BEFORE THE WASHINGTON UTILITIES AND TRANSPORTATION COMMISSION
 

In the Matter of Determining Costs ) DOCKET NO. UT-980311(a)
      )
for Universal Service   )









DIRECT TESTIMONY OF 

ROBERT A. MERCER

ON BEHALF OF

AT&T COMMUNICATIONS OF THE PACIFIC NORTHWEST, INC.

AND

MCI TELECOMMUNICATIONS CORPORATION



June 15, 1998


Q. PLEASE STATE YOUR NAME AND BUSINESS ADDRESS.
A. My name is Robert A. Mercer.  I am the President of HAI Consulting, Inc. 
(“HAI”), a telecommunications technology and economics consulting firm.  The 
address of the firm is HAI Consulting, Inc., 737 29th Street, Suite 200, 
Boulder, Colorado, 80303. 

Q. PLEASE DESCRIBE YOUR EDUCATIONAL BACKGROUND.
A. I received a Bachelor of Science degree in Physics from Carnegie Institute 
of Technology (now Carnegie - Mellon University) in 1964, and a Ph.D. in 
Physics from Johns Hopkins University in 1969.  I have attended and taught 
numerous courses, seminars, and conferences in the field of telecommunications.

Q. PLEASE DESCRIBE YOUR PROFESSIONAL EXPERIENCE.
A. A complete resume is included as Exhibit RAM-1 to this testimony.  Briefly, 
I graduated with a B.S. degree in Physics from the Carnegie Institute of 
Technology (now Carnegie-Mellon University) and a Ph.D. in Physics from Johns 
Hopkins University .  After graduation, I was an Assistant Professor of Physics 
at Indiana University from 1970 until 1973.  I then joined Bell Telephone 
Laboratories.  Over the next eleven years, I held a variety of positions in the 
Network Planning organizations at Bell Labs and AT&T General Departments.  My 
final position at Bell Labs was Director of the Network Architecture Planning 
Center, where I managed an organization that was responsible for early Bell 
System planning of the Integrated Services Digital Network (ISDN), as well as 
systems engineering for new data services being planned by AT&T.
 
 Upon the AT&T divestiture, I joined Bell Communications Research (Bellcore) in 
January, 1984, where I was Assistant Vice President of Network Compatibility 
Planning.  Among other responsibilities, I directed Bellcore's technology 
analysis of various legal and regulatory proceedings at the federal and state 
levels.  I also coordinated and provided direction to Bellcore's activities in 
domestic and international standards activities, and served as a member of the 
Board of Directors of the American National Standards Institute.
 
 After leaving Bellcore in late 1985, I held positions with BDM Corporation and 
AT&T Bell Laboratories before joining Hatfield Associates, Inc., in early 
1987.  I held the positions of Senior Consultant, Senior Vice President, and 
President of the firm.  On October 1, 1997, the former principals and employees 
of Hatfield Associates, Inc., formed HAI, and I became the President of that 
firm.  The firm specializes in engineering, economic, and policy studies in the 
telecommunications field.  Our clients include firms involved in premises, 
local exchange, long-haul and international networks, satellite communications, 
cellular mobile radio, conventional mobile radio, cable television, and data 
and computer networking.

 I also hold an adjunct faculty position in the Interdisciplinary 
Telecommunications Program at the University of Colorado in Boulder, where I  
have taught a course on advanced data communications and computer networking 
for several years, and serve on masters thesis committees.  I have taught many 
other courses and seminars as well, in the areas of the telecommunications 
infrastructure, network technologies, broadband networks, data and voice 
communications, computer networking, and network management. 

Q. HAVE YOU TESTIFIED PREVIOUSLY BEFORE THIS COMMISSION?
A. Yes, on several occasions.  Most recently, in the Docket 
UT960369/UT960370/UT960371 proceeding on the generic costing and pricing of 
unbundled network elements (“UNEs”), I testified during July of 1997 about the 
use of what was then called the Hatfield Model in establishing the cost of UNEs.

