COST-EFFECTIVENESS FRAMEWORK FOR MINIMIZING
TOTAL COST OF EPA REGULATIONS
PHASE I REPORT
Prepared fori
U.S. Environmental Protection Agency
Office of Planning and Evaluation
EPA Contract No.
68-02-2672 (Bottelle)
January, 1979
Development Planning and Research Associates. Inc.
ICF Incorporated
J. Watson Noah Associates. Inc.
Battelle Memorial Institute
-------
COST-EFFECTIVENESS FRAMEWORK FOR MINIMIZING
TOTAL COST OF EPA REGULATIONS
PHASE I REPORT
Prepared for:
U.S. Environmental Protection Agency
Office of Planning and Evaluation
EPA Contract No.
68-02-2672 (Battelle)
January, 1979
Development Planning and Research Associates, Inc.
ICF Incorporated
J Watson Noah Associates. Inc.
Dacrelie Memorial Institute
-------
TABLE OF CONTENTS
Chapter page
1. Executive Sunmary and Introduction 1
Introduction 1
Summary of Results 4
Final Comment 12
2. Theory of Marginal Cost Effectiveness 13
Background 13
Conceptual Development 14
3. Mi! n i rvi 1 Cost Effectiveness Methodology 24
Step 1 Population Analysis 24
Step Entity Analysis 25
Step J identify Elements of the Data Basn 26
Step 4 Create Treatment System Cost and
Abatement Table 28
Step 5 Compute ICE Ratios 37
Step S Analyze BCE Ratios 41
Step 7 Aggregating Entities 4?
4. Textile Industry: A Case Study 51
r>• New Coal Fired Power Plants: A Case Study 65
6. Implementation Concerns 79
Complexity Issues 79
Increased Data and Analysis Requirement;; 31
-------
TABLE OF EXHIBITS
Chapter 2 I. Public Policy Economic Concepts, Total and Matgjial
Social Benefit and Social Cost Function
2, Basic fCE Aggregation Concepts
Chapter i 3. Elements of Data Base
4. Treatment Systems Cost and Abatement Table
5. Typical Total Cost Data
6. Typical MCE Curve
7. MCE for Two Industries
Q. Total Costs of. Control Options Over Time
9. Total Cost for Meeting Interim Standards
10. MCE of Delaying Implementation
Chapter 4 11. Definition of Model Plants for Direct Discharqet s,
Textile Industry
12. Alternative End-of-Pipe Treatment Techno log i<=n ,
Existing Sources
13. Textile Cost/Abatement Table
14. Computation of Marcinal Cost-fif f eci i vnr- Rat i v- Pm
Conventions Is tinder Two Cost Allocation Procedures
15. A Summary of ICE Ratios For Conventionale
16. Marginal Cost-Effectiveness Rat ios For Conventional.';
{as a Target) in Medium Woven Fabric Finishing Complex
Plus Resizing
Illustration of Aggregation of" Marginal Cont-
Effectiveness of Conventionals for Woven Fabric
Finishing
Chapter 5 18. Unit Process/Treatment Chain Table
19. Total Cost Estimates of Pollution Control Unit
Processes
20. Treatment Systems Cost and Abatement Table
21. Total Co?its of the Relevant Treatment Synterns
22. Average Cost-Effectiveness for Six Alternatives
23. Total Cost/Total Abatement Graph
24. TSP Data From Row;; 1 tn 4
25. S©2 Data From Rows 1, 9, and 17
26. MCE of TSP
27. MCK of .SO2
-------
1, EXECUTIVE SUMMARY AND INTRODUCTION
The Environmental Protection Agency has been at the forefront of analyzing
the economic consequences of regulations on those who must. comply with them. As
part of its continuing policy of evaluating both the public and private economic
effects of its regulations, EPA has recognized the need for a more detailed ana-
lysis of the marginal cost-effectiveness (MCE) of alternative pollution control
methods and levels of abatement. Much of the previous work evaluating the im-
pact of a proposed regulation has compared the costs of alternative technologies
or levels of abatement. However, much of this work has focused on a single
technology chosen to represent the "best available" or "economically achieve-
able" control technology without consideration of its cost-effectiveness com-
pared to other methods of control or levels of abatement (effectiveness).
Recent declarations by EPA that it wojld analyze the MCE of pollution con-
trols to improve the evaluation of alternative control strategies in order to
try to obtain the least costly mix of pollution controls Led to the formulation
of this stud". In this context, the study was designed in two phases:
® Phase ' - the development of a methodology and its pilot test in
two i n'iustr ies, and
® Phase II - the application of the methodology to selected indus-
tries to derive marginal cost-effectiveness relationships for
use in policy decisionmaking.
Th u"> report presents the results of the wor k on Phase I, which took place from
September 19 '8 to January 1979, Additional modification and clarification of
the methodology described herein will undoubtedly emerge during Phase II.
Thus, the methodology should be considered as preliminary at this time. The
report is organized as follows;
1. Executive Summary and Introduction
2. Theory of Marginal Cost-Effectiveness
3- Methodology for Marginal Cost-Effectiveness
4. Textile Industry: A Case Stacy
5. Coal-Fired Power Plants: A Case Study
6. Implementat. i on Concerns
THE POLICY CONTEXT
EPA's primary objective i.n sponsoring the development of an MCE methodolo-
gy wis to augment its capabilities to analyze the impacts of its regulations.
-------
- 2 -
The Agency recogn izes that it is no longer satisfactory to measure? only thu
economic burden that meeting a particular standard for a particular pot uu-iit-
places on a specific industry; nor does evaluating the cost-effectiveness
moving to a more stringent control level for a single pollutant provide ade-
quate assurance that EPA is acting in a manner consistent with Its goal of
provid mg the highest Level of environmental protection for thf loast cost..
The nopi for more advanced analytic tools is particularly evident, because
the Agency is increasingly being subjected to legal challenge on economic-re-
lated issues. In adopting more effective regulations, EPA confronts increas-
ingly conf)lex policy decisions. The primary policy issues include determining
control strategies;
• for a single pollutant or group of pollutants,
e for an industry ot industry segment, and
• to determine emphasis among pollutants.
For each of these primary issues, there are a number of aspects which EPA
must determine including which specific industries and pollutants to reguiat,.
^nd the nuier in which they should be considered, tv:>w the regulations might
affect particular regions, the date at which the regulations should be imple-
mented, and whether ot not interim standards should be used.
In developing the marginal cost-effectiveness analysis methodology, it was
a pp a rent that a single analytical scheire could not: be applied to respond to
all the varied policy questions for which such a tool would be useful. We
found that different, issues arise (e.g., weighting, timing, aggregation) de-
pending on the specific policy questions being addressed. Consequently, we
have developed a basic methodology containing a number of steps from which
only those necessary to respond to the particular policy questions at issue
c o u 1 d b >• r, o 3 e<: ted.
THE ANAI.YT fC CONTEXT
In theory, environmental regulations are designed to maximize the differ-
ence between the social benefits derived from the regulation and the social
costs fboth public and private) of compliance with the regulation. Benefit-
cost analysis (BC) is a we 11-developed theory of analysis designed to aid in
identifying socially optimal policies or regulations. Its utility is severely
limited, however, by the requ i renient to quantify both tangible and intangible
social costs and benefits. Cost-effectiveness analysis (CE) was developed to
analyze problems where benefits could not be quantified in a manner commensu-
rate with the quantification of coats. CE is designed to identify efficient
but not necessarily optimal solutions by answering one of the following to
quest ions:
1. What is the least-cost way to achieve a given level of '-Ft'ec-
tiveness?, or
2. tlow can the greatest effectiveness be achieved for a given
level of expenditure?
-------
- 3 -
CE'results have a good deal r,f utility in trie decisionmaking process r marginal change in
cost with effectiveness at a given level of effectiveness.
Marginal cost-effectiveness (MCE) tends to overcome the problems associated
with conventional CE solutions. MCE results can be used for compar isons when
neither cost nor effectiveness is held constant and thus is a suitable basis for
identifying efficient solutions to a broad array of policy questions regarding
pollution control alternatives.
MCE analysis as applied to pollution control issue:; seek.;; to establish the
relation between successively more stringent degrees of abatement (effective-
ness) and the corresponding change in cost of compliance. Theoretically, a con-
tinuous functional relationship between cost and effectiveness can be developed
at both the macro- and microecoooraic levels. In practice, these relationships
can rarely be derived analytically, because the abatement technologies generally
yield a discrete interval of effectiveness at a discrete increase in cost.
Functional relationships between cos!: and effectiveness ran be appro* ima t«>d
statistically through successive incremental analyses. Thus, where a continuous,
cost-effectiveness function does not exist or cannot be easily derived, the MCE
analysis is akin to repeated applications of CE analysis under nearly equal
e ft >x: t lveness conditions, i.e., repeated CE analysis for several closely related
effectiveness levels so that the "marginal™ costs of the "marginal" removals can
b# ascertained.
Consequently, MCE analysis as applied in this methodology is the blending of
successive incremental changes in cost and effectiveness and technically should
be ca 1 I -?d incremental cost-effectiveness analysis. How?v
-------
- 4 -
3ecause marginal cost-effectiveness analysis is not now an integra L part of
the standards-sett i ng process, extensive data requirements must be fuif • 1 ,:>d in
order to under take this type of analysis. For EPA to analyze the marg< -a- cost-
effectiveness of its actions in a compiehensive fashion, it would have to allo-
cate increased resources to broad exam nation of the effects of propose?. roli-
cies. At some future time, EPA must decide whether the costs necessary o col-
lect the additional required data are justified by improvements in the cost-
effectiveness of its. regulations. Because data requirements are an integral
part of performing an MCE analysis, we have incorporated into the early :teps of
the methodology a systematic approach to defining the pollutants and treatment
systems for which costs and abatement levels will be needed. In add it io<-
where data limitations exist, we recommend simpli fications and data ma;,ipu !-
tions which will facilitate MCE analyses in these instances.
We must also enphasize that MCE is only one analytical too 1 among nv-m'
{e.g., average cost—e ff ect iveness analysis, least-cost: solutions, and t.ota.i cost
analysis), and for some specific policy questions, it may be inappropriate
Recognizing the possible limitations of performing MCE analysis, the me"h ,v i ogy
has been designed so that it can also be applied to other relevant econoiii ¦ na-
lysis. In effect, we have expanded the scope of the original workplan an? jvr
attempted to provide EPA with a systematic approach to defining the basi. puts
(cost and effectiveness data for alternative pollution control strategies'
necessary to enhancing its ability to do economic analysis leading to more • f f 1 -
c i ent reg u la tor y s t ra teg i e s -
SUMMARY OF RESULTS
This section presents a summary description of the methodology that wa.;
developed during Phase I and briefly reviews the results of the two pilot tests.
Summary Description of the Methodology
The methodology devised for computing marginal cost-effectiveness ratios is
comprised of seven steps:
1. Per -orm Population Analysis
2. Perform Entity Analysis
Identify Elements of Data Base
4. Compute Treatment System Cost and Abatement Table
'j . Compu te MCE Rat ios
6. Analyze MCE Ratios
7. Aggregate Entities
Each of. this steps is discussed below.
-------
Population Analysis. Population analysts identifies the universe of 3 nduu-
tr ies or processes relevant to the policy issues being examined. For example,
tf the policy question were: "What is the MCE of control J inq pot Lur.ion from tn»--
steel industry?" then the relevant population would he all the processes in th<>
steel i ndu str y. If the policy issue were "Compare the MCK of controlling BOD
in two different indusfc ies," the population wou Id be those processes in each
industry involved in BOD produc11 on and, control. Similarly, if the policy issue
were "Compare the MCE of controlling ait pollution from boilers," the relevant
processes using boilers must be culled from all industries.
