[Federal Register Volume 63, Number 79 (Friday, April 24, 1998)]
[Notices]
[Pages 20392-20404]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 98-10687]
-----------------------------------------------------------------------
DEPARTMENT OF ENERGY
Federal Energy Regulatory Commission
[Docket No. PL98-6-000]
Inquiry Concerning the Commission's Policy on the Use of Computer
Models in Merger Analysis; Notice of Request for Written Comments and
Intent To Convene a Technical Conference
The Federal Energy Regulatory Commission (Commission) hereby
announces that it is requesting comments on the use of computer models
in merger analysis and intends to convene a public conference to
discuss this matter. The purpose of this inquiry is to gain further
input and insight into whether and how computer models should be used
in the analysis of mergers, including whether computer models can be
useful in a horizontal screen analysis that follows the Appendix A
guidelines of the Merger Policy Statement.1
---------------------------------------------------------------------------
\1\ Inquiry Concerning the Commission's Merger Policy Under the
Federal Power Act: Policy Statement, Order No. 592, FERC Stats. &
Regs. para. 31,044 (1996), order on reconsideration, 78 FERC para.
61,321 (1997) (Policy Statement).
---------------------------------------------------------------------------
We are issuing this request concurrently with the Notice of
Proposed Rulemaking on Revised Filing Requirements Under Part 33 of the
Commission's Regulations (Docket No. RM98-4-000). In that NOPR we
identify the use of computer models as an emerging issue in the
analysis of mergers. We are issuing this notice concurrently in order
to inform the Commission's understanding of the current and likely
future role played by computer models in merger analysis. The
attachment to this notice provides a framework for discussion of models
and includes a sample model intended to serve as a starting point for
discussion and comment.
I. Introduction
The use of computer models--specifically, computer programs used to
simulate the electric power market--has been raised in comments on the
Policy Statement and also in specific cases. In comments on the Policy
Statement, the Department of Justice (DOJ) recommended using computer
simulations to delineate markets. DOJ also noted that these simulations
could be helpful in gauging the market power of the merged
firm.2
---------------------------------------------------------------------------
\2\ Appendix to DOJ Merger NOI Comments at A-11, n12.
---------------------------------------------------------------------------
In Primergy, the applicants used a computer simulation in their
market power analysis. We did not accept the results of this computer
simulation, in part because we felt that the model was not properly
structured or tested. However, it was not our intention to inhibit the
use of computer models. We emphasized that ``we do not wish to
discourage the development of computer models for use in merger
analysis''.3
---------------------------------------------------------------------------
\3\ Wisconsin Electric Power Company, et al. (Primergy), 79 FERC
para. 61,158 at 61,694 (1997).
---------------------------------------------------------------------------
The Commission continues to believe that a properly structured
computer model could account for important physical and economic
effects in analyses of mergers and may be a valuable tool to use in
horizontal screen analyses. A computer model could be particularly
useful in identifying the suppliers in the geographic market that are
capable of competing with the merged company. A computer model may also
provide a framework to help ensure consistency in the treatment of
those data in identifying suppliers in a geographic market.
Two important ways in which a computer model could improve the
accuracy of the delivered price test are: (1) by explicitly
representing economic interactions between suppliers and loads at
various nodes in the transmission network and (2) by accounting for the
transmission flows that result from power transactions. We discuss
these and other matters in greater detail in the Attachment.
Interactions between suppliers and loads. In competitive markets
for electric energy, decisions about what suppliers would serve what
loads are likely to be driven by short-run marginal costs, including
the opportunity cost to suppliers of serving one load rather than
another. Because there can be many possible combinations of supplies
and loads, some form of computer model could be helpful in estimating
such combinations.
Transmission flows from exchanges of power. Because of the
properties of electric power flows, exchanges of power between control
areas affect flows throughout the transmission grid. Any reasonable
approximation of these effects may require a computer model to make the
many calculations needed to simulate the electric power flows.
Developing and using a computer model involves a number of choices
about the structure of the model, the level of detail reflected in the
model, the sources of information, and other issues. These issues are
discussed in the Attachment. If these technical aspects of model design
and development can be addressed adequately, a computer program could
be helpful in defining geographic markets. One common approach to
market simulation, discussed further as an example in the Attachment,
is to model the dispatch of generation to meet loads in the
transmission network. The simulation model in the example estimates
market outcomes that minimize the total cost of generation and
transmission. The contribution of such a program to a delivered price
analysis is illustrated by briefly describing the output information
that the model could provide. Typical output from a program could
consist of the following:
Generation levels. The computer model would show the level
of output of each generator.
Power traded. The model would show the net quantity of
power traded between interconnected areas 4 under economic
dispatch.
---------------------------------------------------------------------------
\4\ Typically, the interconnected areas would be control or
planning areas, but the exact geographic area would depend on how
the model was implemented.
---------------------------------------------------------------------------
Flows on the transmission grid. The model would show the
quantity of power flowing through each transmission facility
represented in the model, constrained by any transmission capacity
limits that have been input to the model. The effects of binding limits
would be reflected in model output of generation levels and power
prices.
Prices for power. For each area, the model would show the
marginal cost of power. This price can also be interpreted as the
market-clearing price for the area.
II. Request for Written Comments
If a computer model were available to produce the types of output
described above, we believe that its use could both enhance and
potentially expedite delivered price analyses. However, the
[[Page 20393]]
Commission also recognizes that there are many technical and procedural
questions that need to be addressed concerning whether and how to use a
computer model in merger analysis. To assist in the discussion of these
issues, the attachment presents an overview technical discussion,
followed by a list of questions for comment. These questions are
organized into five areas: basic model structure, alternative
implementations of the basic structure, data issues, application of
models to merger analysis, and model development and maintenance. All
interested persons are invited to submit written comments (not to
exceed 25 pages) on these questions and any other issues that the
Commission should be considering with regard to computer models and
merger analysis. Comments must be filed on or before June 14, 1998, in
Docket No. PL98-6-000. All comments will be placed in the Commission's
public files and will be available for inspection or copying in the
Commission's Public Reference Room during normal business hours.
Comments are also accessible via the Commission's Records Information
Management System (RIMS).
III. Intent To Convene Technical Conference
The Commission intends to convene one or more technical conferences
to discuss the use of computer modeling. We will issue a notice of
conference at a later date.
By direction of the Commission.
Linwood A. Watson, Jr.,
Acting Secretary.
Attachment: Computer Modeling and Merger Analysis
The purpose of this attachment is to present a sample computer
model as a starting point for discussion of issues and questions
about how such models could be helpful in merger analysis,
specifically in reference to the Commission's delivered price test
and potentially in other aspects of merger analysis. This attachment
is a Commission staff paper intended to facilitate technical
discussion. Specific comments on the sample model should be
considered in light of the questions raised at the end of this
attachment.