Q. WHAT IS THE PURPOSE OF THIS TESTIMONY?
A. It is to provide the Washington Utilities and Transportation Commission 
(“Commission”) with the most up-to-date information on the HAI Model Version 
5.0a (“HM 5.0a”).  To make the record clear on the model’s name, let me note 
that the former Hatfield Model has been renamed the HAI Model as of Release 
5.0a.  HM 5.0a  was filed with the Federal Communications Commission (“FCC”) on 
February 6, 1998.

Q. HOW IS YOUR TESTIMONY ORGANIZED?
A. My testimony is divided into four parts.  Part I provides an overview of the 
HM 5.0a.  By relying on the HM 5.0a documentation provided in several exhibits 
to this testimony, this overview is kept reasonably brief and at a high level.  
Part II describes the significance of new customer location and clustering 
algorithms utilized in HM 5.0a, contrasting it with the approach of earlier 
versions of the HAI Model and with the algorithms used in the Benchmark Cost 
Proxy Model (“BCPM”).  The customer location and clustering process is perhaps 
the most dramatic change of many significant changes in HM 5.0a compared to 
prior versions.  This portion of my testimony also addresses the entirely 
unwarranted criticisms the BCPM sponsors have leveled against HM 5.0a.  Part 
III discusses the consistency of the HM 5.0a with both FCC and this Commission’
s criteria to be met by proxy cost models.  Finally, Part IV summarizes why I 
believe  HM 5.0a is the best option available to this Commission for estimating 
the costs associated with providing universal service and  any subsidies 
associated with the universal service fund  (“USF”). 


Q. ARE THERE EXHIBITS TO YOUR TESTIMONY?
A. Yes, my testimony includes four exhibits as follows:
 RAM-1:  Resume for Dr. Robert A. Mercer
RAM-2: HM 5.0a Model Description, Automation Description and User Guide
 
 RAM-3: The HM 5.0a Inputs Portfolio
RAM-4:  AT&T/MCI Ex Parte titled “HAI Model 5.0a: Why It 
Engineers the Appropriate Amount of Distribution Plant”

 I. OVERVIEW OF HM 5.0a

Q. PLEASE PROVIDE A GUIDE TO THE DOCUMENTATION YOU HAVE ATTACHED AS EXHIBITS TO 
YOUR TESTIMONY THAT INCLUDE A DETAILED EXAMINATION OF THE NATURE AND CONTENT OF 
HM 5.0A.

Section 1.2 and Appendix A of the HM 5.0a Model Description contained in 
Exhibit RAM-2 discuss the evolution of the HAI Model in response to ongoing 
intensive internal reviews of the Model, legitimate criticisms of past versions 
by various parties, new developments in modeling techniques and data sources, 
and new regulatory requirements.  For those users who are familiar with the 
predecessor Release of the Model, HM 4.0, Section 2 summarizes the changes 
between HM 4.0 and HM 5.0a.  This description is broken down into one part that 
describes the changes between HM 4.0 and HM 5.0, the version of the model that 
was released to the FCC on December 11, 1997, and one part that describes the 
additional changes between HM 5.0 and HM 5.0a.

Section 3 presents the basic local network structure assumed by the model; it 
does not  differ substantively from the network structure assumed by earlier 
versions.  The same can be said of the overall model organization and structure 
described in Section 4.  On the other hand, Section 5 demonstrates that rather 
dramatic advances have occurred in processes for developing the 
study-area-specific demographic and telephone company data.  Similarly, Section 
6 presents a detailed view of the extensive changes in the assumptions, logic, 
and operation of the Model that are summarized in Section 2. 

Appendix A of RAM-2 provides a history of the HAI Model.  Appendix B  
identifies and defines the more than 1400 user inputs to HM 5.0a.  Appendices C 
through E of RAM-2 provide additional information on the databases and logic 
flow of HM 5.0a.  The document titled HM 5.0a Automation Description and User 
Guide at the end of RAM-2 explains the computer system requirement to run HM 
5.0a, and how to install and run the model.  The description of how to run the 
model includes the mechanism for changing the default values of user inputs, 
and for invoking various features and capabilities of the Model.  

The HAI Inputs Portfolio (“HIP”), included as Exhibit RAM-3, provides the 
rationale and support for the default values of these inputs.  There have been 
a number of changes in the nature and default values of these inputs between HM 
4.0 and HM 5.0a, based on the ongoing exhaustive analysis of the model’s logic 
and inputs.