In short, population analysis involves two substeps:
1. Define policy issues study is to address in broadest possible
terms, and
Identify industries a nd processes io levant to policy issues.
Entity Analysis. For most policy decisions, the ultimate objective is to
analyze MCE ratios for an entire population {e.g., the steel industry, all boi-
lers). But populations frequently will consist of diverse components. For
exanple, the steel industry includes three different types of furnaces are! a
total of twenty-nine processes. Foe the purposes of performing MCE analysis, we
propose that populations be divided into components with similar characteris-
tics. We refer to these subsets of a population in this analysis as "entities."
An entity may be a mole i. plant, a particular engineering prficesa or a mobile
.-source. In de lining entities, several characteristics should hp considered in-
c Iud ino age, siz^, and engineer lng procer,s.
Identify Elements of Data Base. This step of the methodology specifies the
required data. In this step, we identify; the pollutants to he included in the
analysis (Step 3.1} and the unit processes to control each of these pollutants
{S te p .1.2).
Ideally, all pollutants affecting a given entity would be included, but this
may make the analyst.*! unwieldy. No set rules can be ar 11 en la ted for deciding
which pollutants are to be included in the data base. In part, the specific
pollutants included will be determined by the entity being analyzed and the
policy issues being addressed.
A unit process is a piece of equipment, an engineering process, or a raw
material which achieves a level of abatement of a pollutant. It is necessary to
identify each plausible unit process capable of abating a particular pollutant.
Where a unit process affects only one pollutant, this task will be straight-
forward. Where a unit process affects two or more pollutants, a method must be
employed to apportion the costs of operating the process among the affected pol-
lutants.
Create Treatment Systems Cost and Abatement Table. Th<> Treatment Systems
Cost a nd Abatement Table brings together the entire set of key pollutants, the
applicable (unit processes! and the abatement levels they achieve, and their
costs of pollution control. The design of the Table (4.1?, how to determine the
possible number of rout mat ions (4.2), the relevant cost factors (4.3) ar*! the
etfectiveness of the combinations (4.4) ate the substeps involved.
-------
The Treatment Systems Cost and Aba tern e nr. Table is designed to serve as the
fundamental data base for evaluating any MCE-re lated policy questions and must
be as coiiprehensive as possible. It must present the costs of control, broken
Mown into detailed estimates assignable to specific poll itants. It must present
ill the leve.'s ot effectiveness attai nuble by each unit prowss. Most impor-
fantiy, it must examine the costs and effectiveness of" each plausible combina-
tion of pollutants and controls. For each pollutant, the required data include:
(1) techr oI<>jy or unit process employed;
!2) the* cost of that process;
(3) the amount of pollutant removed by the process; and
(4 1 the amount, of pollutant Still emitted.
Calculating control costs for each pollutant is one of the more complex
tasks in the MCE methodology particularly for those pollutants and unit pro-
cesses which affect more than one pollutant. Where possible, based on engineer-
ing judgment, these joint costs should oe allocated among the affected pollu-
tants on the basis of their relative contribution to the total costs of control.
In most cases where joint costs occur, this will be impossible, and a method for
allocating costs among affected pollutants is required (see Step 5).
Quantifying effectiveness results in measures of the level, of abatement and
of the level of emissions from the entity when the unit process combination, is
applied. For each key pollutant, an aopr opr i a te measure should be devised. We
recommend kiloqtarns removed and kilograms emitted per year whenever: appropr iat»-.
For certain pollutants, other standardized neasures will be necessary (heat,
degrees; pH, pH level).
Compute MCE Ratios, The exact data drawn from the data base will depend on
the particular poLicy question being addressed. The policy issue is then used
to identify tiv1 te levant data which directly affect it. Kor example, if the
policy question involved the MCE of changing the standard for a particular pol-
lutant, then only those data which accomplish this while holding all other pol-
lutants constant (or by allocating costs to those affected pollutants where
costs are nonseparable! must be analyzed.
Having identified the data relevant to the specified policy issue, we now
have a measure of costs and abatement Cor each pollutant. Before using this
data to compute the desired MCE ratios, a final manipulation may be required.
To determine marginal costs where inseparable costs exists, some allocation
scheme must be employed. Several ways to assign nonsepar able costs exist and
selecting the one to emp l.oy will depend in part on the availability of data and
the particular po t icy question being addressed. These include:
• target pollutant - all costs ate assigned to one pollutant,
• ratio of separable costs - costs are assigned in proportion to
the ratio of separable costs.
• allocation on effectiveness we lght ing - costs are assigned in
proportion to the ratio of damage averted by the unit process,
-------
n
• ratio of costs of' separate r ici li zied - ecscs ar-? .assigners in
proportion to the ratio of tne costs ,>t buiLdmq separate treat-
ment chains foe each pollutant, and
m equa I ot costs - costs are a.sr. igned in fc adopt a less sr.- nwent interim level ot
required abatement in an effort to lessen the burden its regulations impose on
industry, followed at some later date by a more stringent target standard.
The use of interim standards affects two aspects of the MCE methodoLogy. We
can reasonably assume a firm will select the least-cost treatment chain to
meet each of the proposed standards. Where i he use of interim standards i -i
being considered, examining only the least-cost treatment chain to comply with
the alternative standards imposed at different points in time could be mis-
leading. With the likelihood of shifting over time from a given standard to a
more ;;t l ingent one, it may be less cost ly tnr a firm to select a unit process
which actually costs more than another at the initial period. Thus, from a
marginal cost standpoint, in situations where two ot more time periods are
being c xisidered, it is essential to examine the full range of possible unit
processes to comply with each standard.
Although timing is sua ightfoi ward as it relates tu costs, it severely
compl ieates attests to measure effectiveness. This problem arises when the
MCE methodology is used to compare a proposed standard which is to take effect
immediately to one to take effect at some future date. Comparing costs in
this situation is relatively easy. Although there may he some debate about
the appropriate discount rate when applied to the estimated future costs,
these costs can be directly compared to present investments. But no such
clear-cut manipulation exists for cougaring the effectiveness of the same
standards imposed at different times. If we were to ignore this problem, the
MCE of i standard imposed today wou 1c be exactly the same as that of the same
standard if its implementation were delayed fot a period of time. The problem
is handled by allocating the amount of pollutant removed over a time horizon
appropriate to the policy question at issue.
Aggregate Entities. Marginal cost-effectiveness, although most applicable
at the entity level, can be conceptually extended to apply to industry seg-
ments or across industries, and to geographic regions or nationwide. For many
of the policy issues capable of being addressed by MCE analysis, some level of
aggregation will be necessary. Examples of these issues include the Following:
-------
8
• What ir the marginal cost-effeetiveness or cleaning up one pol-
lutant to the same degree in different, industries?
• What is the marginal cost-efCectiveness curve for cleaninq up
one pollatant in all industries?
• What is the marginal cost-effectiveness of cleaning up all pol-
lutants in on*v region of the country?
• What is the marginal cost-effectiveness curve for cleaning up
one pollutant across all industries in one region?
To respond to these policy questions requires that model plant data be ag-
gregated alonq any of three dimensions:
• pollutant
• industry
• geography
Aggregation is the equivalent to summing the applicable entity-level MCE
curves over a constant range of marginal costs. Before summing the curves, how-
ever, it xs necessary to weight each by the proportional number of equivalent
model entities in the relevant population. Secondly, in order to obtain a con -
r;tctnt range necessary Co sum the MCE curves, they may need to ho extrapolated t<»
>i common marginal cost range. Finally, if different pollutants are being aggre-
gated, they mui:t he weighted (or assumed to be equal).
Coal-Fired Power Plants: A Case Study
This case study shows how the methodology would evaluate alternative new
source perforitiance standards currently being proposed for coal-Ci red power
plants. The population must be defined to include those coal-fired power plants
likely to be built if specific environmental regulation;: ate adopted.
Defining the entity to be analyzed is relatively straightforward in this
case study. Because we are dealing with a new source performance standard,
neither age of the plant rot varying engineering processes are relevant consi-
derations. We will assume that all new facilities employ the same boiler pro-
cesses. For the purposes of this case study, we have defined the entity to be a
5OG-megawatt power plant.
The Development Documents for coal-fired electric utilities and the numerous
studies supporting the development documents have identified over 50 pollutants
which are emitted hy electric utilities. These fifty pollutants include conven
tional, non-conventional and priority water po1lutants plus criteria and hazar-
dous air pollutants. Sludge 13 also created in substantial quantities. Ot.
these fifty, the most serious are:
Air:
SO 2
Fiyash (TSF)
NOX
-------
W*ter;
:l'i -.jiftnifti and Dissolved rolids
Heat.
ph
Chlor *. ne
Oi Cj
Trace Metals
S ludge
Several unit processes are available tor controlling 502 emissions, two
unit processes control particulates, and one unit process is available for
M°x- Sludqt-r is oheriu cai ly treated and land--! 1 I i ed. the data were
insutti- cient, water pollution unit processes wore rn>t analyzed.
The most w»-! 1 -k nown unit process to control SO^ ,¦ ,, f |.ue-qas desulfur izer
fFGD) » \l:io known as a scLubber. Another unit prof-ess to control -SO2 erais-
sitxiia physical cmI cleaning {PCC) and the use of low-sulfur coal (LSC) . For
particulates, two unit processes have b»en identified. The most, common is an
electrostatic precipitator (ESP). The .second unit process is a fabric filter.
The only unit process available to control NGX is two-stage combustion.
Several unit processes exist to dispose of sludge, but their current costs
reduce the practical options to ponding ami chemical treatment and landfill.
Because data were totally inadequate, as compared to the partial inadequacy
of air and sludge data, we were unable to incorporate water unit processes into
this analysis.
Summary. After identifying the relationships among the unit processes and
completing the Treatment Systems Cost and Abatement Table, we were able to il-
lustrate the application of the methodology to specific policy questions. -It is
! nip or t ant to no to that: the limited and unreliable qu 111 ty of air and sludge cost
data and the lark of any water treatment data Sfveie >y compromise the ability to
use these trial 1 osults as anything more chan an example of an application of
the met hodo Uiqy . The policy questions addressed wen-:
• What is the MCE of alternative, more stringent sulfur dioxide
standards?
• What in the MCE of trad ing-off particulate control for sulfur
d t OX 1 dr« control?
In applying the methodology to these issues, it wa.; apparent that, when
used to address a specific policy question, some manipulation of the basic
methodo lojy is required. This is particularly true when only limited data is
available. At the same time, the methodology proved to be flexible enough to
provide the basic framework for analyzing a diversity of policy issues.
Textile Industry: A Case Study
The relevant population of textile mills was defined as existing direct
d tsoharqeri;. The population is composed af about 220 mills which are subject
-------
10 -
t«> best ava i J ab i* technology economically achievaoi*. This |>opulsti««- -an -»,¦*
represented by 26 entities composed of different processes and mill sizes.
The data base used was Erom recent Effluent Guidelines Document on the
U'xnle industry. Although a number of pollutants are present, seven key uo!
Infants and pollutant, parameters wore repor ted including:
Conventional BOD, COD, TSS, O & G
Honconventiona11 Phenols, Chromium, Sulfide
The Textile Industry case study demonstrated that marginal cost-effectivp-
ness can be developed but only after significant analytical effort not normal-
ly included in the industry engineering studies. Model plant information can
tie agg r eg a ted to industry or regional totals provided that t.ho study include:-,
geographic as well as size/type of plant parameters.