Background and Organization of Attachment
This Attachment discusses computer models and their use in
merger analysis. A computer model is a computer program designed to
implement a specific mathematical procedure. The specific procedures
discussed here are typically called ``models'' because they are, or
at least contain, abstract representations of real world processes.
We concentrate here on two such processes: power markets and
electric power flows over transmission networks. Computer models
hold great potential in merger analysis because they can simulate
both market processes and the electric power flows that results from
market processes.
Computer models of electricity markets and networks have many
potential uses, but we are primarily concerned here with how the
market simulations produced by such models can be used in performing
a delivered price test described in the horizontal analysis section
of this NOPR. In the context of a delivered price test, computer
models--in the sense of simulations of markets or electricity
networks--must be distinguished from other types of computer
programs. A wide range of computer programs could be used to
automate parts of the delivered price test. For example, a computer
program could be used to identify all generating units that could
supply a destination market at a particular price, given the
variable cost of power at each plant, and the transmission cost to
the destination, as inputs. Such a program would not typically be
called a model, because it does not simulate either market
interactions or electricity flows.
For purposes here, the computer models for our consideration can
be grouped into three broad categories:
Electricity Market Models. These models simulate
electricity production and trade between regions, but do not attempt
to represent the underlying electricity network in the model.
Examples of such models include the Electricity Market Model (EMM)
from the Energy Information Administration (EIA), and the more
detailed Policy Office Electricity Model (POEMS) developed for the
Policy Office of the Department of Energy.
Electric Power Production/Transmission Power Flow
Models. Generally, these are detailed models that simulate electric
power generation and/or electric power transmission, but do not
attempt to represent the market interactions or power trade between
regions. There are several models that implement standard power flow
simulation techniques.1 Detailed production cost models
(e.g., PROMOD and GE-MAPS), when they are designed for detailed cost
analysis of a single utility, could also be placed in this category.
---------------------------------------------------------------------------
\1\ For example, the FERC Office of Electric Power Regulation
uses a load flow program called PSLF from General Electric that is a
package of programs handling loadflow, fault analysis, and stability
calculations.
---------------------------------------------------------------------------
Hybrid Models. Hybrid models combine a market
simulation component with an electricity production and transmission
component. We know of no standard model designed specifically for
this purpose. Some production cost models, such as GE-MAPS, have
been expanded beyond single utility territories and used as
simulations of a competitive regional electricity market. However,
these models remain highly detailed and may be more difficult to use
for simulating electricity market trading of electricity over large
regions than a regional market model with a more aggregated
representation of the power transmission network. We seek comment on
currently available models in the questions at the end of this
attachment.
For examining the competitive aspects of mergers, hybrid models
are the computer models of interest, because both market processes
and actual power flows are important for the analysis. To understand
the role of a computer model in the analysis, it is essential to
distinguish between the computer model itself and its application. A
run of the computer model simulates power generation and power
transmission for a particular scenario. The outputs from the
simulation are then applied to a particular problem--for example,
power generation and transmission levels from the simulation output
might be used in the identification of suppliers in a delivered
price test. In this attachment, we will restrict the use of the term
computer model to the first function--simulating results for a
particular scenario--but also discuss how these simulation results
could be used in a delivered price test. In addition, we seek
comment on other potential uses of a computer simulation model in
the competitive analysis of mergers.
This attachment describes one type of computer simulation model
we have been considering and its potential use in merger analysis.
It then raises a series of questions about the framework and
examples presented. These questions are intended to serve as a guide
for commenters and perhaps for discussion at technical conferences
on computer modeling and merger analysis. The Attachment is
organized into five sections, as follows:
Overview of a modeling framework for electric power
trading over a transmission network. This framework is presented to
facilitate a discussion of whether the Commission should consider a
computer model for use in the analysis of mergers, and what role a
computer model, if utilized, should play in the analysis.
Description of a simple model implementing the general
framework, presented both qualitatively and as a mathematical
formulation. The purpose of this simple example is to provide a
structured starting point for technical questions about the design
and development of a more complex simulation model for use in merger
analysis.
Data considerations in model implementation using
currently available public sources of data. This section discusses
the data needed for a computer model and the availability and
limitations of publicly available data.
Application of a computer model in merger analysis.
This section addresses the question of how computer model simulation
runs would play a role in a delivered price test.
Questions for discussion at a technical conference or
conferences. These questions extend the earlier discussion by asking
questions about the design and development of the framework and
sample model, how a model should be used in the competitive analysis
of mergers, what data sources are available, and how the Commission
should proceed in developing and maintaining a model.
Overview of Model Structure
The role of computer modeling in merger analysis can be
identified by first reviewing
[[Page 20394]]
the Commission's delivered price test. For a delivered price test,
applicants are expected to estimate the cost of economic
transactions to acquire power and transmit it to a destination, and
also to determine how much power is available to be generated and
transmitted to a destination, given the limitations on power
transactions imposed by the transmission system. For example, given
a particular destination market, an applicant should:
Determine an appropriate competitive price for
wholesale electric power in that destination market that is
consistent with available information, and adequately support the
method used to determine the price.
Estimate the available generating capacity and variable
cost of wholesale electric power from potential supplier facilities
at the level of individual generating units to the extent possible.
Estimate the cost of transmitting power (including
ancillary services) from the source of generation to the
destination, using maximum applicable tariff rates or other
conservative estimates that can be supported.
Make other adjustments, as appropriate, to reflect a
supplier's competitive presence in a destination market, and support
such adjustments with adequate analysis, data and assumptions, and
Evaluate the impact of transmission system limitations
on the ability of potential suppliers to deliver power to the
destination market, using simultaneous estimates of transmission
capacity limits to the extent possible.
These requirements help delineate a framework for analyzing
electric power transactions over a transmission network. This
process of analysis can be made more explicit by first constructing
a general representation of the analysis and then incorporating this
general picture in a mathematical formulation of the economic
problem and the constraints imposed by the physical electricity
transmission system limits. Figure 1 gives a general representation
of the problem of combining the analysis of electric power
transactions with an analysis of the physical limitations imposed by
the electric transmission grid. The upper diagram represents the
economic network of power transactions, that is, the production and
consumption of power in each area, as well as trades of power
between interconnected areas. The amount of trading that occurs
among areas depends on the load requirement of each area, on the
price and availability of power in each area, and also on the cost
of transmitting power between the areas. The lower diagram
represents the actual physical transmission network in which these
economic transactions occur. It would comprise primarily the
transmission lines and transformers that are called ``flowgates.''
Transactions between areas (in the upper diagram) cause flows across
these flowgates in the physical network (in the lower diagram).
