7, a discounted SLC cost of $88 achieves an aggregate fiber main feeder cost for the five CBGs that closely approximates the total cost for these five CBGs assuming copper main feeder. This analysis further illustrates the fact that the BCM's default SLC electronics cost of $500 does not produce an economic copper/fiber crossover point at 12,000 feet, but instead, the effective, discounted SLC electronics cost necessary for an economic copper/fiber crossover point of 12,000 is less than $100.
Table 6.7 Approximation of SLC Electronics Cost Necessary for Economic Copper/Fiber Crossover Point at 12,000 Feet Average Total BCM "Grand Copper Total BCM "Total Copper Feeder and Total Loop" "Grand Total Distribution Feeder Cost" "Total Distrib. Cost Loop" Cost Cost Feeder" Cost Distance* 12,753 $1,193,044 $738,901 $574,225 $618,819 $164,676 SLC Base Cost $120 $110 $100 $90 SLC Discount 20% 20% 20% 20% Final SLC Cost $96 $88 $80 $72 Total Fiber $178,931 $167,355 $155,779 $144,203 Main Feeder Cost * Average Total Feeder Distribution Distance and aggregate costs are based upon the data for the five CBGs presented in Table 6.6.
This section has identified the serious flaws in the BCM's assumption regarding the cost for digital loop equipment and the economic crossover for deploying fiber in the feeder plant. Both of these related problems with the BCM can and should be corrected, the effect of which will be to lower the average cost that the BCM computes per CBG. For example, using the BCM costs for digital loop equipment, but changing the crossover point to 21,000 feet (from the default value of 12,000 feet) reduces monthly average cost by approx $1.40.
As demonstrated by the analysis in this section, the per-line costs for SLC and AFC digital loop carrier equipment are significant inputs to the final per line cost for fiber main feeder. As such these per-line SLC and AFC electronics costs should be viewed critically in determining the cause of the copper/fiber crossover anomaly revealed in the above analysis. The SLC and AFC costs of $500 and $550 per line respectively are user inputs to the Loop Module as are their respective discounts for SLC and AFC Electronics of 20% and 10%, however, these figures are not substantiated in any way in the Joint Submission. Because of the difficulty in obtaining more accurate cost data, we also examined SLC and AFC price data based upon approximations made by Hatfield Associates, Inc. (HAI) in the California USF proceeding,[34] i.e., undiscounted costs for SLC and AFC of $250 and $500, and discounts of 40% and 25%. We conducted a sensitivity analysis based upon these data and determined that the result of changing these data (and making no other corrections to the BCM) is to lower the average monthly cost by approximately $5.00 to an average cost of $11.94.[35] Clearly these inputs are critical components of the final cost results.
6.5 The assumption of uniform household density should be revisited
Among the areas of possible enhancements that the Joint Sponsors have identified is a modification to the existing BCM assumption of uniform household density. The Joint Sponsors indicate that the assumption of uniform household density may not be sustainable in CBGs with density of less than one household per square mile.[36] Indeed, while in the plains, there could be uniform household density, in most other parts of the country, cluster developments are more likely.
The Joint Sponsors have indicated that they plan to address this problem as follows:
For CBGs with less than 20 households per square mile the road network within the CBG will be identified. A buffer will be established around each road as an approximation of the area within the CBG where households have the highest probability of being located. Buffers will be set according to the following parameters: 10-20 Households/Sq. mil. -- 500 ft; 5-10 Households/Sq. mi. -- 1000 ft; <5 Households/Sq. Mi. -- 1500 ft. This buffer area will be used to form a new polygon for purposes of network design.[37]
This revision is clearly essential to correct the overstated distances for outside plant that the current version of the BCM now computes (and thus the overstated costs) and appears to be a reasonable method for addressing what would otherwise be a serious flaw in the model.[38] This correction will clearly lower the results of the cost proxy model and should be incorporated before the BCM is adopted as a policy making tool.
[2]California PUC, Consolidated R.95-01-020 and I.95-01-021, Rulemaking and Investigation on the Commission's Own Motion into Universal Service and to Comply with the Mandates of Assembly Bill 3643.