These positive results are offset by the following factors. The results
actually generated ate useful only as an example application of the methodolo-
gy and cannot tie used foe policy decisions. Moreover, the additional ana lys < -
required is substantial and requires a degree of sophistication not normally
needed in engineering studies ¦— the methodology relies on data that may be
available Airing the engineering analysis but not generally required to
achieve study objectives. Finally, MCE results of the type possible exclude
such salient consi rteiaM ons as economic impacts and are therefore only one
tool of several required for sound policy decisions.
Because of the data limitations the MCE analysis was confined to one of
the 26 segments; "confilex plus desiz ing" mills of the woven fabric finishing
segment- Also, the analysis was confined to one of conventional pollutants.
This stenmed from the lack of technical background information for detailed
cost assignment.
Within a re levant range of the avallabLe data, we found a marginal cost
curve lor convent i on,i I pollutants Cor wwen fabric finishing composed of 17
small. ; nd 10 large plants. This components of this curve are;
ABATEMENT
Ma r g i na 1
Cost
Amount
Percent
1 . 00
1.50
2.00
2, SO
1.00
3. 50
4.00
2415
3875
4 9 50
5310
6 4 50
6955
7125
25
40
52
6 1
67
7 1
76
'The inclusion of chromium was done as a matter of convention and ease
of presentation of the analysis.
-------
- li -
At a marginal cost of 51.50, the small mills could reach a 32 percent,
abatement Level and medium milts could remove-46 percent of the conventionale.
This illustrates the different amounts of abatement among a subsegtment for a
given marginal cost.
The overall finding of this analysis was that, even though the source
document used for data points is one of the better ones we have reviewed, a
large amount of-" additional background inf Dtmat ion i s required for carrying out
the doLa l 1 el ana ly s is as represented by M"E analysis.
Implementation Concerns
The development and testing of methods for measuring the marginal cost
effect ivory ss of SPA requlations was much more complex and difficult than ori-
ginally thought. A number of factors affect the complexity of the analyses,
and, in turn, raise theoretical issues which impact on the rigor o€ results
obtainable. The specific issues which introduce complexities into the imple-
mentation of the methodology include: the types of data required and theoreti-
cal complex it ies. The specific issues in each area are:
* Types of Data Issues
1. Number of Entities
2, Number of Unit Processes
1. Interdependency Among Pollutants
• MCF; Theoretical Issues
1. Well Ordered Treatment Chains - unit processes must be combined in
such a manner that total systems cost increases and that the
amount of each pollutant removed does not decrease.
2, Number of TCE (MCE) Data Points - to aggregate to industry,
region, or national totals requires a reasonable number of data
points covering a reasonable range of MCIi values. Each data point
requires a properly constructed treatment chain and data points
are required over an extensive effectiveness range.
J. ICE (MCE) Interval - the data available Cor the case studies was
not sufficient to develop a statistical approximation to the
underlying cost effectiveness function. Instead MCE results werr
approx ima t ed by ICS a na ; y s i s.
Thr above issues introduce complexities into implementation not so much
because they cannot he resolved but because of the scope and detail of the
tasks required. We believe that significant gaps exist in the data available
but that these gaps can be closed by more: complete and effective systems
analysis at the entity level. Moreover,, it seems probable that, the quality ot
data available in completed engineering studies can be improved by interact ion
with selected labs or contractors.
-------
Information gaps include:
I - Raw Influent, BPT Effluent and BAT Effluent Cnaracter istics -
No information on caw influent was available and only aggre-
gated characteristic data foe BPT and treatment chain effluent.
Knowledge of the effluent character istics for each unit process
in c.ich treatment chain is essen t. i a 1 £01 cost- assignment..
2. Treatment Chain Lo<] i c - The reasons for s ^ I >x: t: i ng par !. i n i1 a r
unit processes are! the sequence in which they are applied
should be fully explained.
3- Complete Cost Analysis - Both investment and operating cost
estimates should be prepared for each unit process in ea-~h
treatment chain tested. This i:; essential when cost issiqnnieni.
in r»iuired.
* 4¦ Cost Effectiveness Analysis - A cost effectiveness study, not a
mere reporting of cost and effoetiveness is required. Each
treatment chain presented should be the least cost method for
achieving the desired level of effectiveness.
5. C r os s -Med i a Con s i.de rati on - It is important to include impact.-:
on other media.
The above data gaps were generalized from the Textile Industry case study.
The Coal-Fired Power Plan Case Study revealed similar problems. More specifi-
cally, information on the interrelationships among poLlutants ami unit pro<> es
was sketchy, cost arid effectiveness points were limited even when a continuous
function was available and data was: lacking on at least sevora i feasible comb i
nations of unit processes.
The data and analysis problems cited above can be solved but at some cost in
both money and time. The problems suggest that our Phase II effort be concen-
trated on a tew Key industries so that truly useful results can be obtained. it
does not ajjpear that, a broad-brush study of a large number of industries bas.-d
on available data will produce results much more applicable than the two case
studies completed in this Phase.
The Phase II study will, of course, clarify the issues raised here. It
seems ike ly, however, that a comprehensive application of MCE methods will re-
quire no tea sed resources for more detailed aaa Lyses of abatement option. HPA
must tleietoro assess the cost e fT t iveness of MCF. at souse future time wh<->n
better estimates of resources are available.
FINAL COMMENT
Marginal cost-effectiveness analysis;, when properly performed," can be a
powet tul and useful tool for policy analysis. However it. should be recognized
that it requires extensive data. Al:>o it. should be cnnsidei ed as on 1 y on'1 tool
to be used in policy analysis. It does not replace conventional economic impact
analysis aimed at plant closure, production and employment impacts. Both types
of analyses should be performed in setting pollution control standards.
-------
- 13 -
2. THEORY OF MARGINAL COST-EFFECTIVENESS
Marginal, cost -f=> f 1 ec !. iveness { MCE) ana 'vsis of pot 'ut ion abatement seeks to
egtabish the relationship between successive increments of pollution abatement
{effectiveness) and cor responding incremental costs of abatement. Theoretical-
ly, continuous functional relationships exist between cost and effeetiveness at
both microeconomic and macroeconomic levels, but in practice such theoretical
MCE functions can only be statistically estimated through successive incremental
analyses. This arises because most abatement technologies require discrete com-
ponents having discrete costs for a given level of pollutant somovaU Thus, th<-
marginal costs of pollution abatement, and the marginal effectiveness are not.
norm.i 1ly precisely related, although they may be concurrentiy analyzed and an
MCE relationship may b<.- approximated.
BACKGROUND
As a point of reference, marginal cost-efEectiveness analysis is contrasted
with that of cost-ef feetiveness analysis which involves one of two criteria:
• holding effectiveness constant and determining
the least cost alternative for nulling the spe-
cified cfhvtivencss, or
* holding total cost constant and determining the
alternative which maximizes effectiveness for
the specified cost.
MCE analyis is most like performing repeated applications of the first cri-
terion, i.e., repeat the cost-effectiveness analysis for several closely-related
effectiveness levels so that the "marginal* cost {also least cost for each ef-
fectiveness lev--1.) <>f the "marginal" removals can he ascertained. Next, MCE
analysis may requii o that an MCE curve be fitted to successive "marginal" va Lues.
Such an approach to MCE poses analytical subtleties and may present burden-
some data requirements even for a single pollutant, because the methods of
achieving improved pollutant removal usually involve new or modified treatment
¦//stems {rather than changes in operating procedures of a given treatment sys-
tem) . The necessary data may not be readily accessible.
Most pollutant, abatement issues deal with multiple pollutants that are in-
tetrelated within treatment systems. In this situation, two fundamental ques-
tions arise - both of which may require solutions for many policy problems to
be answered:
a) What is the cost of each pollutant's incremental removal, i.e., how
are the treatment system's joint costs to be allocated (assigned) to
individual pollutants?, or.
-------
b) What is a common measure of of fee t lveness, i. , how ire the mui; >pi •
pollutants to be separately weighted to produce an overall index ot
abatement?
Th 13 mu Lt1 poll utant case and the quest Lons tt raises are complex and hav e
been the core of this research effort. Although there appears to be no single,
best answer to either of the questions posed, the research bar, developed some
,-il terna! i v<- approaches to each which .ire presented below.
In addition to the multipollutant issue, other di fflcult ies were encountered
that s ign 1 £icantly affect the applicability and value of MCE analysis. Two of
these issues are of major consequence — aggregation and time phasing, Their
theoretical and methodological implications are discussed separately below, as
are t hos. > lot other remain, i rig concern:;
CQNCEPTtJAI, j jEVIr;LO_PM.HHT
Marginal cost-effeetiveness analysis is Coremo.it a micoieoonoiiii e procedure,
although with proper aggregation it provides macroeconomic results. The poten-
tial usefulness of MCE analysis in making environmental management decisions ,:an
best be described by considering a simplified mactoeconomic case in which the
economic concepts of public policy arc embeddec.
Economise Con c_ep ts _of Public Policy
The generally accepted criterion for selecting among pub LipoI icy alterna-
tives {such as those regarding pollution abatement) is to choose those which
maxirar/.i- social welfare, the sum of both public and private net benefits. Such
a criterion theoretically requires per fee t knowledge of all components of the
social costs and social benefits of all feasible alternatives.^ Thus in these
terms the question "What degree of pollution abatement should be achieved to
maximize sociaL welfare?" can be posed. The question is extremely conplex, but
the conceptual basis for answering this question can be illustrated in a
.implied . vu-»o. For exaii^le, assume Shah only cne pollutant ex jst r, in a it."it ic
environment .i L settinq and that we know both the total costs and the total, bene-
fits associated with all levels of abatement, i.e., costs and benefit functions
from 0 to 100 percent abatement. Based on these functions, the associated mar-
ginal cost and marginal benefit functions can be derived. These relationships
are illustrated in Exhibit I.
With these relationships, it can be shown that maximum social welfare will
be achieved at abatement level A where marginal social costs equal marginal so-
cial benefits. In relation to the total, functions in Exhibit I, this optimum
point (A) also corresponds to the maximum distance between the total benefit and
the total cost, functions.
l
This oh-jeetive involves both issues of efficiency and equity, a 1 though
these issues will not be pursued here.
-------
-.Milrpit i
PUB I, IC POL _ XJ3C )NOHi: ¦,'. NCZV'::, ^
TOTM. AND MARGINAL SOCIA , BENEFIT AND* SOCIAL >X'ST FUNCTIONS
Total !
Value
(S) !
X
ISC
Abatement
100"
Mar 11 rial
Val tie
(S)
MSB
\
MM
/
A
Abatement
lOO1:
-------
In the single pollutant case, MCE analysis is equivalent to e st ima c i ng t'ne
margins 1 social cost function of Exnibit 1. Est imar my the marginal, r-oci i L
benefits is extremely difficult and must be approached using a proxy bjot-d on
effectiveness. Thus, the optimum level of abatement, i.e. , {A) , wi 1.1 not be
known. Presumably, some independently derived criterion in terms of the cost
level associated with A may be specified to indicate the desired marginal cost
IpvI that will approximate the opt imun abatement level.
As suggested, to the extent that MCE analysis produces an accurate estimate
of marginal social costs of pollution abatement, it becomes an important compo-
nent in deciding whether proposed (or existing) pollution regulations may be
"reasonable" or perhaps too costly in relation to perceived criteria.
Analysis Techniques
Benefit-cost ana Lysis applied to alternative treatment, systems will identify
the treatment system that maximizes net social benefits, i.e., the [joint A on
Exhibit 1. This technique cannot be applied, however, unless the total benefit
function (TSB) is known. When the tote 1 benefit function is not precisely known
(and in general such is the case), some relaxation ot the rigor of public-poi:cy
economic concepts is required.