These flows are then subject to the actual physical limits imposed
by the electric transmission network.
Most of the key elements in the Figure 1 are the same elements
that would need to be considered in a delivered price test without a
computer model. In order to explain the structure shown in Figure 1,
we explain these common components first:
Areas. These are locations in the transmission network where
electric power is injected by generators and withdrawn by loads.
Although in principle they can be any part of the network for which
generation and load data are available, in practice they often
correspond to control areas. In any case, the considerations that go
into defining the locations of generating plants and loads can be
the same, whether or not a computer model is used to conduct a
delivered price test.
Generators. In Figure 1, the generators located in each area are
shown as supply curves. In the model, the width of each step on the
supply curve would correspond to the capacity of a specific
generator located in an area. The height would correspond to the
variable cost of power from that generator. To construct a supply
curve, generators may be arranged in order of the variable cost of
generation, just as they would be for a delivered price test without
a computer model. Supply curves can be constructed in others ways,
and we seek comment on such alternatives.
BILLING CODE 6717-01-P
[[Page 20395]]
[GRAPHIC] [TIFF OMITTED] TN24AP98.000
BILLING CODE 6717-01-C
[[Page 20396]]
Loads. Loads in Figure 1 represent demands to be met by
generating power and transmitting it over an electricity network.
Although a computer model of power transactions would be expected to
include more than just destination market loads explicitly
considered in setting the destination market price, the information
sources for these loads should be the same as the sources for a
delivered price test without a computer model.
Power Transactions/Area Interconnections. The specification of
interconnections and the cost of transmitting power between areas
included in the analysis should be the same with and without a
computer model. In particular, transmission prices should represent
a conservative estimate of the cost of transmitting power (e.g., by
using maximum tariff rates).
As noted above, a computer model of market interactions would
contain more loads than just those at a particular destination. To
be adequate, it should represent all relevant loads that would have
a significant impact on the market for power at a destination. This
type of computer model could then calculate the suppliers'
opportunity cost of selling power, and market prices that reflect
these opportunity costs, because the cost of power at each
destination would be considered in the model. Although this
opportunity cost can be informally considered as an adjustment to a
supplier's competitive presence when doing a delivered price test
without a model, a model removes the ambiguity in this informal
consideration by explicitly calculating the opportunity cost.
A computer model should also represent the physical electrical
network and model the relationship between power transactions and
actual power flows and the limitations on power transactions that
must be imposed when actual power flows approach transmission
capacity limits. These two considerations--the relationship between
electric power trading and physical power flows, and the effect of
transmission capacity limits--should be included in any analysis of
a merger to the extent that information is available. One value of a
simulation model lies in incorporating both of these considerations
in the computer program, where the needed calculations can be
performed in an efficient, standard way. The treatment of
transmission flows and limits in the computer simulation model are
discussed in more detail below.
Estimating Transmission Flows from Power Transactions. The model
structure presented in Figure 1 shows the link between transactions
and transmission using power transfer distribution factors (PTDFs).
As shown in Figure 1, these factors are used to superimpose the
effect of power transactions shown in the upper diagram on the
underlying electricity network shown in the lower diagram of the
figure. These flows may be on individual lines or groups of
lines.\2\ The lines represented in a computer model may correspond
to tie lines between areas, but they may also correspond to other
lines in the transmission network that are internal to areas and not
part of an interface between areas.\3\
---------------------------------------------------------------------------
\2\ Groups of lines are referred to here as ``flowgates,''
discussed further below.
\3\ For example, in the DC flow model used by the NERC to
generate the draft PTDFs, 20 transmission lines make up the flowgate
representing the interface between APS and PJM, 12 lines represent
the interface between APS and AEP, 3 lines make up the interface
with Ohio Edison, 3 lines make up the interface with Duquesne and 7
the interface with Virginia Power. In addition to tie line
flowgates, the NERC model includes 34 flowgates representing lines
internal to the APS control area.
---------------------------------------------------------------------------
Figure 2 shows how the PTDFs are applied. The exchange of power
between areas shown on the left side of the figure corresponds to
the injection of power (100 MW in the example) into the transmission
grid in Area 1 and the withdrawal of the same quantity of power in
Area 2.4 Because of the nature of the electricity flows
in networks, this exchange of power induces flows on all lines in an
interconnected grid. While a precise estimate of the electricity
flows from a specific change can only be determined from a
complicated power flow model, the flows can be approximated by a
standard modeling technique, known as the DC Load Flow
model.5 Distribution factors can be used to capture the
DC Load Flow estimates as shown in Figure 2. The quantity of flows
on each line in the actual transmission network is estimated by
multiplying the quantity exchanged by a PTDF. For example, 70 MW of
the 100 MW power (a PTDF of 0.7 times power trade 100 MW) exchanged
between Area 1 and Area 2 flows on the lines from Area 1 to Area 2.
---------------------------------------------------------------------------
\4\ For purposes of the example and discussion, we are ignoring
losses.
\5\ See Fred C. Schweppe, Michael C. Caramanis, Richard D.
Tabors and Roger E. Bohn, Spot Pricing of Electricity, Kluwer
Academic Publishers, Boston, 1988. Appendix D describes the DC Load
Flow.
---------------------------------------------------------------------------
The Distribution Factor Task Force of the North American
Electric Reliability Council (NERC) estimates PTDFs for input into
the interim Interchange Distribution Calculator (iIDC).6
A computer program for market and merger analysis could use these
PTDFs, but other forms of distribution factors are standardly used
in DC load flow analysis.
---------------------------------------------------------------------------
\6\ NERC plans to use the iIDC to support a flow-based
transmission reservation and scheduling process and line loading
relief procedures. In response to an NERC Board of Trustees
recommendation, the Engineering Committee and Operating Committee
approved the creation of a Transmission Reservation and Scheduling
Task Force to ``develop a process for the reservation of
transmission services and scheduling of energy transfers recognizing
the actual use being made of the Interconnection''. The task force
developed a detailed recommendation for a flow-based transmission
service methodology (FLOBAT) based on flowgates and PTDFs. See
``Transmission Reservation and Scheduling, Transmission Reservation
and Scheduling Task Force'', Report to the Board of Trustees,
December 12, 1996.
[[Page 20397]]
[GRAPHIC] [TIFF OMITTED] TN24AP98.001
We seek comment on the most appropriate source for information
on distribution factors for modeling purposes.
Transmission Capacity Limits. NERC has compiled distribution
factors for the Eastern Interconnection 7 that relate
control area power exchanges to flow across area tie lines and their
corresponding flowgates. These flowgates are groups of transmission
facilities that are monitored for security purposes. Using these
factors, it should be possible to model flows at points in the
transmission system that are most likely to constrain the economic
use of the transmission grid. These flows become important for
market analysis when any flows reach a physical limit on the
flowgate. When the limit is reached, power must be redispatched if
the destination loads are to be met. Redispatching power means
changing which generating units produce power, so that power
generation does not cause transmission flows to exceed the limit on
the flowgate.