Q. PLEASE DESCRIBE THE FUNDAMENTAL NATURE OF HM 5.0A.
A. Like earlier versions of the Hatfield Model, HM 5.0a is a bottom-up  
economic-engineering costing model of basic local exchange service developed by 
HAI at the request of AT&T and MCI.  The Model estimates (in a consistent 
fashion) the costs that an efficient firm would incur to provide unbundled 
network elements (“UNEs”), universal service, and interconnection services.  
Specifically, the HAI Model estimates the costs that an efficient LEC would 
incur to provide narrowband, voice-grade telephone services in a manner that is 
also capable of providing access to advanced services. The network modeled by 
HM 5.0a is capable of supporting advanced services (i.e., ISDN services), but 
does not include all of the costs necessary to institute ISDN service over the 
network.

Q. WHAT DO YOU MEAN WHEN YOU SAY HM 5.0A IS A BOTTOM UP ENGINEERING AND 
ECONOMIC MODEL?
A. I mean that HM 5.0a constructs a network based on detailed and granular – 
“elemental,” if I may use that term – information as to service demand, network 
component capacities and costs, and expenses, as opposed to, say, attempting to 
decompose total costs or revenues of existing telephone companies into their 
constituents . Specifically, the Model carries out the following eight major 
steps: 
First, it determines the amount and location of demand for local exchange 
service, network elements, and network interconnection for the Incumbent Local 
Exchange Carrier (“ILEC”) and jurisdiction under study, using geo-coded 
customer data and publicly-available census data, as described in Section 6.1 
and Appendix C of the HM 5.0a Model Description included as Exhibit RAM-2.

Second, it incorporates jurisdiction- and/or company-specific data on local 
terrain attributes for each relevant Census Block (“CB”), based on the Census 
Block Group (“CBG”) to which that CB belongs, to identify circumstances in 
which the terrain attributes will cause installation costs to increase over 
their normal levels.
 
Third, it provides over 1400 user-adjustable inputs that enable adjustments to 
reflect specific local conditions and circumstances, and/or to permit 
sensitivity analyses to be performed.  Default values are set for each of these 
inputs that reflect industry practices, suitably adjusted to be consistent with 
the forward-looking orientation of the Model.  Appendix B of Exhibit RAM-2 
identifies, defines, and gives the default values of these user inputs.  
Exhibit RAM-3 provides the rationale and support for each of these inputs.

Fourth, based on the forward-looking network architecture being deployed by 
ILECs today, it determines the amounts of various network components needed to 
support the known demand for the elements and services in question.

Fifth, using public information and opinion from subject matter experts on the 
availability, capacities, and costs of network assets and facilities available 
in the marketplace today, which are provided to the Model through the user 
inputs, it estimates the investment required to purchase and deploy the 
requisite quantities of each identified component considering both material and 
labor.

Sixth, it determines the cost of operating the network, taking into account all 
relevant capital carrying costs, network operations, customer operations, and 
corporate overhead costs (with forward-looking adjustments where appropriate).  
Again, various parameters required to make these calculations are provided to 
the Model through user inputs. 

Seventh, it calculates per-unit UNE costs, network interconnection costs, 
and/or the cost of universal service, depending on the nature of the particular 
proceeding in which the Model is being utilized.  At the user’s discretion, 
these results can be displayed by line density range, wire center, CBG, or 
individual customer location “cluster.”

Finally, the Model run produces outputs and associated intermediate results 
that are available for public scrutiny, both in hard-copy and electronic form.