Cost-effectiveness analysis can be applied (usually at the micro level)
under these relaxed conditions. The available abatement technologies could be
examined to determine several levels of: (1) cost of c omp lianc e or (2) abate-
ment. A total cost of compliance (TC) curve similar to TSC in Exhibit. 1, can
thus be developed. Because social, costs2 are excluded, TC would be below the
TSC curve although it would have a similar shape. Each point on the TC curve
would be efficient, because the technology selected either would be the least
cost method of providing the desired level of abatement, or would yield the
greatest abatement for a given cost of compliance. Thus, TC can be viewed as
the envelope or boundary curve of efficient applications of all available
technologies. Note, however, that location of the optimal point A cannot be
ident i f i ed.
Most cost-effeetiveness analyses yield discrete points on the TC curve (un-
less cost of compliance is a continuous rather than discrete function of effec-
tiveness) ; thus, successive analyses are required. The marginal cost curve
(similar to but below MSC on Exhibit 1) can be derived mathematically or throuqh
successive analyses at very small intervals around selected levels of effective-
ne ss.
Cost-effectiveness results may be sufficient for some decisions such as
establishing a standard for a particular model plant or process. This presumes,
of course, that, the proposed standard has already been deemed socially efficient
and equitable. Superior solutions can be obtained for many policy considera-
tions by use of marginal cost-effectiveness analysis. It can be shown, for
example, that the least cost method of reducing pollution over an entire indus-
try occurs when MCE ratios are equal fcr all processes in the industry.
2 .
This need not be done in general but is the normal and ((k.ammended
p rocedure.
-------
This being so, determining the best estimator of the true marginal cost
function is desirable. The precision of tne techniques discussed above can h
illustrated in the following example. Assume that cost (C) is a non-d<5croas
ing function ot abatement or effectiveness (E) and further that the abatement
has 5 finite limit. One such function can be written:
r = aE
b-E
where a is a scaling constant and b is the limit of effectiveness.
The marginal cost-effectiveness (MCE) function can be derived mathematically
a nd is
McE - :,-a~2
(b-E)
and the MCE ratio at any point x i:
HCE =
x (b-E )
x
Now, as discussed above, the MCE ratio can be approximated by repeated analy-
ses of small intervals around an effectivness level of x. This procedure,
applied to an interval of x, that is moving from x - x to x yields an in-
cremental cost effectiveness (ICE) ratio of
ICE
ab
x (b-x) {b-x+ fix)
This shows that ICE is a good estimator of MCE when x is small, and further,
that the mathemat ic definition ot MCE
LTK AC
MCE = E o Al
holds. Note, however, that ICE is not a good estimator of MCE if x is large
and that, for a given x, the error of the estimate increases as the point x
moves farther along the effectiveness axis.
A cost effectiveness (CE) ratio is a poor estimator of MCE. The ratio is
obtained by dividing total cost at abatement point x by the abatement, i.e.,
x. Assuming the same functional relationship
„ ab
CE = ——
x b-E
a ni
so that for
(b-x) <1, CE MCE
(b-x) = 1, CE = MCE
CF
x - (b-E )
MCE X
x
(b-x) ~y 1, CE MCE
-------
18 -
The relationship between MCE, ICE and CE have implicat l ir.s for the applica-
tion of the theoretical approach to practical situations. Considerable care
will be required to develop reasonable imatot s of MCr: when the cost effec-
tiveness function cannot be derived.
B,1 s ic Agg i egation Concepts
The previous example as illustrated in Exhibit I could have represented
either a single polluter or many polluters each with its own marginal cost func-
tions. It' more than one polluter were involved, the "nor i/.onta I ->um of the mar-
ginal cost curves" of all sources is r-jquired. The abatement levels must he
specified in absolute terms in order to weight each source in the summation pro-
perly; percentages are inadequate.
This agg regat ion concept is iHust'al.<*d in ExhibLt 7. tor 3 hypot.het u-a L two-
source case in the same environmental control region. Only the marginal cost
functions are shown, thouqh implicitly the total abatement cost curves exist.
The example illustrates that X he amount of the pollutant abated by each
source is to be summed for each marginal cost level. ft new total abatement axis
must be specified to represent the aggregate pollution levels in the control
region, Important characteristics of the summation process are to use only the
r jsing marginal cost portion of the marginal cost functions and to be sure that
all marginal cost functions span the cost-range being aggregated.
By aggregating in this manner, the sum of the sources marginal cost func-
t. ions yields the same aggregate marginal social cost function as portrayed in
Exhibit S. Furthermore, it can be shewn that the roost pfficient pollution
abatement procedure would be to establish the marginal cost level associated
with point A (in the aggregate) and then requite that all sources (two in
Exhibit 2} "spend" that marginal cost level, i.e., MCft = MC-j = MC-2 for
sources I and 2. This is the least-cost (and most efficient) manner to achieve
the abatement level A in this hypothetical case.
Other aggregation procedures, thotgh perhaps useful, will rot posses.; the
properties described here. The main concern is to maintain comp»rahiIity across
sources in the aggregate.
Marg ina L Cost-Ef t «*ct iveness Concepts
The preceding discuss ion was bases upon the stngle-pollutant case for which
theoretical constructs are readily determined. However, in fact, multiple pol-
lutants are often involved in treatment systems, and this comp i i cat.es the con-
ceptual basis (and the analytical requirements) of MCE analysis. Three general
cases exist for assessing the nwitipoliutant problem. First, one might allocate
all treatment costs among the impacted pollutants and consequent ly complete se-
parate MCE analyse;; for each pollutant, i.e., cost allorat ion approach. Second-
ly, one might develop effectiveness weights for each pollutant in order to de-
fine a single-valued abatement index, i.e., effectiveness weighting approach.
Thirdly, one might combine the cost allocation and the effectiveness weighting
approaches for specified classes of pollutants with consequent separate MCE ana-
lyses for each class of pollutants. Each of these general cases has underlying
-------
MC
5
3
2
1
Source A
—r - j
4 6
Abatement
10
(100?O
EXHIBIT 2
BASIC MCE AGGREGATION CONCEPTS
MC
2 1
1 J
Source B
MC
Abatement
(100?
Of \
ll
Aggregate
6 8
Abatement
~~!
10
i
-i~ i
14 I'",
; ion
i ¦
Source A
Source B
MC
n
JL
2
3
4
A
2.0
4.5
6.0
7.0
7.5
MC
1
2
3
4
5
1.0
2.5
3.5
4.0
4.2
Aggregate
MC A
1
2
3
4
3.0
7.0
9.5
11.0
11.7
-------
20 -
theoretical complications as a is-JUiised below. Not«? that the selection 01 ono
(or more) of the general approaches for a particular policy application is Im-
pendent primarily ipon the po Licy-re 1 a ted questions to be answered rather than
on the inherent nature of the approach itself.
Cost-A 1 location Case. In a multipolltant treatment environment, separate
MCE analyses may be made for each of the impacted pollutants if the treatment
system's component costs (e.g., by unit process) can be uniquely assigned to
individual pollutants. Often, however, costs cannot be uniquely assigned, be-
cause some unit processes abate more than one pollutant and such costs are non-
separable, i.e., joint. Theoretically there is no correct way to allocate -joint
costs; they must be arbitrarily allocated among the impacted pollutants.
Conceptually, when joint costs have been allocated, any ¦ ;ubsequent analysis
of a sinqle pollutant will potentially lead to decisions that. in^d not maximize
public arid private welfare. A partial cost allocation may understate the "true*
cost of abating a given pollutant. In contrast, allocating all costs of a unit
process to a single "target pollutant.™ —¦ so as to reflect the "true" cost of
abatement -- will necessarily understate the cost of treating other germane pol-
lutants, Also, the potential allocation of all costs to more than one pollutant
may lead to erroneous abrogation r^sizlts.
Despite the conceptual problems associated with allocating joint (nonsepar-
able) costs, there are many situations where decision rules may he applied to
allocate costs "reasonably**- Several alternative decision rules that may often
be applicable are described in Chapter 3 on methodology. The reasonableness of
each is principally dependent on how the marginal cost effectiveness results are
to be used, i.e., the policy issue being assessed including aggregation require-
ments.
Effectiveness-Weighting Case, This second general case requires the estab-
lishment of e ff ect iveness weights (e.g., environmental damage function values)
that are assigned to each pollutant to provide a single-valued abatement index.
In this case, the marginal (incremental) costs of treatment chains do not have
to be allocated among pollutants; hence, joint cost issues are not encountf1 .
On the oi hor hand, suitable e f f ec 1.1 v pne ss weight:; .n e generally unknown.
Conceptually, the most appealing type of weighting is one that rates pollu-
tants accordir>g to their relative damage in the environment per standard units
of pollutants. Such environmental damages will vary among control regions and
sub regions within such regions. Fut. t.her, an environmental damage function
weighting (versus a single-valued weight) is implied m relation to the aggre-
gate 1eve 1 s o f po1lu tant s emit ted in to the c on t ro1 reg ion.
Simple relative weights among pollutants (versus damage estimates) are ade-
quate for aggregating in this marginal cost-effectiveness general case, though
they too must be based on an underlying value system that should be known. The
weights used will, i n si r umenta 1 ly aff>ot. the final results oht.nn«v) and, there-
fore, the conclusions that can be drawi regarding she marginal cost effective
ness of the pollutants abated.
-------
Cost-Allocation and Effectiveness-Weighting Cas-e. A eomoi nation ol the
first, two cases can be utilized, perhaps most ' suitahi/, for most policy ques-
tions. This general case involves both partial cost allocation among classes or
pollutants and partial effectiveness weighting. Foe instance, a common distinc-
tion is made among conventional, nonconventional and priority pollutants, and
this may be an acceptable distinction for certain analyses. In this situation,
costs could be allocated among the three classes (thereby reducing joint cost
allocation problems) and effectiveness weights would only be required within
classes (perhaps equal) for purposes of the marginal cost-eCfectiveness computa-
tions. This procedure would help reduce the scope of the problem yet allow a
needed separation of results for each class of pollutant analyzed including the
ox post assignment of effectiveness weights as deemed appropriate for each case
studi ed.
The confo i na t i on case has advantage of overcoming some of" the joint cost al-
location issues, but, also, the disadvantage of introducing subjective (of ten
unknown) effectiveness weights within each class of pollutants. Information is
lost concerning the marginal cost effectiveness of individual pollutants, though
certain of the individual-pollutant results would be theoretically suspect be-
cause of joint cost allocation problems. At best, one should carefully consider
the issue to be analyzed, the availability and quality of data input, and the
degree of confidence in using needed effectiveness weights in deciding to use a
confo i nation cost allocation and effectiveness weighting approach.
Types of Data Required3
Basically, only four types of data are requited to conduct MCE analysis at
the entity, e.g., model plant, level. (Aggregation requires further data regar-
ding the "population" being analyzed as described separately below.)
These four types of data are:
a) Treatment Options
b) Treatment Chain Costs
c) Treatment Chain Abatements
d) Effectiveness Weights
The latter effectiveness weights are not needed for the cost allocation ap-
proach, whereas much of the detailed treatment chain costs (bv pollutant) are
not needed where the effectiveness weighting is applicable.
Each of these types of data are briefly described here, to emphasize the
data that ace conceptually required tor MCE analysis. Oftentimes available data
dees not approach the "ideal", though the data to be used can be accepted as
representing such an ideal. Further descriptions of these types of data, inclu-
ding common limitations, are presented in Chapter 3.