---------------------------------------------------------------------------
\7\ The Eastern Interconnection is the portion of the
transmission grid that covers the eastern part of North America,
extending from the Rocky Mountains to the Atlantic Ocean (but
excluding the Electric Reliability Council of Texas (ERCOT)).
---------------------------------------------------------------------------
The physical limit on a flowgate is not a simple, static
quantity. Flowgate limits are set for individual elements of the
transmission network to assure they are not operated beyond safe
loading, depending upon such cond itions as thermal limits,
generating resource availability, line outages, loop flow, stability
and voltage conditions, and so on. Because the limits reflect system
conditions at any point in time, the limits are dynamic and care
must be exercised if single quantities limits are used in a computer
model. These considerations about the nature of transmission limits
are not limited to the particular example of flowgates; they apply
as well to the Total Transfer Capability (TTC) and Available
Transfer Capability (ATC) quantities posted on OASIS. We focus here
on flowgate limits because they appear to be the limits most
directly related to the distribution factors used to estimate
network flows. Other approaches to estimating physical flows and
associated limits are possible; we ask questions about such
approaches in the last section of this attachment.
NERC is developing an Interregional Security Network (ISN) that
may include data on flowgate capacities, but these limits are not
currently available. Estimates of the capacity limits of these
flowgates are important data for the implementation of a model based
on that network. The availability of these limits would be of
considerable value even if a model is not used, since they could be
used to estimate limits on transmission flows for many types of
analysis of transmission grid transactions, including conducting
delivered price test without a model.
Specification of a Simple Model
The two main benefits of implementing the electric power
modeling framework through a computer program are: (1) Better
representation of the market interactions, in particular the
opportunities presented to suppliers by the presence of other loads
in addition to the loads at the destination market and (2) better
representation of the impact that transmission limits will have on
economic transactions. In order to make the general structure
specific for use in a computer program, the mathematical structure
of the algorithm must be described and the data used as input to
this algorithm must be specified. As a starting point for
discussion, this section describes an algorithm that can be
implemented using most standard mathematical programming software
packages. The algorithm is described qualitatively and also
presented as a mathematical formulation.
The problem solved in this example is finding the lowest cost
combination of supplies (generating plants) and power transactions
between areas, to meet fixed demand (loads) over an electricity
transmission network, given costs for power, charges for
transmission of power within and among areas,8
transmission loss factors, and physical limits to moving the power
over the grid. Solving this cost minimization problem simulates the
actions of a competitive market. Under this least cost dispatch,
buyers of power can't make any more trades among suppliers to lower
their purchase costs. This is the expected result in a purely
competitive market, where buyers have alternatives and are permitted
to trade among these alternatives until they get the best value for
their money.
---------------------------------------------------------------------------
\8\ As discussed above (page 4), these areas would typically be
control areas. Since the sample model is general, we drop the
specific qualifer.
---------------------------------------------------------------------------
In the ``real'' world, conditions are more complex than in a
computer program. The clearest differences between generation and
transmission in the computer program and the real world are
assumptions about information (the model assumes it is perfect and
costless) and the cost of transactions (the model assumes no costs
for searching for
[[Page 20398]]
suppliers, negotiation of trades, or costs of interruption.) The
computer model makes any trade that can lower costs, even if it
involves large and complicated combination of individual trades
among buyers and generators across a transmission network. Even
simple transactions are assumed to involve only variable costs of
generation and maximum transmission rates.
While these idealizations are limitations, some idealizations of
this sort are inevitable, and point out the need to view computer
simulation model as a tool in an overall analysis. These issues can
be addressed with model runs where assumptions change--i.e., by
conducting sensitivity analysis under different scenarios. In
addition, computer program results need to be validated by checks
against other sources of market information before making use of the
outputs from the program.
The model specified here is a basic model that could be used to
examine electric power transactions and transmission flows. This
model is presented as a ``strawman'' point of departure for
discussion. It represents only a single period solution of the
problem, that is, it does not attempt to address startup costs or
other multiple period effects. It also includes some parameters as a
single constant that may need to be varied across areas, for
example, adjustments for losses. Further, other factors would need
to be addressed through adjustment of input data (for example,
through adjustments to plant capacities for availability in each
time period analyzed). These issues will be raised below in the
section on issues and questions for comment. However, even without
such modifications, staff believes that this basic model does
capture important market and transmission effects. Even the use of a
simple model, not much more complex in structure than the model
presented here, could potentially enhance the delivered price test
and expedite the analysis of mergers, if data are available to
implement the model. In the next section we discuss data issues
related to this implementation.
The objective of the model, the constraints that must be met in
reaching this objective, and the model inputs and outputs are
described below. The model is stated mathematically in Figure 3.
Model Objective. Minimize the total cost of delivered power,
calculated as the sum of generation and transmission costs to meet a
fixed set of demands (loads) in each area, given costs for power
generation in each area and rates to transmit the power between
interconnected areas.
Subject to constraints that satisfy:
Generation capacity requirements. Generation does not exceed a
maximum capacity for each unit or fall below a minimum level if one
is specified.
An energy balance in each area. The sum of generation in each
area plus power imported from other areas over the transmission
network, adjusted for losses in generation and transmission, is
equal to the demand in each area.
Flowgate requirements. The flow across the flowgates defining
the electricity network does not exceed the maximum flowgate
capacity or fall below the minimum flowgate level if one is
specified.
Transmission system balance requirements. The total power
injected into the transmission system equals the total power
withdrawn from the transmission system, adjusted for losses.
The model inputs needed to compute the objective function and
determine the constraints are:
The variable cost of generation at each unit in each
area.
The capacity of each generating unit in each area (and
the minimum run level if needed).
The demand (load) in each area.
The applicable transmission rate between each pair of
interconnected areas.
Power transfer distribution factors for each
interconnection between control areas.
Losses in generation and transmission.
The maximum capacity of each flowgate.
BILLING CODE 6717-01-P
[[Page 20399]]
[GRAPHIC] [TIFF OMITTED] TN24AP98.002
BILLING CODE 6717-01-C
[[Page 20400]]
This information is used to determine the generation levels and
transmission interchange between control areas that minimizes the
sum of generation costs and transmission charges as specified above
in the objective function. The key outputs from this algorithm are:
Power production at each generating unit in each
control area.
Net power interchange between areas.
Power flowing on each flowgate.
Marginal cost of power in each area.