Q. WHAT ARE THE KEY ATTRIBUTES OF HM 5.0A?
A. In a number of ways, HM 5.0a represents a revolutionary advance in the 
modeling of local telephone network costs.  Specifically, HM 5.0a incorporates 
the following: 
Actual geocoded customer locations;
An algorithm that identifies clusters of customers that may be served 
efficiently together – without recourse to arbitrary geographic limitations 
other than wire center boundaries;  The model could be modified to eliminate 
even this limitation, but constraining a cluster to fall within a single wire 
center is consistent with the FCC “scorched node” approach.
Numerous optimization routines that ensure 1) the use of outside plant that is 
most technically and economically suited to particular local conditions; 2) the 
appropriate economic choice of feeder technology between copper cable and 
fiber-based digital loop carrier systems; and 3) at the user’s option, the 
appropriate economic choice between wireline and wireless distribution systems;
Explicit specification of host, remote and stand-alone switches, and 
appropriate subsequent treatment of their costs;
An optimizing algorithm for the creation of efficient interoffice SONET 
transport rings; 
The ability to portray costs at different levels of granularity, including 
individual serving areas if desired; and
Opportunities to flexibly allocate certain expenses based on lines or relative 
investments.

Q. IS THIS A COMPLETE LISTING OF THE DESIRABLE ATTRIBUTES OF HM 5.0A?
No, it identifies only the key attributes, focussing on those that I consider 
to be the most dramatic departures from earlier versions.  Even a cursory 
examination of Section 2 of Exhibit RAM-2 will demonstrate many more desirable 
attributes even compared to the prior version of the model, HM 4.0.  Other 
valuable aspects of the model include its public nature, the ease of running it 
on standard personal computers, the ability to audit its results by examining 
intermediate outputs of various modules, and the user-friendly interface that 
enables users to change the more than 1,400 user-adjustable input values, as 
well as readily keeping track of those changes.

 II: CUSTOMER LOCATION AND CLUSTERING IN HM 5.0a

Q. YOU HAVE CHARACTERIZED THE CUSTOMER LOCATION AND CLUSTERING ALGORITHMS 
UTILIZED BY HM 5.0A AS BEING AMONG THE KEY FACETS OF HM 5.0A.  PLEASE SUMMARIZE 
THE OVERALL SIGNIFICANCE OF THESE ALGORITHMS.

The location algorithm ensures that customers are located at, and in many cases 
much better than, a CB level of specificity.  This is a marked improvement over 
the location process used in earlier versions of the Hatfield Model or BCPM, 
which essentially located customers using CBG information.  This difference in 
specificity is conveyed by the fact that nationally there are more than 28 
times as many CBs as CBGs.  The clustering algorithm ensures that the 
identified customer locations are served by outside plant that is configured to 
be economically efficient and consistent with design guidelines that are based 
on the characteristics of currently-available outside plant technology.  
Together, then, the location and clustering algorithms provide a much higher 
degree of certainty in the loop cost results produced by the Model.

Q. PLEASE SUMMARIZE THE HM 5.0A CUSTOMER LOCATION PROCESS.
A. This process is described in detail in Section 5.4 of RAM-2.  Sections 5.1 
through 5.3 are also useful in understanding how residential and business line 
counts are determined.

 Briefly, HM 5.0a utilizes the most precise customer location information that 
is available.  Wherever customer locations have been precisely determined by 
geocoding, that information is used.  Geocoding, as applied in HM 5.0a, locates 
customers with a 50-foot accuracy.  The percentage of all customer locations 
identified using geocoding is growing steadily over time, as the need of postal 
patrons, E911 service providers, zoning boards, and utility mapping act to 
increase the importance of this activity.  In Washington, geocoding currently 
locates approximately 60% of all customer locations.

 For those locations for which no geocoding information is available, the Model 
uses the next most precise information source available; namely, the U S Census 
Bureau’s location of residential households by CB.  It assumes those customers 
are located uniformly around the boundary of the CB. 

Q. WHY IS IT APPROPRIATE TO ASSUME THAT THE HOUSEHOLDS THAT ARE NOT GEOCODED 
ARE LOCATED ON THE BOUNDARIES OF CBS?
A. Obviously, it would be desirable to know the actual locations of every 
household (and, equivalently, of every business).  One attractive aspect of 
using geocoding information is that it gets better and more complete over time, 
as more and more households are geocoded in response to pressure from the 
postal system and emergency services.  Until such time as all households are so 
located, the CB data is the next best available source of information.  It is 
crucial to recognize that this data provides no detail below the CB level, and 
no model, whether HM or BCPM or the FCC’s HCPM, can do better than that.  Any 
attempt, such as BCPM’s, to claim that customers can be located more precisely 
is misleading, and is subject to the concern expressed by the FCC that efforts 
to report costs using “excessively small geographic units . . . creates a false 
sense of precision because the input data is still not disaggregated at that 
level.” Federal-State Joint Board on Universal Service, Forward-Looking 
Mechanism for High Cost Support for Non-Rural LECs, CC Docket Nos. 96-45, 
97-160, at p. 39.