¦^In addition to the data needs specified here, MCE analysis ultimately
requires, also, a criteria or a ™ th res hole!" MC Sevel that approximates the
optimum level if it is to be enployed by all sources. Such a level corres-
ponds conceptually to the MSG level associated with abatement level A in
I'xh ib it 1, above, where MSC = MSB.
-------
- 22 -
Treatment Options. An ordered set of treatment options — theoret..:,/ .7
the leant -co:;t- set4 -- ace required/ each of which :,..-:cesj i ve 17 iTniev^
increasing levels of abatement for one or more of the spec 1. f iced pollutant,,
for which the treatment system is designed- The abatement level of no pollu-
tant should decline throughout the specified treatment sequence.
Unless a oost-effectiveness analysis is performed to assure for each f-
fectivenesa level the chosen treatment option is least cost, then the MCE ana-
lysis per termed may not be optimal. Hence, decisions oasod on ihr- anaLysi-*
may not max imize achievement of pubLic-poLicv oh]^:tr;°3.
Discrete, increased levels of abatement, Cor one or more pollutants, will
occur as a result of the ordered, sequencing process. Still, though, each
pollutant may have varied increments or abatement, and "ill po! i.utdats may not
have the same number ot applicable 1 nc remental aba foments, i.e., no change in
abatement is valid for some pollutants (but not all) affected by the designa-
ted treatment system.
Ideally, the abatement (effectiveness) increments for each pollutant will
be small throughout the sequence of treatment options. The smaller the incre-
ments, the more nearly the assessment vi LI represent a "marg ina I" -;na lys i.\,
and, hence, a true MCE analysis. However, the discrete nature ot known tech-
nological pollution removal alternatives may limit the aval lable options.
Treatment Chain Costs. The cost allocation approach is highly demanding
of treatment-system engineering are! economic data to perform the requisite
cost allocations among multiple pollutants. Generally, treatment systems are
comprised of unit processes that are designed to attack spec 1 11 e pollutant:;,
Therefore, data pertaining to these buil.ding blocks are essential, e.g., pol-
lutant influent and eftiuent levels, investment and operatinq costs, and engi-
neering design parameters. In contrast, if on Ly the effectiveness weighting
approach is to be applied, then only total treatment costs for each option are
required. These data requirements ate fully described below in Chapter 3.
Two types of costs are generally applicable when the cost allocation ap-
proach is rtxjui red: separable and nonr.eparabl e. Nonseparable costs may b>-
further divided as either semi-separahLe or unseparable. Separable cost-; ire
those which are clearly applicable to a given pollutant, e.g., a specific unit
process attacks only one pollutant. Semi-separable costs are those that can
be shown to apply to only a subset of the pollutants; any subsequent cost a 1 -
locations will also apply to this subset. Finally^ unseparable costs pertain
to those cost items which cannot be clearly allocated to any specific poll. 1-
tants.
" " 1 """ 1 •——— - '
In practice, alternative ordered seta of treatment options may not be
"least-cost" over the entire range of effectiveness. Furthermore, many there
exist technological constraints to shifting from one set of treatment options
m another set. Theoretically, howevei , in the static case, one can con:= ld^r
the "envelope curve" ol least-cost options.
#
-------
Conceptually, the nonseparabl" costs present vie greatest burien of so« Iv5i=i
m the cost allocation approach. At best, only arhi r.rar y cost allocation deci-
sion rules can be applied to obtain the desired data for individual pollutant
MCE ana lys is. Various, potentially applicable dec 1 s ion rjles are descr ibed in
Chapter 3.
Treatment Chain Abatement. A full disclosure of the treatment: chain's
abatement performance is needed for each treatment option. Furthermore, a
breakdown of influent-effluent characteristics for each unit process within each
treatment opt ion is needed in order to determine those pollutants affected 'and
the degree of abatement) as a partial basis of the preceding cos!" allocation
decisions. Such detailed unit process abatement data are not required for ef-
fectiveness weighting; instead, only abatements for each treatment option are
needed.
Obviously, the cost allocation approach requires additional detailed data
pertaining to abatements as well as more detailed cost data. However, unless
the joint cost allocation requirements are cea I ist icaliy approached, trie results
of MCE a;;scssmenl will be highly .'inspect. Theory 1ically, there is no con fit
way, because joint costs cannot be separated.
Effectiveness Weights. Effectiveness weights are required for MCE analysis
when either the effectiveness weighting or the comhinat »on (coot .! Llocation and
effectiveness weighting) approaches are used. In the first case, at least rela-
tive, differential weights are required foe each pollutant in the analysis. tn
the combination case, at least relative weights are necessary within classes of
pollutants mo that abatement levels of each may be .appropriately ago rega ted.
Conceptually, m the most general case, the effectiveness weights for all
pollutants would be based upon their relative environmental damages. However,
such damages will be functionally re La ted to the aggregate levels of each pollu-
tant, and others, in the environment. Hence, it follows that environmental da-
mage function weights are the most suitable. Even more complex functions exist
when dynamic versus static relationships are considered and/or synergistic In-
terpol lu t ant environmental reactions are embedded in the I unction:; posited.
As is evident, effectiveness weights are not based on any of the calcula-
tions or data that are germane to MCE analysis, per se. That is, such weights
arc to be determined exogenously for use in certain MCF. analyses as the
lor aggregating pollutant abatements into a common effectiveness index.
Although effectiveness weights are theoretically definable, such weights
seldom exist and are necessarily subjective when they do exist. This "ta te-of-
the-art limitation handicaps the potential application ot MCE analysis in many
cases where at least partial effectiveness weighting is required to answer some
policy questions.
Whenever effectiveness weights are used in MCE analysis, then effects
should be easily identifiable from the analysis. Ideally, a sensitivity analy-
sis can also be made using alternative weights so that decisionmakers can care-
fully assess the effectiveness weighting effects of their decisions.
-------
1. MARGINAL COST-EFFECTIVENESS METHODOLOGY
This chapter presents the ana lytic.il scheme devised for computing marginal
rost-e ff ectivenesr. {MCE) ratios. The rrethodology is comprised of seven step^:
1. Perform Population Analysis
2. Perform Entity Analysis
3. Identify Elements of Data Base
4. Create Treatment Systems Cost and Abatement Table
5. Compute Marginal. Cost-Effectiveness Ratios
6. Analyze Marginal Cost-Effectiveness Ratios
7. Aggregate Entities
The methodology is organized first to define the universe being studied
{population analysis)f then to disaggregate the universe into homogenous
segments (Step 2) to facilitate marginal cost-effectiveness analysis. Steps 1
and 4 define the data requirments to compute marginal cost-effectiveness and
conjpi le the information in a systematic mariner. Steps 5 and 6 describe
alternative procedures for computing and analyzing IKE ratios. The last step
discusses how to manipulate the entity-level analysis to make policy decisions
relating to either segments of or the entire population.
As suggested in the introduction MCE analysis can be a useful analytical
too) to address a variety of relevant policy issues. Therefore, we have aimed
at developing a comprehensive, yet flexible mechanism that can be incorporated
into EPA's standards-setting process. The diverse nature of the policy ques-
tions that may be posed preclude the mechanical application of the methodology
to a particular policy Issue. Depending on the policy question being address-
ed, some substeps (such as weighting), may not be required. for other sub-
steps, we have presented several acceptable alternative methodologies and
suggest guidelines to aid the analyst in deciding which of the approaches is
most applicable to a particular policy issue.
A second major consideration in applying the methodology is assembling thf>
required data. As constructed, the methodology requires information on the
costs and abatement levels for ail relevant combinations of pollutants and
control processes. We believe that this level of detail is critical to doing
a useful, credible, and defensible MCE analysis. We also recognize that the
additional costs associated with these requirements in some cases may not be
Justified, In the methodology (especially Steps 3 and 4), we address what
data ideally would either be available or developed Cor the purpose of this
analysi-s. We then discuss what modifications must he made where data
limitations exist.
STEP 1; _POPULATION ANALYSIS
Population analysis is the first siep in undertaking an MCE study. This
step identifies the universe of industries and/or processes relevant to such a
study. The objective in defining the universe is to establish the parameters
that the policy analysis is to address; to accomplish this, it is necessary to
articulate fully the policy issues themselves. We emphasize that, at least in
the early stages, these issues should he defined as broadly as possible. By
-------
- 25-
definind these issues broadly, it will then in subsequent -.reps of the method -
oLogy, be possible to address a variety of specific policy issues contained
within the initial framework.
For some policy questions, the population would be all the relevant pollu-
tion- related processes in a specific industry, e.g., the steel industry. But.
for more complex policy questions, comparisons are required across industries
and even across pollutants.
If the policy issue were, "Compare the MCE of controlling BOD in two dif-
ferent i ndustr ies, " the population would be those processes in each industry
involved in BOO' production and control. Similarly, if the policy issue were,
"Compare the MCE of controlling air pollution from boilers," all processes
using boilers in all industries would be examined.
Thus, population analysis invo Ives two sub steps:
1.1 Define policy issues to be analyzed in
broadest possible terms,
1.2 Identify industries and processes relevant to
policy issues.
S T E V 2 : ENTITY ANA I ,Y SI S
For most policy decisions, the ultimate objective is to analyze MCE ratios
for an entire population (e.g., the steel industry, alt toilcis). But popula-
tions frequently consist of diverse components with different pollution
characteristics, control processes, etc. For example, the steel industry in-
cludes three different types of furnaces and more than twenty processes. To
compute the MCE of a proposed government action on such a diverse population
is extremely complex; it is most workable to divide the population into rela-
tively homogeneous segments. Such subsets of a population in this analysis
ace referred to as "entities". In most existing EPA development documents,
the model plant defined for each industry segment is the equivalent of an
"entity". Frequently, this same model plant can be adopted as the entity for
the purposes of MCK analysis. But for some policy issues, an entity may be a
particular engineering process {e.g. boiler units, auto engines}. To avoid
this confusion, we have elected to use the term "entity" Cor the purposes of
this study.
In defining entities for a given population, the key characteristic which
determines whether distinct entities are required is the extent to which dif-
abatement strategies (i.e., unir processes and costs) are used within
the population. Oth^r characteristics os the population that should be
ex amined i nclude:
ftge - 0Idej plants may have markedly different poliut ion and abate-
ment characteristics than new facilities. This may be true even
where the same processes are employed in new and old facilities. If
a known relationship exists between the old and new segments of the
population, some computed "factor* can be used to estimate the MCE of
one segment based on compu tat ions of the MCE for the other. In most
-------
instances, this relationship will be unknown and therefore, will
necessitate the definition of two or more independent entities and
MCE analysis Cor each. This analysis would then be followed by
aggregation based on the proportional weighting of each entity as
determined by its representation in the population.
Si ze - Pollution abatement costs, abatement levels, and techno Loq
wi1! significantly vary based on the size of a particular facility.
Large plants sometimes can achieve economies of scale; in other
instances, certain control technologies may he limited in their
ability to clean large volumes of emissions. To determine whether it
is necessary to define distinct entities by size, the range of sizes
in the population must be exanrined. Secondly, the extent to which
different controls, levels ot" abatement and costs vary by size must
also be determined. Again, if the variation is minor, or if a known
relationship exists over the range of capacities present, only one
entity will be required and a factor can he applied to aggregate to
the population level. Moat frequently, however, MCE analysis for two
or more distinct entities will be required.
Engirteerinq Process - Where pcpuiations are defined as an mdustiy or
a region, numerous distinct ergineer inq processes will he relevant to
an examination of the MCE of » nv i ronment a 1 regulations. Because
these processes involve different control technologies and costs,
they must be examined separately as distinct entities and later be
aggregated based on their proportional representation in the industry.