Implementing the Basic Model: Data Considerations
In principle, the sample algorithm in the last section could be
implemented at a high level of detail, where areas were
geographically small, for example, at a level of detail below a
utility service territory. This level of detail could approach the
level of detail used in detailed power flow and transmission system
analysis. In practice, data limitations may make a such a detailed
model generally impractical as a screening tool for merger analysis
(although in specific cases, more detail can be developed as
needed). A reasonable starting point for data considerations is the
information currently required to conduct a delivered price test. As
discussed above, one would expect many of the sources of information
used for computer modeling to be the same as the sources for the
non-model application of the delivered price test. Variable
generation costs and capacities by area, area demands,
interconnections between areas, transmission tariff rates could be
the same in both analyses. A computer model would need data on a
larger geographic area than a delivered price test for a single
destination. However, most of the publicly available sources are not
limited to single regions, but provide nationwide coverage.
Sometimes this coverage is limited to a particular class of market
participants--e.g., Investor-Owned Utilities (IOUs), Municipal
utilities, etc. However, it is generally possible to compile
nationwide data on the key variables needed in the analysis;
consequently, data for the larger geographic areas that may be
required for a computer model should be generally available and
relatively easy to incorporate in the analysis.
The availability and format of data circumscribe the ways in
which key variables in a model can be defined. For parameters that
are common to calculations with or without a model the issues of
definition are the same in either type of analysis. As an example,
consider the question of what areas to use in an analysis. Answers
to this question depend on how data are reported geographically, as
follows:
Generator locations can be assigned to specific
geographic locations within control areas.
Tariffs are filed by utility areas (or sometimes for a
single holding company such as Southern Company).
For load scheduling purposes, interconnections are most
naturally defined by control area, and Form 714 data are reported on
that basis.
System lambda data are filed on a control area basis.
Historical loads are most easily derived from the Form
714 filings which are reported on a planning area basis.
These data limitations suggest that areas for modeling purposes
might be defined by combining control and planning areas. This
definition would permit a modeling analysis to consider different
time periods defined on the basis of hourly load data, and to
estimate the system lambda corresponding to the load data on a basis
that is consistent with the requirements for a delivered price test
without a model. Staff seeks comment on this and related issues
below.
PTDFs are needed in the model specified in the previous section,
but would not be needed if the merger analysis did not use a
computer model.10 Recall that PTDFs relate power
exchanges between areas to flows across flowgates. The sample model
assumes that the areas in the model are the same ones used to define
PTDFs. Although PTDFs are not needed in an analysis that does not
use a computer model, they are nevertheless a valuable piece of
information for any analysis that needs to examine the implications
of loop flow and transmission limits.
---------------------------------------------------------------------------
\10\ This is the only data element required for the sample model
that would not be needed without it. However, a more complex model
might impose additional data requirements. These additional
requirements are addressed in the last section of this attachment on
questions for a technical conference.
---------------------------------------------------------------------------
Transmission limits are also important data inputs to the
computer model. As discussed above, flowgate limits have not yet
been defined for the flowgates identified in the NERC data on PTDFs.
The best currently available information for estimating limits
appears to be OASIS values for Total Transfer Capability (TTC) and
Available Transfer Capability (ATC), and transmission capacities
reported in various NERC studies and other systems assessments.
Since these are the same sources that are needed for a delivered
price test analysis, the model does not impose additional data
requirements beyond those of the delivered price test. One caveat
may be noteworthy, however. A computer model may be more sensitive
to data limitations, because the model automatically enforces the
transmission system limits on electricity trade. This automatic
nature of the computer model is a great benefit if consistent and
accurate data are available, because the model can automatically
capture the effects of trade across an interconnected electricity
grid. However, this characteristic of a computer model can also make
results more sensitive to data imperfections than an analysis
relying more directly on the analyst's judgment, and suggests that
analysts should conduct studies to determine the sensitivity of
market simulations results to a range of transmission limits.
Finally, a computer model simulation is a valuable tool for
examining the consistency of the data used in the analysis. The
model uses all the same information used in the current delivered
price analyses for the key parameters: generation costs and
capacities, transmission tariffs and limits, and destination market
loads. From this information, the computer model simulates
generation levels, generation costs, control area prices, and
transmission flows between areas. It should be possible to reconcile
these simulation results with corresponding reported information.
For example, the simulation results (such as control area prices and
the costs the marginal generator) should be consistent with reported
values for system lambda. Inconsistencies may indicate deficiencies
in either the model or the information sources, or both, and large
inconsistencies need to be understood before proceeding with the
analysis. This is particularly important for system lambda data,
since the system lambda data may be used to set the destination
market prices. If estimated prices from a simulation are not
consistent with system lambda data, the cost information used in a
delivered price test (such as the generation costs reported on Form
1) may not be consistent with the destination market prices. Since
inconsistencies between estimated and reported values can also arise
because of the limitations of the model itself, however, some degree
of inconsistency may be inevitable. However, the model would still
provide a valuable tool for linking the different sources of
information used for the delivered price test and potentially
corroborating the system lambda data as a destination market price
indicator. As experience is gained in calibrating a model with other
sources of information on prices and generation levels, judgments of
what destination market prices to use in an analysis should improve.
Applying a Computer Model to Merger Analysis
The discussion has not yet considered the role of a computer
model in a delivered price test. It is important to distinguish
between the computer model itself and use of the output of the model
for merger analysis and the delivered price test. A model simulates
generation and power flows in the transmission network based on
economic and electrical engineering principles. It is then applied
to a particular analysis as defined by a particular procedure. Using
a model as a tool in this way does not alter the basic objectives or
principles underlying the delivered price test.
To assist the discussion of applying the model to a delivered
price test, we divide this section into three parts, as follows:
A Delivered Price Test Without a Model. The delivered
price test is not intended to be applied in a rigid, inflexible
manner. Accordingly, staff has tailored the basic steps described
here to fit the circumstances in each case.
Model Outputs Relevant to the Delivered Price Test.
This part briefly reviews computer modeling methods and results that
are important in the delivered price test. These features are
described without reference to technical details of model design and
data discussed in previous sections.
A Delivered Price Test With a Model. A delivered price
test with a model will follow the same basic pattern, but details of
the procedure will change. This section describes where the model
would fit in the context of a typical DPT application.
[[Page 20401]]
Staff's Framework for a Delivered Price Test Without a Model
The competitive screen analysis focuses on one aspect of merger
analysis: whether the merger would significantly increase
concentration. The four steps in the competitive screen analysis
are:
Identify relevant products.
Identify affected customers.
Identify potential suppliers to affected customers.
Analyze effect on concentration.
For purposes of comparing a delivered price test with and
without a computer model, the key step is the identification of
suppliers in the market. This step will be described in detail, but
other steps will be also be briefly described for completeness.