As to why HM 5.0a assumes the customers are located on the boundary of the CB, 
rather than at internal points, this is motivated by 1) the belief it is 
appropriate to err, if at all, on the side of assuming too much customer 
dispersion, rather than assuming some arbitrary clustering that is not known 
from the data; 2) the observation from examining geocoded data that customers 
tend to be located along CB boundaries; 3) the observation that whenever a CB 
definition would otherwise cause there to be a substantial number of households 
inside the CB, additional CBs are defined internal to, and surrounded by, the 
original CB, which tends to ensure there are no substantial population clusters 
located far inside a CB boundary; and 4) perhaps most significantly, the 
important role that roads play in the Census Bureau’s definition of CB 
boundaries.  Concerning the last of the points, the six top entries in the 
rank-ordered hierarchy of features that are used to define CB boundaries are as 
follows:
Must-hold census block boundary; 
Water areas;
Named, addressable divided roads;
Named, addressable undivided roads;
Unnamed, addressable divided roads; and
Unnamed, addressable undivided roads. 

Thus, roads figure prominently in the definition of CB boundaries, suggesting 
that 1) the CB boundaries are highly preferentially located along roads; and 2) 
significant roads are unlikely to be located internal to a CB.  Furthermore, 
the first two elements on this list would create boundaries where there are no 
roads, so that the assumption customers are located on the boundaries has the 
effect, if any, of creating additional customer dispersion and additional 
distance. 

Q. PLEASE SUMMARIZE THE CUSTOMER CLUSTERING PROCESS.
This process is described in Section 5.5 of RAM-2.  Its purpose is to identify 
all the customer locations determined from the customer location that are close 
enough together to be efficiently engineered as a single telephone plant 
serving area.  Customer locations  must meet the following criteria to be 
considered members of  a particular cluster:
No point in a cluster may be more than 18,000 feet distant (based on right 
angle routing) from the cluster’s centroid;
No cluster with a non-zero area may exceed 1,800 lines in size; and
No point in a cluster may be farther than two miles from its nearest neighbor 
in the cluster.

Clusters so identified are classified as “main clusters” if they have five or 
more customer locations, and “outlier” clusters if they have less than five 
locations.  
Once main clusters are identified in this fashion, the clustering algorithm 
calculates and records the location, area, and aspect of a rectangle that has 
the same centroid, area, and North-South vs. East-West aspect ratio of the 
original main cluster shape.  Thus, customers belonging to main clusters end up 
within the confines of a “rectangularized” cluster shape that allows the model 
to practically estimate the type and amount of outside plant required to serve 
each cluster.  Outlier clusters are assumed to lie along roads, and customers 
in those outliers are served by a “linear” cable configuration appropriate for 
such a situation.  The cluster type and shape information, as well as other 
data about each cluster as listed in the Cluster Input Data Table in Section 
6.1.1 of RAM-2, become the demographic input data for the Model calculations.

In summary, the customer location and clustering approach is fundamentally 
different from any approach that uses arbitrarily-determined geographic 
delineators, such as CBs, CBGs, or latitude and longitude grid cells.  Thus, 
while the CBG approach of HM 4.0 and earlier versions of the Hatfield Model was 
the best available when those versions were developed and, in my opinion, gave 
overall results that were as accurate as possible, the HM 5.0a results are 
likely to have a greater degree of accuracy and certainty associated with them 
at a more granular level of detail.

Q. WHAT DOES THE MODEL DO WITH THE INFORMATION THAT RESULTS FROM THE LOCATION 
AND CLUSTERING PROCESS?
HM 5.0a treats each main cluster identified during the clustering process, 
along with its associated “outlier” clusters, as a serving area.  Thus, as 
described in Sections 6.3 and 6.4 of RAM-2, the model extends copper or fiber 
feeder cable to each main cluster.  From there, distribution cable, including 
an arrangement suitable for long loops where necessary, extends throughout the 
main cluster, and to and within outlier clusters that subtend the main cluster.