[3]California Universal Service Proceeding, A Discussion of the Input Assumptions Used in the Hatfield Proxy Model, Appendix A, page entitled Output Module, "inputs" worksheet.
[4]However, the Joint Sponsors have specifically rejected the use of a different business line multiplier for each density group, although they have not yet provided the rationale for this position. Ex parte submission, February 21, 1996, op. cit., footnote 75.
[5]The Cost of Basic Universal Service, Hatfield Associates, July, 1994, at 6.
[6]Pennsylvania PUC, Universal Service Proceeding, Direct Testimony of Dr. Robert Mercer, op cit, footnote 6, at 12.
[7]This figure of 1.44 is the ratio of all lines in the state of Washington to the number of households in the state of Washington. The number of residential lines is 2,062,385. SOCC, 1994-1995, Table 2.5. The number of households that have telephone service and that also have second lines is 7.2% based on ETI's calculation of additional residential lines using FCC methodology. The subscribership rate in Washington State is 96.0%. Monitoring Report, May 1995, Table 1.2 (1994 Data). Therefore, using these same data yields a computed number of households of 2,004,028 (i.e., [2,062,385 divided by 1.072] divided by 0.96). The total number of lines (including second lines, business lines, and public access lines) is 2,881,344. Statistics of Common Carriers, 1994-1995, Table 2.5. Thus, the ratio of the total number of lines to the number of households is 1.44. (Because the BCM does not document the source of its business gross up factor of 1.75, we cannot explain the discrepancy between our figure and that of the Joint Sponsors.)
[8]Because in this "baseline" run we use 95% fill factors, we adjusted the distribution cable multipliers as discussed in 6.3.
[9]These data reflect Cost Factor No. 2. Appendix 8B.
[10]Throughout this report, ETI's corrections are intended to modify the BCM so that it reflects the appropriate scope of service, i.e., the deployment of a single line to each and every household, so that the BCM quantifies the stand-alone costs of that basic residence local exchange service. However, conceptually, a separate BCM should be run to quantify the stand-alone costs of meeting all other local exchange service. Because the combined cost of serving both "universes" (i.e., single-line residence and all other local service) should be less than the sum of the two parts (contrary to the aphorism, "the whole is greater than the sum of the parts"), the results of the stand-alone analysis of residence costs should then be scaled back appropriately.
[11]See Appendix 3C.
[12]For example, CBG 80379534002 and CBG 80379534006 of wire center AVONCOMA, Colorado are both assigned to the second Density Zone (5 to 200 per square mile), but the monthly cost per loop in the former CBG is $50.15 and the monthly cost per loop in the latter CBG is $34.94. See output derived from CODTIN1.XLS, Rows 225 and 229.
[13]The Joint Sponsors have indicated that they will not make a modification to the model that would allow a user to calculate the average cost at the wire center level. Ex parte submission, January 26, 1996, op. cit., footnote 75. The Joint Sponsors indicate that an "interested user of the BCM could perform an aggregation of all CBGs in a wire center to obtain an approximation of cost at the wire center level." Ex parte submission, February 21, 1996, op. cit., footnote 75.
[14]Certain RBOCs have raised this problem. See, e.g., Comments of Bell Atlantic, Docket 80-286, at 8: "[a]s the Commission acknowledges, use of such a small measurement area [CBGs] could substantially increase the total subsidy requirement and with it the size of the USF."
[15]Under the architecture assumed in the BCM, each individual CBG is served by one SAI. In fact, of course, the actual design of the feeder and distribution networks bear no relationship whatsoever to CBGs and CBG boundaries.
[16]To be precise, the BCM would not assume the use of fiber to serve CBG A, because it is less than 12,000 feet from the CO. Thus, the model would engineer copper feeder cable as far as CBG "A" along the very same route as the fiber, with the fiber bypassing "A" and serving only B, C and D. Such a configuration would, of course, make no real sense; hence, if anything, the arbitrary use of the 12,000 copper/fiber crossover point will almost invariably lead to an overstatement of optimally engineered feeder costs.