These descriptions are not meant as a fixed prescription, and we do
not mean to imply that there is a single way to conduct a delivered
price test. Rather, they describe a set of choices we have found
appropriate in previous cases. These choices are guidelines that
staff believes can be improved upon as analysis evolves. Their
purpose is to distill experience and provide reasonable common
ground as guidance, without restricting innovation in future
applications.
Identify Relevant Products. Although other products can be
appropriate, the relevant product for the delivered price test has
typically been short-term energy. Short-term energy has been further
differentiated by time period. For most purposes, staff has divided
time periods into nine time categories, defined by season and hourly
load conditions: winter, summer and spring/fall seasons, with peak,
shoulder and off-peak periods being identified for each season.
Short-term energy is then analyzed as a separate relevant product
for each of the temporal categories.
Identify Affected Customers. Customers have generally been
identified based on the facts of each case, the Applicants' filing,
and analyses filed by intervenors. The result has been the
identification of destination markets with higher probabilities of
negative effects. Each destination markets has been analyzed
separately for each time period.
Identify Suppliers to Affected Customers. Identifying suppliers
to each destination market in each time period involves several
choices and related calculations. The identification starts with a
decision on how to limit the total group of suppliers included; that
is, with how many ``wheels'' away a supplier must be in order to be
excluded from consideration. Generally, three wheels has been deemed
adequate, but no rigid number of wheels can be determined a priori,
so the boundaries need to be fitted to the facts of each case. The
main remaining components in supplier identification are:
Competitive price in the destination market.
Generation costs and capacities.
Transmission prices and transmission system capability.
``Native'' loads.
A general summary how each of these components has been included
in the delivered price test is given below.
Competitive price in the destination market. The destination
market system lambda provides a default indicator that can be
calculated for each of the time periods considered. However,
differences in methods underlying the system lambda and well as
differences in reporting (such as inclusion or exclusion of
purchases) mean that system lambda data should to be compared with
other indicators such as published spot prices for consistency. One
approach to the problem of uncertainty in any estimate of the
competitive price is to analyze concentration for different price
levels, in order to determine how sensitive the concentration
results are over a plausible range of prices.
Generation costs and capacities. The primary source of
information for the capacity and variable cost of generation has
been the FERC Form 1 and related forms.11 These data are
available for individual generating plants, but do not provide
information on specific units when there are multiple units at a
plant. However, it does provide information by prime mover type
(e.g., fossil steam, internal combustion) and type of fuel. For
purposes of variable cost estimation, this level of detail is a
reasonable approximation to unit level information in most cases.
---------------------------------------------------------------------------
\11\ For example, the Rural Utility Service Form RUS-12 provides
information on generators owned by cooperatives, and the Energy
Information Administration Form EIA-412 provides information on
municipals.
---------------------------------------------------------------------------
Generation capacity is adjusted for availability, based on
estimates of planned and forced outages. Planned and forced outage
rates should be based on historical outages, and varied at least by
fuel type. If more detailed data are not available on the temporal
patterns of outages, outage rates should be applied to represent
typical patterns. For example, forced outages are applied equally to
all time periods, unless another allocation can be supported.
Planned outages are assigned to spring/fall where they would be most
expected, except where more explicit scheduling patterns can be
supported.
Transmission prices. In general, staff has used firm ceiling
rates from open access tariffs. Generally, the maximum applicable
hourly rate, in $/MWh, is used. In cases where discounted rates a
generally available and posted on OASIS, these discounted rates are
used.
Transmission rate structures are undergoing changes, so no
single approach is always the best one to use. Where new rate
structures have been adopted, the new rate structure should be used.
For example, MAPP rates are distance-based, and these current
regional rates are used for transmission analysis involving MAPP
companies.
In order to determine the transmission costs for a supplier to
reach a destination market, it is necessary to trace a ``contract
path'' between the supplier and the destination market. The basic
information source for identifying the individual companies in these
interconnections has been the FERC Form 714. Where there are
multiple paths between the supplier and the destination, staff has
chosen to assign suppliers to the path with the lowest transmission
cost.
Transmission capacity. There are two different publicly
available sources that can be used to estimate transmission
capacity: NERC Regional Reliability Council transmission assessment
studies and OASIS reports of Total Transfer Capability (TTC) and
Available Transfer Capability (ATC). Staff has used both of these
sources, but the specific uses have been based on the strengths and
weakness of each source. NERC data provide better supporting detail
and can be used for estimation of simultaneous transmission
capabilities. However, NERC reports generally report simultaneous
transmission capability at the regional or sub-regional level, not
at the more detailed geographic area reported on OASIS. OASIS data
provide a desirable level of detail (the control area and some sub-
control-area detail), but the reporting is not generally on a
simultaneous basis and reporting has not fully matured. For example,
different OASIS sites report differing TTC/ATC capacities between
areas over the same path. Therefore, OASIS data, while detailed,
need to be reviewed closely for use in estimating transmission
capacity in the delivered price test.
The total generation capacity on a particular path from a
supplying area to the destination market is determined by the
suppliers assigned to that path. When the available transmission
capacity on a path is less than the total generation capacity
assigned to the path, it is necessary to allocate capacity to the
suppliers comprising the path. The merger policy statement does not
endorse any particular method for making this allocation, but the
two approaches used by staff are to reduce each supplier's capacity
pro rata and to select suppliers in order of generation cost.
Native load estimation. When the measure of capacity used is
available economic capacity, an estimate of native load in each area
is needed. This estimate is used to reduce the generation capacity
available for sales to the destination markets that are being
analyzed. For this purpose, FERC Form 714 data on hourly loads can
be used to estimate the load in each time period. Because these data
are reported on the basis of ``planning areas'', some adjustments to
these data are necessary for use in estimating native load by
control area.
Analyze effect on concentration. The final step in the analysis
is to examine the pre- and post-merger concentrations and compare
them to the appropriate thresholds. These concentrations are based
on the estimated supplier shares from the supplier identification
step, for pre-and post-merger combinations of the following cases:
Products--short term energy.
Periods--nine periods by season and load conditions.
Capacity measure--economic capacity (supplier capacity
deliverable at 105% of the competitive price) and available economic
capacity (subtracting native load from a supplier's economic
capacity).
Model Outputs Relevant to the Delivered Price Test
The steps in supplier identification described above could be
conducted using a
[[Page 20402]]
computer program that uses information on generation costs and
capacities, transmission costs and capacities, and other inputs.
Such a program would provide a list of suppliers and capacities
making up the supply to each market. Without a computer model of the
market and transmission grid, these programs cannot take into
account certain factors that are important in determining what
suppliers can deliver power economically to a particular
destination. The two main factors not accounted for are:
Interactions between suppliers and loads. In a
competitive environment, decisions about which suppliers will serve
which loads will be driven by opportunity costs, in particular the
opportunity cost to suppliers of serving one load rather than
another. Because there can be many possible combinations of supplies
and loads, some form of computer model could be helpful in
estimating such combinations.