Thus, if any of these factors are prevalent, and no factor (known rela-
tionship) exists capable of aggregating any variations, then the use of more
than one distinct entity is requi red.
STEP 3: IDENTIFY BLEMBMTS OF DATA BASE
This step of the methodology identifies arid establishes the data specifi-
cations required for subsequent MCE analysis. In this step, we identify: the
pollutants to be included in the analysis (Step 3.1) and the unit processes to
contro1 eacb of these poiIutants (Step 3.2) .
Identification of the appropriate data elements is a critical step to a
workable MCE methodology. By systematically pulling together the relevant
pollutants, potential unit processes, and abatement, levels, this step provider
a use ful roadmap to guide the efforts of EPA's engineering and economic con-
tractors and Lilt imate ly the Agency's own decisionmakers in analyzing the MCE
of proposed actions,
3.1 Identify Key Pollutants
The first: step in creating the data base is to identify the pollutants to
be included. Ideally, a 11 pollutants affecting a given entity would be in-
cluded. For most entities, however, t.ms would make the analysis unwieldy;
there are no significant advantages from including relatively unimportant pol-
lutants. On 'the other hand, "key" pollutants cannot be determined by consi-
dering mass emitted alone. This would eliminate from the analysis most toxic
-------
- 27-
and non-conventiona 1 pollutants which cause significant damage to the
environment and which may be extremely costly to abate.
No set: rules cart be articulated tier deciding which pollutants are to he
included in the rlua base. In part, the specific pollutants included will be
determined by the entity being analyzed and the policy i ssues being
addressed, In the context of development documents supporting standards, EPA
has frequently addressed this exact issue and used reasoned judgment to make
its determinations. If a question arises concerning a particular pollutant,
we do recommend, however, that the analyst err on the side of being over in-
clusive.
It would also be useful at this point in the analysis to group pollutants
into recognized categories. For water pollutants, there would be conven-
tionaLs (e.g., BOD and COD), priority pollutants, and nonconventionals. For
air, the groupings would be criteria pol utants, toxics and all others. Sludge
should be treated as a distinct pollutant . The purpose of these grouping?; is
to facilitate the analysis of pollutant control costs and abatement (Step 4).
3.2 Identify Unit Processes and Treatment Chains
A unit process is a piece of equipment, an engineering process, or a raw
material which is enployed to achieve a level of abatement of a pollutant. In
its most simple; construction, it is a. pi we of equipment which controls one
pollutant to a specified level at specified costs. Frequently, however, a
unit process will aff ect more than pollutant. Where this occurs, it must be
incorporated into this analysis. A treatment chain is one or more unit, pro-
cesses used to abate pollution. For example, physical coal cleaning and a
scrubber {both distinct unit processes} when used together to abate sulfur
dioxide form a treatraent chain. (A single unit process, such as an electro-
static precipitator, would also be consiaered a treatment chain.) Treatment
chains occur most frequently in controlling water pollution. Here, higher
levels of abatement, require adding new cnit processes to existing controls.
In many of these instances a particular ordering of additional unit processes
(e.g., remove a certain amount of suspended solids before going after phenols)
is required and must be followed in the development of treatment chains for
the analysis to be cred ible¦
The first task in this step is to identify all process-specific treatment
chains {made up of one or more unit processes) capable of abating a particular
pollutant. Where a treatment chain affects only one pollutant, this task will
be straightforward; where process or chain affects two or more pollutants,
complications arise. Specifically, a met hod must be developed to apportion
costs of abatement, among the affected pollutants. This problem is discussed
in Section 4.4. For this step it is necfssary only to identify those treat-
ment chains which affect more than one pollutant.
A second coraplication is that many treatment chains can achieve more than
one level of abatement. For example, an electrostatic precipitator can
achieve a full range of abatement levels for particulates by changing the area
of magnetic plates it contains. In cases where treatment chains can be
designed to achieve a full range of abatement levels, a continuous function
would exist for its costs. This problem is dealt with in detail in Step 4.4.
-------
- 28-
Again foe this step, all that need be noted is whether a unit process is cap-
able of achieving only one level of abatement, can achieve a full range of
abatement levels or only several distinct points. The latter would be the
case if an electrostatic precipitator could only be designed to meet 0.5
1 bs/10BTU or 1.0 lb/106 BTU and not any abatement levels in between.
This is necessary because processes or chains which control to different
levels at different costs will be entered separately into the Treatment
Systems Cost and Abatement Table derived in Step 4 and must be analyzed as
distinct control alternatives.
The process used in this step is almost identical to that currently used
by EPA in its standards setting process. The one divergence is that EPA estab-
lishes standards based on available technology (here termed a treatment chain)
and then does cost analysis for the proposed alternative standards, MCE
analysis stops short of using standards and instead focuses on level of abate-
ment. This is necessary because a particular unit treatment chain may abate
pollution to a specific level, but when used in the presence of or in combina-
tion with other equipment, or with different material inputs (e.g., higher
sulfur coal), it may attain a slightly higher or lower level of abatement.
The costs of moving back to the original standard may be very high- Thus, by
using abatement levels, we are able to achieve a more flexible analytical
framework for MCE evaluation.
The final product of this Step is a data base table (Exhibit 31 listing
all "key" pollutants and their relevant treatment chains. For each treatment
chain. It is essential that the particular' unit process or combination of unit-
processes be identified. Where relationships exist among pollutants, control
processes or confcinations of the two, these must also be identified in this
table. For example, total suspended solids raay be defined separately from BOD,
but they are indistinguishable from the perspective of abatement (i.e.,
reducing one also reduces the other) ; t his should be noted under these pollu-
tants. When unit processes are combined, the costs of the existing controls
may be altered. This too must be notec in the table. Finally, the use of a
treatment chain to control one pollutant sometimes affects the control of
other pollutants and must be described in the Table.
A hypothetical example illustrates the contents of the Data Base Table.
In this example, SO2 and conventionaIs {e.g., BOD and COD) are the only key
pollutants. For each, we describe the alternative treatment chains, the rele-
vant relationships which exist, and the abatement characteristics. This table
defines in general terms the inputs Into the Treatment System Costs and Abate-
ment Table developed in the next step.
STEP 5; CREATE TREATMENT SYSTEM COST AND ABATEMENT TABLE
The Treatment Systems Cost and Abatement Tablel brings together the
entire set of key pollutants, the appl,cable treatment chains and the abate-
ment levels they achieve, anc their costs of pollution control. This section
first describes the Table# followed by sections which discuss how to determine
the possible number of coabinatlons (4.2), the relevant coat factors (4,3) and
the effectiveness of the treatment chains (4,4).
-------
- 29
4 - i Dps i gn i rig the Tab l_e
The Treatment Systems Cost arx^ Abatement Table is destine! to serve a;i
the fundamental data base for evaluating any MCE-re la ted policy questions
which could be asked about the specified population. To accorap1ish this ob-
jective, it must be as comprehensive as possible. It must present the costs
of pollution control broken down into detailed estimates assignable to speci-
fic pollutants. 11., milst present all lev* Is of effectiveness attainable by
each treatment chain. More importantly, it must examine the costs and effec-
tiveness of each plausible combination of pollutants and treatment chains.
For the purposes of this analysis, combinations of pollutants and applic-
able treatanent chains are referred to as "treatment systems." Exhibit 4
illustrates the design of the data base. The initial task is to categorize
the key pollutants into criteria and hazardous pollutant groups for air;
groups of convent iona Is, priority pollutants and ncn-convent. iona Is for wa t r ;
and sludge.
^Thi:j table is presented only as a prototype for a specific analysis.
The data required and that available will vary significantly and could be com-
bined in a number of equally acceptable alternative formats.
EXHIBIT 3
BLMEHTS OF DATA BASE
Abatement
Characteristics
Creates sulfur sludge; Requires met eas-
ed use of water; Can be designed to
various levels.
Creates sulfur sludge and requires dis-
posal of coal tail washings; PCC "re-
duces sulfur content and requires h'ss
scrubbing to meet a specified standard;
Can be designed to abate to various
levels.
Western coal contains less sulfur
but has a lower heating value so
more must be used to produce the
sane amount of electricity; Use of
LSC adversely affects operation of
o iec trost.a t ic p rec i p i ta tor.
Treatment
Pollutant Chain
so2 Flue gas desulferizer
(TOD)
Physical Coal Clean-
ing (PCC) and PGD
Low Sulfur Coal (LSC)
-------
EXHIBIT 4
TREATMENT SYSTEM COST AND ABATEMENT TABI.F
rs e aw e f*r
SYSTEM
COKB IfiAT JONS
Convetuinntl*.
Treat . Cos Abac »
Chair. R f
CC
HTPOTMF~:<;AL est it
POL! I'TANTS
WATER
Pr i ,-iri t v
Ph«?rsa!*
Treat. " n% t Abate
Chain R E
Hon i: i ->ra ' s
K»i:
Treat . € .« t
Cha "•
50 400 300
XX
100 100 50
CC+VF mm- 300 200 OC+VF non- .01 ,001 CT
sep
YY ?0
CC+VF r.cxi -
sep
SOb liO
•rp
X-VT nm~ .01 .001
sep
0
:o
mp
SL'JDOE HONSFPARABI.E3
Abate
R E
f ri ";er i a
rsr
Treat . C,?s ; Aba".
Chain '4
ESP i 50 35
ESP 150 115
Hazardous
Trace Metals
Tsvjr . Kasl Aba;
f~ i, a . n P
0
0
Treat .
CS* s n
CF+L 25
CPU 25
. I <'.1S . •. os i
Chiui
cr»VF :on
CC»VF I 50
fast
11 ii
? 1'
I
r -;: , t «m' v. sr.: •*
?atari: Err;; t.! ;
-------
- 31-
Conventionals (BOD and COD)
Chemical Coagulation Affects both BOD and COD
Vacuum 1 titration (VF) Woi ks only after 1 nst: a I ) a-
and CC t Kin of CC.
The specificity of these groups will be determined by the level of detail
of the available data. Ideally, all costs of treatment chains will be allo-
cated to specific pollutants, thus eliminating the need fox groups. Realis-
tically, however, whom joints costs exist or where i nadequa te data is avan-
ahin the analysis can best be per formed w th groups of similar pollutants.
In Exhibit 2 we present the Treatment Systems Table. In doing so, we
recognize that, frequently only limited data will be available, thereby
requiring modifteat ions to the prescribed format. As discussed above, one
simpl ication is to group similar pollutants, A second simplication involves
the absence of all the possible combinations of treatment chains. The MCE
analysis can still be employed with a limited number of treatment systems, hut
the results may be less than optimum. The format of the Table allows the
analyst to determine when other points are possible but currently unavailable,
and require additional data is required.
For each comb i nat \ on of pollutant s, the required data includes:
(I) treatment system employed;
{2) the total cost of that system;
(3) the amount of pollutants removed by each treatment system; and
(4) the amount of pollutant still emitted.
A separate column to the right of the specific pollutants isolates those
costs which are "nonseparable". This ref ers to those treatment; chains which
affect more than one pollutant and, which m some instances, must be assigned
among the affected pollutants. As 111 ust.rated by the example, where non-
separable costs exist tor those affected pollutants, "nonsep" would be written
under their cost headings, and the total cost of the treatment chain would
appear under the column labelled "Nonseparable Costs" to the right.
The final two columns would be summations of total costs and abatements.
Nonseparable costs must be distinguished from situations where two or more
pollutants are affected, but where the treatment chain is employed primarily
to remove a particular pollutant and ancillary effects are incidental.