[17]This is not to say that the price the carrier charges for serving D should be set to recover only the $40,000 incremental cost. For pricing purposes, it may be appropriate to spread the $160,000 in total efficiency gains among those other customers whose existence made those gains available to the carrier when serving D. However, for purposes of determining the effect of a requirement that the incumbent serve an entire wire center area, only the $40,000 cost incremental is relevant.
[19]Of course, there are also many wire centers for which aggregating costs to the wire center level would not affect the support requirement. Such is the case when a wire center serves all high-cost CBGs (e.g., a completely rural area) or all low-cost CBGs (e.g., an area of moderate density). The incidence of exchanges containing both CBGs that would require support and ones that would not is also directly affected by the specification of the affordability threshold: with a lower threshold (e.g., $20), a significant number of wire centers have CBGs falling both above and below the threshold, whereas with a higher threshold (e.g., $40), there is a significantly lower incidence of "mixed" results.
[20]The current version of the BCM does not include the costs of the SAI. Apparently the next version will include these costs. California PUC, Universal Service Proceeding, A Discussion of Assumptions used in the Hatfield Model, op. cit., footnote 27, at 10.
[21]FCC Monitoring Report, May 1995, Table 1.1, at 18 (Subscribership Data for November, 1994).
[22]Joint Submission, September 12, 1995, at 30-31.
[23]According to one of the Joint Sponsors, were a user to increase the fill factor, it would not be necessary to make any other changes to the BCM, because the structure costs would and should change in a corresponding manner. Conversation with James Dunbar (Sprint), March, 1996. We believe that this effect is implausible, however, and thus have conducted runs of the BCM with and without modifications to the structure cost multipliers.
[24]As discussed below, we also examined the sensitivity of the BCM to varying this structure multiplier adjustment.
[25]This result reflects solely the correction to the fill factor. See Appendix 8B. Specifically, we multiplied the distribution cable multiplier by 20 if the household density for a CBG was less than 200 and the total number of households was less than 200. We multiplied the distribution cable multiplier by 3 if the household density was less than 200 and the total number of households was between 200 and 400. These adjustments were made in the Data & Calcs sheet of the Data Module.
[26]Appendix 8B.
[27]We multiplied the distribution cable multiplier by 10 for CBGs with a density of less than 200 households per square mile and fewer than 200 households. We multiplied the distribution cable by 3 for CBGs with a density of less than 200 households per square mile and between 200 and 400 households.
[28]Digital loop carrier equipment is multiplexing equipment that combines multiple electronic signals onto a single bit stream for transmission over fiber.
[29]AT&T's "SLC" (Subscriber Line Carrier) is assigned to CBGs in the five largest household density classes and Advanced Fiber Communications' "AFC" is assigned to CBGs in the lowest household density class.
[30]2,000 foot main feeder distance + 11,000 foot distribution distance = 13,000 foot total distribution. The BCM assigns fiber main feeder to CBGs with total distribution distance over 12,000 feet.
[31]The Main Logic Sheet of the Loop Module where the copper/fiber crossover algorithm is found, is password protected. We were able to overcome this restriction.
[32]To simplify replication of the BCM, we selected CBGs that were "single CBG" quadrants and which did not require a sub-feeder segment. In other words, we chose CBGs that were directly in the path of the main feeder route and which did not share main feeder costs with other CBGs.
[33]CBGs chosen for this analysis had household densities greater than five and so were served by SLC and not AFC equipment.
[34]California PUC, Universal Service Fund Proceeding, A Discussion of Input Assumptions Used in the Hatfield Proxy Model, op. cit., footnote 27.
[35]See Appendix 8B.
[36]Joint Submission, September 12, 1995, at 38.
[37]Ex parte submission, February 21, 1996, op. cit., footnote 75.
[38]If feasible, this revision should include a data overlay that entirely excludes areas that have roads that serve areas that are entirely uninhabited (e.g., logging areas) and that possibly includes areas such as national parks that may well be inhabited (e.g., by forest rangers).