Transmission flows from exchanges of power between
areas. Because of the properties of electricity, exchanges of power
between areas affect flows throughout the transmission grid. Any
approximation of these effects may require a computer model to make
the many calculations needed to estimate electric power flows.
Developing and using a computer model involves a number of
choices about the structure of the model, the level of detail, the
sources of information, and other issues. These issues are discussed
elsewhere in this attachment. The main question to raised here is
what information the computer program provides to the analyst. Once
this question is answered, the discussion turns to the question of
how that information can be used in a delivered price test.
For purposes of this discussion, the computer program is assumed
to be a simple representation of dispatch of generators to meet a
fixed set of loads in a single time period. The program is assumed
to simulate the economic dispatch of power over an electric
transmission network, by finding the dispatch of generators and
exchanges of power between areas that gives the lowest total cost of
producing and transmitting the power. Output from this computer
program would include generation levels, the quantity of power
exchanged between areas, flows on the transmission grid, and the
marginal cost of power in each area. Each of these computer model
outputs is described briefly below:
Generation levels. For each generating unit, the
computer model estimates the level of output of each generator. It
does not estimate which generator sells to which load, but only how
much power is generated by each generator when dispatch of that
power is at least overall cost.
Power exchanged. For each pair of interconnected areas,
the model gives the net quantity of power exchanged between the
areas under economic dispatch.
Flows on the transmission grid. For each of the
transmission facilities represented in the model, the model outputs
the quantity of power flowing through that facility. These flows
will be limited by any transmission capacity limits that have been
input to the model.
Marginal costs for power. For each area, the model
would find the marginal cost of power under economic dispatch. For
purposes of this analysis, this cost can be interpreted as the
market clearing price for the area.
These model outputs can be used to apply the model in a
delivered price analysis. This application is discussed in the next
section.
A Delivered Price Test With a Model
One use of a computer model is to use it in a delivered price
test analysis. A computer model would be used only in the supplier
identification step. The model could be helpful in two parts of this
analysis: determining the destination market price and identifying
the suppliers that can deliver to each destination market. The role
of a computer model in each of these steps is described below:
Determine destination market price. The default
approach to market price determination would still be the system
lambda data. However, a computer model could be used here to help
corroborate the price used for the destination. As discussed above
(p. 14), a computer model could be used to simulate a destination
market price for the loads in each time period. This simulated price
would not be a substitute for a price estimated from system lambda
data, but could be an additional factor in determining how to
establish the price and whether to examine a range of market prices
rather than a single estimate.
Identify suppliers to the destination market. A
computer model could be used to determine what suppliers could
deliver to the destination market. It could simulate the supplier
identification procedure of the delivered price test. In the
delivered price test, suppliers are considered in the market as long
as they can deliver to the destination market at a price less than
or equal to a threshold price equal to 5% above the destination
market price. A computer model could simulate the same test by
considering only the load in the destination market (i.e., assuming
all other loads to be zero). Under these conditions, the computer
model would be run with increasing destination market demand until
the market price reached threshold price. All suppliers running at
this price would be identified as supplying the destination market.
In addition to these steps, adjustments to supplier capacity
that can be delivered to a destination may be appropriate. One
possible adjustment could be to consider other destinations that
provide selling opportunities for suppliers and the likelihood that
supplier's opportunities may alter their capacity available for
delivery to a particular destination market. A computer model is one
tool that could be used to assess the effect of these alternatives
in a delivered price test. Staff seeks comments on whether these
types of adjustment may be appropriate in a delivered price test and
how a model could be used for this purpose.
Finally, computer models hold additional potential for
application in other areas of the competitive analysis of mergers.
In the next section, staff seeks comment on these and other issues.
Issues/Questions for a Technical Conference
Below are questions for comment and perhaps also discussion at a
technical conference. Commentors should also raise any other issues
they believe need to be considered. In considering these questions
or in raising further issues, it is important to specify whether the
model is intended primarily as a screening tool or as a detailed and
full analytical tool. In the former case the model must therefore
strike a balance between detail (with the presumption of greater
accuracy and precision) and ease of application within the
requirements for a screen.
Questions are listed in five groups: basic model structure,
implementing the basic structure, data issues, application to merger
analysis and process issues.
Basic Model Structure
The sample model assumes the general form of a mathematical
programming problem. Is this the most appropriate technique to
simulate economic equilibrium problems in the electricity market?
Please be explicit about any proposed alternatives.
The sample model is structured as a linear program. Would
another mathematical programming form be better (for example, a
quadratic program with piecewise linear supply curves)?
Demands are assumed to be fixed in the sample program, so the
demand side of the market is not represented in the sample model.
Should demands be made responsive to price? If so, what is the
appropriate price elasticity? Should the objective function then be
to maximize social welfare (the sum of producer plus consumer
surplus)?
The sample model uses distribution factors to estimate power
transmission flows. Is this approach adequate? Should Commission
staff rely on transmission distribution factors supplied by others
(either NERC or another third party) or perform its own transmission
system analysis to derive distribution factors for market analysis?
In the sample model, the generator cost functions are
represented as a constant variable cost for a unit, even though unit
efficiencies vary over the operating range of a generating unit. Is
a formulation with a constant variable cost sufficient for purposes
of a screening model? Are there alternative formulations of the cost
function that can be easily implemented with available information?
How should generating unit availabilities and losses be
represented in the model? Could availabilities be treated outside
the model, as adjustments to available capacity for each time period
studied? Should losses be represented only for transmission flows,
or for all generation and transmission, and should different loss
factors be supplied for each area? Should losses associated with
generation or load within each area be treated differently from
losses associated with transmission exchanges or flows across areas?
Should losses be transaction based or flow based?
How should generation and transmission reserve requirements be
modeled? How should transmission reserve margin (TRM) and capacity
benefit margin (CBM) be used?
[[Page 20403]]
What additional adjustments are required to account for generation
operating reserves, generation planning reserves, or transmission
reserves?
Are there other operating conditions that would need to be
represented in a model for screening purposes? For example, would a
model need to represent operating costs for startup or ramping in
order to capture whether particular unit might be available to
respond to price increases? Are there any special design
considerations for hydropower that need to be incorporated in the
model, and how can these best be added?
Alternative Implementation of Basic Model
Is a geographic level of detail corresponding to control areas
the best level of detail for purposes of a screening model? If a
greater level of detail is necessary, please explain how this detail
can be represented with public sources of data or how it can be made
part of the filing requirements. Also explain how a more complex
analysis with a detailed model could be conducted within the time
requirements of a screening analysis. If geographic areas larger
than control areas are recommended, please explain how the approach
could adequately capture competitive issues required in a merger
screen.