Instances of ""incidental removal" costs should not be considered nonseparable,
all costs should be borne by the pollutant which requires the treatment
chain. The rows would consist of all the possible combinations of treatment
chains and pollutants. By costing out each of these treatment systems, it
should be possible to determine which treatment chains achieve a given level
of abatement for the least coat and how MCE can be improved by trading-off
abatement levels among pollutants.
-------
- 32-
4.2 Determining the Number of Possible Combinations
In order to guide the analyst and to determine the amount of resources
necessary to undertake the analysis, it is useful to identify the total numbetr
of combinations or treatment systems to be examined. This computation draws
from Steps 3.1 and 3.2 where we identified the key pollutants, the related
treatment chains, and levels of abatement.
For each pollutant, the number of combinations (rows in the table) is
determined by the number of different treatment chains Identified in Step
3.2. For example, particulates can be controlled by an electrostatic precipi-
tator to reach abatement levels of 0.1 and 0.05 x 10^/BTU and by fabric fil-
ters capable of reaching a control leval of 0.03 x 106/BTU. Thus, there are
three different options for particulates, each of which must be compared to
all the other possible combinations of options for other pollutants. In a
four pollutant example having three options for two pollutants and two options
for the remaining two pollutants, the total number of rows would be 36
(3x3x2x2).
The number of possible treatment systems calculated in this manner repre-
sent a ce ili ng, with the actual number of options being cons ide rab Ly fewer.
Some combinations of treatment chains are mutually exclusive or redundant and
can be eliminated at the early stage of the analysis. Where the policy issue
being addressed concerns a static situation (i.e. , only one time period), any
treatment system which does not achieve a certain level of abatement at the
least cost can also be eliminated.
The one situation where the method understates the true number of treat-
ment systems would be where a continuous function exists for a particular
treatment chain. Because the chain can achieve any abatement level over a
specified range, it does not fit neatly into this part of analysis and must be
ex eluded.
4.3 Calculate Cost of Control
Calculating the cost of controlling each pollutant in each row is one of
the more complex tasks in the MCE methodology. The costs used should include
all elements relevant to the particular system application. These elements
include investment and operating costs associated with the systems, generated
pollutant disposal costs, interrupted production costs, and, for in-plant/
process change systems, costs associated with changes in production capacity.
Cost coverage is not the only complexity. Costs are best computed by the
basic building blocks of the method, i.e., unit processes. Some unit pro-
cesses attack only one pollutant, but others attack several pollutants. When
considering treatment chains and systerss, cost assignment problems exist.
When a unit process involves only one pollutant, assignment is easy. This
situation describes what are called "separable coats". Where mult ipollutants
are involved, costs are "nonseparable". In some cases nonseparable costs are
assignable to a group if not all pollutants are involved. These costs are
termed "semi-separable". Costs are termed "unseparable" when they can be
assigned only to all pollutants attacked by the particular chain.
-------
- 33-
The complexity of the cost analysis is reduced to the extent that separ-
able coats can be identified. Sound engineering analysis can enhance this
identification. In some cases, for example, a particular unit process is in-
cluded in the treatment chain to attack a particular pollutant. The reduction
of other pollutants by this process is often termed "incidental removal." It
seems logical to assign all costs associated with the unit process to the
prime pollutant where incidental removal can be demonstrated.
It will not be possible to apply the incidental removal criterion to most
cases where multipollutants (hence joint costs) are involved. Some other
method for assigning costs is required. ?or the purposes of filling in the
Treatment Systems Cost and Abatement Tabls, all that is necessary is to enter
these nonseparable costs under the specified column to the right of the dis-
tinct pollutants and to place "nonsep" in the cost column under those affected
pollutants. The problem of allocating these costs need only be addressed when
calculating certain MCE ratios and is discussed in Step 5.
A continuous cost-effectiveness function also requires special handling in
completing the Table. In this instance, we may want to identify specific
levels of control, calculate costs for these derived from the continuous func-
tion, and enter these in the Table. Alternatively, it may be more desirable
not to include distinct points for this pollutant in the Table. Instead,
these would be incorporated directly into the MCE analysis in Step 5.
Regardless of form, it is essential to include all relevant costs of com-
pliance in the Table. Different types of unit processes require different
types of analysis to determine total cost of con?) 1 iance. These are discussed
below.
Costs for"End-of-Pipe" Treatment. Costs for end-of-pipe treatment, used
genericly to mean treatment upon discharge for a particular production plant
or process, can ba estimated in a relatively straightforward manner. Basic-
ally only two types of costs are involved: (1) investment costs, and (2)
annual costs. Investment costs are defined to include the one-time costs
associated with designing? procuring, installing and checking the unit treat-
ment process. Annual costs are defined to include all recurring costs
associated with operating, manitoring and rehabilitating the unit process.
Various general definitions of the relevant investment costs and annual
costs may be presented. The intent, however, is to include all items that are
affected by a decision to proceed with each prospective treatment option. In
most cases, these costs included at least -
Investment Costs
Con s tr uc t i on
Equipment
Monitoring Equipment
Eng i neer ing
Contingencies
-------
- 34-
Annual Costs
OSrM Labor
Maintenance
Sludge Disposal
Energy
Chemicals
Mon i tor i ng
Energy cor.t
-------
1r, _
-J* ,3
pollutants (chlorinated hydrocarbons) can be removed using activated carbon,
and others {heavy metals) may be removed by filtration. Thus, depending on
influent and treatment process, the sludge generated could contain concentra-
tions of priority pollutants high enough to require special sludge disposal
techniques. It so, these costs are relevant to water pollution abatement and
woo I d show up in the Table under the sludge column.I
Interrupted Production Costs; The installation o£ abatement equipment
could cause a production stoppage or slow-down at the model plant or process.
IC this occurs, the costs associated with the interruption are relevant to
pollution abatement. Quantifying these coats can be a difficult task
requiring a case-hy-case analysis. The following guidelines will be helpful
l n such an analysis;
(1 } Value ot" Lost Production: Lost produc t ion can, when data are
available, be valued in terms of contribution to profits -- that
is, the selling price less the variable (or out-of-pocket} cost
ot production. It is generally more difficult to determine if
production is lost or merely delayed, Unless a plant is
operating at or near capacity ami the interruption is signifi-
cant, the cost of lost production may be more illusionary than
real. In those cases where production can easily be made up at
a later" Ate, the value of lost production approaches zero.
(2) Shutdown or Slowdown Costs; Some costs may be incurred whether
or rot the plant is in operation. Many of these costs are fixed
in nature and would include auperviBcry salaries and rent. All
such costs are relevant to pollution abatement.
Ijv-jP Lant Control/Process Change Costs: ft variety of in-plant controls or
process changes can be postulated. Changing from high-sulfur to low-sulfur coal
to reduce SO2 or: using staged combustion to reduce M0X ace two examples from
the coal-fired power plant case. The textile case provides several examples
useful to Illustrate cost analysis methodology so that the discussion here cen-
ters on water pollution in general and the textile industry specifically.
In-plant. controls and process changes are measures a firm can take to reduce
water usage (e.g., water reuse and reduction) and the production of pollutants
{e.g., substitution and material reuse). In some instances, there may be both
costs and benefits associated with in-plant controls. Hence, "relevant costs"
can be either positive or negative. Cost will include investment and operating
categories such as those discussed above, These costs will, however, be reduced
by the benefits derived which include;
(!) energy savings associated with the reuse of cooling water
(2) potential reduction in the cost of water
(3 J potential reduction in process material coats through reuse
(4) 3ale of any residuals from control process.
Process changes fas opposed to in-plant controls) tend to be complex and
costly. So Ivfjnt- processing can, for example, be substituted for conventional
processing for scouring wool and some knit fabrics and for finishing fabrics
^Notn that sludge disposal was included as an element of operating costs.
-------
- 36-
sensitive to water, but with limited eff ec t lveness. Changing material and pro-
cess flow procetij res from batch to continuous, substituting standing baths for
moving baths or containing operations where possible, tend to decrease hydraulic
loading. Newer equipment tends to be "ess polluting. Pressure dyeing uses dye-
stuffs more ef£iciently, uses less wateT and reduces the levels of toxic dye
carriers compared to atmospheric dyeing.
The determination of costs relevant to process changes is slightly more com-
plex than determining the cost relative to in-piant control. Coat estimates for
both the new and the existing process are required using investment and opera-
ting categories similar to those discussed above. These costs should be time
phased over some reasonable planning horizon with any interruption of production
noterf so that the value of lost production can be considered. The new process
costs should also include the cost of .abandoning the old system Jess salvage
value, A first estimate of the net cost of the process change is the cost of
the new process less the cost of continuing the old process totalled over the
planning horizon.
This net cost (which could in theory be negative) represents the cost of
abatement relevant to the process change. In some cases, the new process may
increase production capacity significantly. When this occurs? it does not seem
logical to charge ail of the costs of the change to pollution abatement provided
that the increased capacity can be utilized at the model plant. The separation
of costs associates with capacity from the cost associated with abatement re-
quires some add i t i ona I analysis.
When capacity is increased 'and the coat of production does not decrease
costs can be allocated between abatement and extra usable capacity on a per-
centage basis, that is
P
R
A P
n.
where RA is the percent allocated to abatement, PQ is the original capa-
city and Pn is the new capacity. The percentage allocated to capacity (R(.!
is
~ po
pn
These ratios would assign too much costs in those cases where unit produc-
tion costs decreased. Under these conditions, the percent of net cost allo-
cated to capacity (R ') jg
P M
nuo
Rc
Pnun
where 0o and Un are the old and new unit costs. Since unit cost is the
ratio C./N, the equation reduces to
'"nPo
"o^n
-------
- 37-
whera CQ and Cn ace the old and new process costs respectively. The per-
cent charge to abatement {%*) is
ra = ...
C P
o n
Note that RA' is always less than Rft so long as unit costs have decreased,
4,4 Computing Effectiveness in the Table
Quantifying effectiveness in the Tabl<- results in me asus of the Level of
abatement and of the level of emissions ftorn the entity wli^n the unit process
coirt:>ination is applied. For each key pollutant, an appropriate physical mea-
sure should be devised. For example annual kilos removed and kilos emitted
should be used for most conventional pollutants. To the extent possible, all
measures should be expressed In the same units.
Both the amount of pollution removed and the amount emitted must be in-
cluded for clarity. For some treatment chains, the amount removed will
increase, but at the same time the amount emitted will also increase. An
example of the phenomenon would be the use of high-sulfur coat with a.scrub-
ber. Even though the scrubber is operating properly, the amount of sulfur
emitted may increase because more sulfur is being put into the system. It will
be important in answer ing policy questions', and performing marginal analy- sis to
identify this occurrence. By reporting both kilos emitted and removed, we also
will be able to calculate percent removal, which may be useful to com- pare MCE
ratios.
One other aspect of effectiveness in the Table is the lack of data in the
"Nonseparable" column and Total Effectiveness column. Although it would be
useful to have an overall effectiveness measure for nonseparables and the total
system, it is not possible to create one until weights are assigned to
pollutants. In the next section, this wii1 be discussed further,
STEP 5; COMPOTE MCE RATIOS
Having identified all possible unit process combinat ions and their asso-
ciated costs and levels of abatement in the data base Table {Exhibit 2), the
essential elements are now present for doing MCE analysis. By enumerating all
the possible relevant combinations of unit processes, the data base contains
enough detailed information to address a variety of policy issues.