Q. PLEASE SUMMARIZE WHERE IN THE EXHIBITS TO YOUR TESTIMONY THE CUSTOMER 
LOCATION AND CLUSTERING ALGORITHMS ARE DESCRIBED IN DETAIL.
Sections 5.3 and 5.4, as well as Appendix C, of the Model Description contained 
in Exhibit RAM-2 provides a detailed portrayal of the customer location and 
clustering algorithms, respectively.  These processes in turn require an 
accurate determination of LEC wire centers and customer line counts by type; 
the process of obtaining these data is described in Sections 5.1 through 5.3.  
Sections 6.1 and 6.2 describe how the Model engineers outside plant to provide 
service to the identified clusters of customers. 

Q. IN THE DOCUMENTATION THE JOINT SPONSORS OF BCPM FILED WITH THE FCC ON 
DECEMBER 11, 1997, THEY CLAIMED THAT “BCPM 3.0’S CUSTOMER LOCATION ALGORITHM IS 
SIGNIFICANTLY MORE PRECISE IN LOCATING CUSTOMERS THAN THE CUSTOMER LOCATION 
APPROACHES USED IN HATFIELD 5.0 AND THE HYBRID COST PROXY MODEL (HCPM).” 
Submission of the BCPM3 Model by Bellsouth Corporation., Bellsouth 
Telecommunications, Inc., U S West, Inc., and Sprint Local Telephone Companies, 
Introduction, at p. 6.  DO YOU AGREE WITH THIS STATEMENT?
No, I do not.  There are substantial flaws in the mechanism BCPM3.0/3.1 uses to 
“locate” customers and then provide outside plant to serve them.  First, it 
actually does not locate any customers whatsoever, because it uses road 
locations, not customer locations.  Second, it assumes that customers are 
uniformly distributed along those roads, so that road distance in a given 
“ultimate grid” can be used as a surrogate for the number of customers in that 
grid.  A cursory examination of geocoded data, as well as everyday experience 
driving through rural areas, demonstrates this is an unreasonable assumption.   
Third, even accepting the use of road locations as a surrogate for customer 
locations, BCPM does not provide outside plant to those locations.  Rather, in 
each quadrant of each “ultimate grid” in the model, it arbitrarily moves the 
customers from the roads on which they are allegedly located to a different 
area centered on the road centroid of the quadrant.   The road centroid is a 
mythical point that may have no physical meaning – if more than one road is 
present in a grid, the centroid may not even lie on one of those roads, but be 
located between roads -- and whose use may therefore cause the model to deploy 
plant inefficiently.  Finally, it arbitrarily assumes that the size of the new 
customer cluster is 1,000 feet times the road distance, compacting the 
customers it estimates to be present into an arbitrarily-sized cluster.  Even 
while doing all of this, the Model attempts to create an illusion of a 
precision in locating customers that goes beyond any level of precision 
available today.  Based on that illusion, the BCPM Joint Sponsors would have 
regulators believe that BCPM is significantly more precise in locating 
customers.

Q. MORE RECENTLY, SPRINT HAS CLAIMED IN EX PARTE FILINGS WITH THE FCC, AND IT 
AND OTHER ILECS HAVE REPEATED IN STATE PROCEEDINGS, THAT THE CUSTOMER LOCATION 
AND CLUSTERING ALGORITHMS IN HM 5.0A CONTAIN A “SERIOUS FLAW.”  WHAT IS THE 
NATURE OF THIS CLAIM?
A. According to Sprint, “The HAI preprocessing and distribution module take 
known customer locations and move them closer together.  In many cases, the 
distance they are moved is literally miles.  By ignoring actual customer 
locations and moving them closer together, the model builds less distribution 
plant than is really needed to serve customers”.  Testimony of Brian Staihr 
contained in Supplemental Direct Testimony of Brian Staihr, Jim Dunbar, Kent 
Dickerson, Talmage Cox, and Joseph H. Page, filed in the Compliance Proceeding 
for Implementation of the Texas High Cost Universal Service Plan, PUC Project 
No. 18515, Public Utility Commission of Texas, June 5, 1998.
Q. PLEASE RESPOND TO THIS ALLEGATION.
I have attached an ex parte presentation made by AT&T and MCI to the FCC on 
June 9, 1998, that addresses this allegation.  It is accompanied by charts that 
are referenced in the presentation.  I can briefly summarize this material as 
follows: 
Notwithstanding claims and concerns to the contrary, the model provides 
sufficient distribution plant to reach customers in the lower density zones, 
and potentially a little too much in the higher density zones (Slide 2) ;