The model represents transactions between control areas.
Transactions between control areas follow a contract path and pay
for each control area transfer between source and destination. As
rate structures change and power pools evolve, these rate structures
will also change. What design elements should be incorporated to
ensure that the model is sufficiently flexible to accommodate these
evolving structures?
How should firm sales and contracts be represented in the
modeling structure? For example, should generation capacity be
reassigned from the selling region to the purchasing region? If
capacity is reassigned, which generating units should be associated
with the reassignment? Should the transmission capacity be made
unavailable for both scheduling and use, that is, should it be
assumed that the purchaser is obligated to use the power rather than
resell it, so capacity will be used and not available for short-term
trading in the model?
The model can simulate a market (minimize costs) over any
arbitrary area for which data are available. Should the overall area
be broad, for example, the Eastern Interconnection, or should it be
limited to a smaller area surrounding the parties to a merger?
Discuss how trade with areas outside the area represented in the
model should be analyzed and incorporated in the model.
Should different modeling structures be used to simulate the
different characteristics of power trading and power flows for
different regions? For example, is the sample model considered
equally applicable to the analysis of the Eastern Interconnection
and WSCC? If not, what key differences between regions should be
reflected in the structure of the model, and how should they be
represented?
Data Issues
Are there alternatives to using FERC Form 1 data (and data from
related public sources) for generator costs and capacities that
provide comparable geographic and company coverage?
What are the best data for estimating the fuel cost component of
variable cost? Should historical costs, such as those reported on
Form 1 be used? Or should other estimates, such as spot prices, be
used? If a single heat rate is used for each unit to convert fuel
costs to a cost per unit of electricity, should that heat rate be
taken from Form 1? Or are other heat rates, such as those filed by
unit on the Energy Information Administration Form 860, a better
estimator of the cost of power from the unit?
Should variable cost include non-fuel operating and maintenance
costs? What components should make up non-fuel operating costs? Can
these costs be estimated from Form 1 data with sufficient accuracy
for a model? If they can, what methods should be used for estimating
these costs from Form 1 data? If they cannot be estimated from Form
1 costs, what sources of information should be used in their
estimation?
Should NERC PTDFs and flowgate limits (if available) be used?
What are the strengths and weaknesses of using the NERC PTDFs and
flowgate limits? If flowgate limits associated with NERC-calculated
PTDFs are available, can they be used in the way they are
represented in the sample model discussed in this attachment? If
they should be incorporated in a model using an approach that is
different from the one described in this attachment, what should
that approach be?
If NERC flowgate limits are unavailable, is the approach of
using PTDFs and flowgate limits to represent the physical network
still practical? If the PTDF approach is practical in the absence of
flowgate limits provided from NERC, how should other sources of
transmission limit information (such as OASIS TTC or ATC data or
system reliability studies) be used to estimate flowgate limits? If
the PTDF approach is not practical, how should actual power flows
and transmission limits be modeled?
Environmental factors can influence the variable cost of
operating plants. For example, the variable cost of operating coal
plants is affected by the cost of SO2 allowances, and
environmental programs in California and the Northeast could have a
significant impact on costs. Are these costs adequately captured by
publicly available sources, such as the reported costs on Form 1, or
do they require separate cost estimation?
Application to Merger Analysis
Can the model be straightforwardly applied to simulate the
supplier identification step of a delivered price test that is
consistent with a delivered price test performed without a model?
First, consider the delivered price test as it is described and
applied currently, without adjustments to supplier capacity. Then
consider how a model might be used to adjust supplier capacity for
the presence of loads at other destination markets, and how such
adjustment could be made in a manner consistent with the purposes of
the delivered price test.
In addition to using a model in a delivered price analysis, what
are the other areas of market definition or of the analysis of the
competitive effects of mergers where a computer model could be used?
Comments may address the general use of computer models in antitrust
analysis, such as their use in a hypothetical monopolist test or
their use in simulating dominant firm behavior. However, comments
should address how these applications might function as a screening
tool and in the Policy Statement. In your comments, specify what
these areas of application are and what benefits are provided by
using the model, how the model would be used in the analysis (in as
much detail as possible), and how use of the model can be made
consistent with the practical constraints of time and resources
available in the screening context.
Process of Model Development and Maintenance
The staff believes that a computer model can be a feasible part
of a horizontal screen, and will aid the analysis. The model may
also have the potential to expedite the analysis by providing
agreed-upon standard methods that can be applied in merger analysis.
Are these beliefs sound, or are there limitations in principle or
practice that make the use of models infeasible as part of a
horizontal merger screen?
What should the Commission require with respect to computer
modeling in merger analysis? Should it endorse a specific computer
model, a particular modeling approach (such as an economic dispatch
model), or only a general framework? Or should it only seek to
provide guidance on how a model should be used if applicants choose
to include one in their application?
Are there existing models that meet the requirements for use in
a horizontal screen? Explain how any candidate model could be used
by staff, applicants and/or intervenors in the context of a merger
application? Address issues of technical adequacy, practical issues
such as complexity and ease of use, and procedural issues such as
the proprietary nature of third-party commercial software products.
If there are other existing models, should the Commission staff
acquire a existing model, or should Commission staff develop a model
for its own use and the use of applicants and intervenors?
If the Commission staff were to develop a model rather than
acquire an already existing model, what development approach should
be taken? Should the model be developed by Commission staff based on
technical discussion and input from industry, by industry groups
with Commission oversight, or some other way? If the Commission
adopted the approach of issuing guidelines only, but not developing
a single model for general use by staff and applicants, would
independent development of models by others provide models of
sufficient quality and standardization for merger analysis purposes?
How should a model be tested prior to use in specific merger
cases? If a model has been used in other contexts, under what
conditions should that use be regarded as sufficient to validate its
use as part of a horizontal screen analysis? If the Commission staff
were to develop or adopt a
[[Page 20404]]
new model for use in merger analysis, how should it be tested to
ensure that the design criteria have been met?
How should a model and associated databases be maintained and
updated? What process should be followed to identify needed
modifications to the model and create new versions of the computer
code? Should a fixed set of data inputs be identified, in order to
avoid this potential difficulty and provide consistent a starting
point for analysis (assuming applicants can file additional data for
further analyses if they choose)? As an alternative, should
applicants be permitted to substitute the most recent data from the
same sources even if these data have not previously tested in the
model? Or should a standard set of model inputs be maintained and
updated as a group? If a standard set of inputs is maintained,
should Commission staff be directly responsible for the maintenance
of these data or can this responsibility be carried out by third
parties?
[FR Doc. 98-10687 Filed 4-23-98; 8:45 am]
BILLING CODE 6717-01-P