The claims and concerns are based on a misunderstanding of how HM 5.0a locates, 
clusters, and serves customers (Slides 3-6);

PNR is now able to reorient the rectangles that represent customer locations so 
those rectangles better represent customer locations rather than always having 
a N-S/E-W orientation, thereby addressing the one valid point Sprint has made, 
but doing so has a negligible impact on the results for all areas tested 
(Slides 7-8 and Chart 1); 

It is surprising that Sprint has raised the issue in the first place, because 
even the existing HAI “rectangularization” of customer locations is superior to 
the BCPM’s treatment of customer locations as consisting of squares, and the 
new PNR capability will further increase the superiority of HM 5.0a over BCPM 
(Chart 9);

While Sprint focuses on one aspect of the HM 5.0a engineering algorithms that 
could, in limited circumstances, provide too little distribution cable to reach 
all the customer locations within a main cluster, it ignores offsetting aspects 
that together ensure there is sufficient cable, such as the treatment of 
outlier clusters, the conservative treatment of surrogate locations, the PNR 
practice of offsetting actual locations from the roads on which they are 
located, and the fact that the Model assumes no additional clustering within 
the boundaries of a cluster (Slides 10-22 and Charts 2-3)

In particular, to properly compare HM 5.0a distribution distances with “actual” 
customer locations, one must consider not only the branch and backbone cable 
produced by the model, but the drops as well, and if this is not done (as 
Sprint has not done), then the result is an “apples to oranges” comparison 
(Slides 16-20);

Specifically, Sprint’s comparison of HM 5.0a distances to the so-called Minimum 
Spanning Tree (MST) connecting customer locations is flawed because it ignores 
drops, the conservative dispersion of customer locations, and the fact that the 
MST, as Sprint defines it, is not the minimum amount of cable that is needed to 
connect customers together, and furthermore, it has relied entirely on a 
theoretical concept rather than presenting actual ILEC data on distribution 
route distances (Slides 21-29 and Charts 4-5);

An analysis by FCC Staff member Jeffrey Prisbey has been used by Sprint in a 
way that Prisbey did not intend (Slides 30-32); and

Furthermore, when the Prisbey analysis is corrected, it demonstrates that the 
cable lengths engineered by HM 5.0a are adequate to reach all customers (Slides 
33-41 and Charts 6-16).

In summary, there has been no meaningful demonstration by Sprint or any other 
party that there is a flaw in HM 5.0a that causes the model to  build less 
distribution plant than is really needed to serve customers .  On the contrary, 
all of the analyses we have done as a result of these claims provide further 
confidence that the model is behaving in an appropriate fashion.
 III. CONSISTENCY OF HM 5.0A WITH  
WITH FCC AND THIS COMMISSION’S CRITERIA


Q. WHAT ARE THE COSTING METHODOLOGY CRITERIA ESTABLISHED BY THE UNIVERSAL 
SERVICE ORDER AND FNPRM?
A. In summary, the FCC's criteria is as follows:
The technology assumed in the cost study or model must be the least-cost, 
most-efficient, reasonable technology for providing the supported services that 
is currently being deployed.  However, the model should:

   - use existing ILEC wire center locations;

- use a loop design that should not impede the provision of advanced services;

- have wire center line counts equal to actual ILEC wire center line counts;

- reflect the incumbents’ actual average loop length.

Any network function or element, such as loop, switching, transport, or 
signaling, necessary to produce supported services must have an associated cost;

Only long-run forward-looking economic costs may be included.  Accordingly,

the period considered must be long enough that all costs are variable and 
avoidable;

costs must not be the embedded cost of the facilities, functions, or elements;

costs must be based on current cost of purchasing facilities and equipment