S.1 Define the Policy Issues
The exact data dt awn from the Table will depend on the particular policy
question being addressed. Drawing from the table of costs and abatement, it is
possible to analyze the MCE of the relevant policy questions defined in Step I
including:
-------
- 38-
• What is the MCE for controlling a particular pollutant
rat alternative, more stringent standards?
• What is the overall MCE of pollution controls affecting
an industry?
• How does the MC^ of controlling a pollutant in one
industry compare to that of controlling the same pollu-
tant in another Industry?
5.2 Identify Relevant Rows in Table
Once the policy issues are defined, the next step is to identify the rele-
vant rows in the Table which directly a ffeet them. As expressed in the pre-
vious section, the Table was constructed to be comprehensive and therefore
represents an extremely flexible policy tool. If a policy question involved
the MCE of changing the standard for a particular pollutant, then only those
rows which accomplish this, while holding all other pollutants constant (or hy
allocating costs to those affected pollutants where costs are nonseparable)
must be analyzed.
Furthermore, the Table presents data for all possible comb mat ions of
treatment chains, abatement levels, and costs for each pollutant. In static
situations (only one time period being examined), where two treatment systems
can achieve the same level of abatement or where one can achieve a higher
level of abatement for leas money, it is possible to eliminate the more expen-
sive treatment system from this analysis. Where more than two time periods
are being considered {e.g.# the use of interIm standards, the MCE of alterna-
tive dates of compliance), all unit processes, regardless of costs for meeting
any one level of abatement, must be analyzed. This Issue will be examined in
greater detail, in Step 6.
'). 3 Assign Nonseparable Coats
Having identified the rows relevant to the specified policy issue, we now
have a measure? of costs and abatement for each pollutant. Before using this
data to compute the desired MCE ratios, a final manipulation may be required.
To determine marginal costs where noneeparable costs exists, some allocation
scheme must be enployed. Ideally, all unit processes can be defined in such a
way that there exists a direct, one-to-one relationship between that treatment
chain and the level of abatement for a particular pollutant. Where this is
not possible, some form of cost allocation will be required. Exhibit 2 iden-
tified these instances. Remember however, that a distinction must be made
between situations where two or more pollutants are significantly affected
{i.e., where joint costs must be allocated) and where only one pollutant is
significantly affected, with others being incidentally affected and therefore
not assigned any costs.
We analyzed numerous different ways to assign nonseparable costs and pro-
pose several alternative met hods. Each of these methods is described and
illustrated using the following example of nonseparable costs.
-------
- 39-
Treatment Pollutant Pollutant
Cba ins
Affected
Costs
Removed
A
13
10
UP,
A,B
30
10 CAS
MB)
UPj
B
50
5 '
The choice of these methods to be employed will depend in part on the
availability of data and the particular policy quest ion being addressed. One
of these methods may require significant amounts of added resources to perform
the analysi;; (i.e., separate facilities) ; another is based on the questionable
assumption that credible relative weights can be assigned to pollutants {i.e.,
effectiveness-weighting approach). We emphasize that there appears to be no
correct way to assign joint coats. We have listed the methods of allocating
costs in an order of preference established for the general case. For .•speci-
fic cases, the best advice we can offer is to apply two or more of the methods
and compare result;--;. What must be at all times avoided is a mechanical appli-
cation of any of these methods. If done, any of these methods could produce
results which will be misleading.
Target Pol]utant; Where a treatment chain was employ el for a particular
pollutant, it may be most credible to assign all the costs to this one pollu-
tant. The reasoning behind this approach is that the abatement of target pol-
lutant was the primary objective and could be achieved only by incurring the
total costs; thus, any effects on other pollutants would be borne by the tar-
get pollutant. This would however, tend to understate total effectiveness for
the costs incurred (i.e., other non-targe ; pollutants increasing or decreas-
ing) . In many instances, a target pollutant cannot be identified. For
example if tJP2 was used because it was th-a least costly way to meet the
requirement of 20 units of removal of pollutant ft, we could assign ail of its
costs to A. Thus, the total cost of removing 20 units of A would be 40.
Ratio of Separable Costs: In some situations where non-separab Le costs
occur they may only comprise part of the total costs of controlling the
affected pollutants. Where separable costs do ^xist their ratio can be used
as the basis for allocating nonueparable costs. Using the example above, the
separable costs for A and B are 10 and 50. Using a ratio of 1/6 and 5/6, the
total costs of UP7 (30| would be allocated to 5 to pollutant A and 25 to
pollutant B.
4 - 30 To-Tsir ° 5
B = 30 To^V - 25
Allocation Rased on Effectiveneas-Wej.ght inq: This alternative taKos info
account the environmental damage averted by the unit process and uses this as
the basis ot cost allocation. It serves as a proxy for the benefits received
by abating pollution.
-------
To use the above example, we mast assume a damage function exists between
pollutants A and B. We will assume that B is a toxic substance and therefore,
cause 1.5 times more damage to the environment than A. A's share of the cost
of UP2 would be;
A => 30 (5) (12) = 25.7
5 (12) +-10
B = 30 10 = 4 . J
5 112} + 10
Separate Facilities: Another option is to determine the costs of con-
trolling the pollutant to the des ired abatement level using exogenous means.
This method requires several separate engineering analyses to identify the
costs of each control option.
In our example, we would determine separately the costs of control Iinq
pollutants A and B at the incremental levels (e.g., for A going from 10 to 20
units of pollutant removed).
Equal Allocation of Costs: This least preferred method allocates cost
equally across all affected pollutants. In doing so, however, we are forced
to ignore the amount of each of the relevant pollutants removed and the rela-
tive costliness for each. In the above example, using this allocation method,
the costs would be equally divided, to pollutants A and B, 15 units each.
5.4 Assign Weights
The denominator of the MCE ratio derived from the Table will show some
physical measure {amount removed or emitted) of the effectiveness of the
treatment system. In sonae situations, this information alone will be adequate
Cor making a policy decision. For example, if the decision involves a single
pollutant an MCE ratio can be derived.
Where several pollutants are invo ved in the policy question some method
is necessary for bringing them together into one MCE ratio. Without this
step, the policymaker has a aeries of disjointed MCE ratios Cor each pollutant
and no acceptable means of comparing them. Ideally, the method of comparing
effectiveness across pollutants would be baaed cm the relative damage func-
tions of each pollutant. The state of the art for developing damage functions
for pollution emissions has not yet developed to a point where it can he
directly used appropriately in this methodology. In its place we suggest
using weights for each pollutant baaed on subjective estimations of relative
damage done by each pollutant.
We have designed the methodology to avoid this problem whenever possible
ami to minimize its impact when it is unavoidable. First, when only one pol-
lutant is being analyzed, weights are not required, so this step of the
methodology is inapplicable. When two or note pollutants are being compared
and a single MCE ratio is desired- weights are employed as one of the last
steps in the methodology. They are inserted 3imply as a term to be multiplier
by the physical measure of effectiveness (e.g. usually kilos emitted) in the
demoni nator of the MCE ratios. If no weight is assigned, it Implicitly means
that ail pollutants are considered to be equal.
-------
- 41-
Because weights are imposed as a distinct substep late in the methodology,
it is a simple matter to redo the analysis using different weighting schemes.
No additional data will be required. Thus, where some weighting scheme is
required we suggest the use of sensitivity analysis to determine the effect of
alternative weighting schemes.
STEP 6: ANALYZE MCE RATIOS
The next step in the methodology requires the analyst to evaluate the MCE
ratios before drawing conclusions. This seep is critical because the derived
ratios are subject to misinterpretation; they must not be mechanically applied
in the dec 1sionmaki ng process. These situations where misinterpretation us
likely are; (a) where thresholds are used as the determination of acceptable
MCE ratios; (h) where major polluters are permitted to continue at the ex-
pense of minor pollutors; and (c> where the percentage of removal var ier,
(e.g., one firm moves from 10 to 50% abatement, another moves from 80-90%
abatement). Two additional situations are discussed in this section. These
involve situations where different time periods are included as an aspect of
the policy issue being addressed. In theory, MCE analysis is timeless. To
use this analysis in situations which are not static requires certain manipu-
lations discussed in sections (d) where interim standards phased-in over a
period of time are being used; and (e) where the tilting of implementation of a
standard remains at issue. Each of these is discussed below.
A. Using a "Threshold" to Evaluate MCE Ratio
Among the most important considerations when analyzing the MCE ratios
is to understand its role in choosing standards. Ideally we need the MCE
curves and pollutant-based curves. However, these curves do not exist. In-
stead we have discrete points for particular industry segments and industries,
none of which is truly comparable. But in many cases it may be possible by
manipulating the available data to fit a representative curve through the
existing points. A typical set of points is shown in Exhibit 5,
BXHXBIT 5
TYPICAL TOTAL COST DATA
TC
• B
• A
Abatement
25 50 75 100
-------
The points represent unit processes or treatment chains and correspond to
rows in the Treatment Systems table. When only these poi ntn exist, MCE th-*n
becomes incremental cost effectiveness (3ee Chapter 2 tor a more detailed dis-
cussion) . So long as the analyst is aware of this distinction no problems
should necessarily arise.
However, the lack of continuous curves does create problems when "thres-
holds™ are iimportant. A threshold is a dollar amount per unit of abatement at
the margin which has been judged reasonable value for firms to spend. A thres-
hold could be set based on the costs of similar treatment at a putalicly-ownwd
treatment works or it could einploy the existing MCE as a floor. For example if
municipal waste treatment facilities spend $3/ton at the margin, this might hp
considered a threshold for private industry to meet. Because the Table uu.-i tides
discrete poi nts, it is sometimes impossible to find a value at the threshold.
Furthermore? the threshold may be such that for one industry compliance will
require that only 2 percent of the total pollution of the industry be abated: In
another industry 100 percent removal may be neces sary. Yet the economic pro-
blem faced by these industries of hew to pay Cor the cleanup is not considered.
It could be that the industry which should remove 100 percent of its pollution
cannot obtain the resources to install the necessary equipment. Each of these
two problems is discussed in the succeeding paragraphs.
Exhibit 5 shows what the Treatment System Table, looks like in graph form
when discrete points are identified. Theslope of the total cost curve is thp
in.u gi nal cost curve. This is shown in Exhibit 6.
EXHIBIT 6
TYPICAL MCE CURVE DERIVED FROM EXHIBIT 5
• D
A
• • C
• B
25 50 75 100
WSATEMEHT
The Threshold Isn't Defined By The Technologies: If the marginal cost per kilo
were set at ,05 dollars per kilo aa a threshold, would this graph supply an
answer? The answer is that we could not recommend an action because the
threshold lies between two discrete points — C and D. If the new standard
imposed the industry will be exceeding the threshold; yet at I he old standard it
is not at the threshold.
Mc. .06 -j
.05 -
.04 -
I
.03 -
.02 -
.01 -
-------
- 43-
EXHIBIT 7
MCE CURVE FOR TWO INDUSTRIES
Industry A
Industry B
MC
25 50 75 100
25 50 75 100
ABATEMENT
Industry Equity Issues and the Threshold Concept; Exhibit "7 shows hypotheti-
cal MCE graphs for two industries. The MCE graphs again are discrete points
represented by the available technologies. In addition to the threshold pro-
blems, in this case an additional problem arises. If the threshold is 52 per
kilo, then Industry A will have to remove 100 kiloa, and Industry B will have to
remove #5 kilos. Thus, it Is possible that Industry A will have to pay far more
than Industry 8 in order to remove the efficient amount of pollution. The issue
then is "Can Industry A afford these costs?"
B. Relative Significance of Control. Consider the example where several
industries have different existing standards and different proposed future stan-
dards for some pollutant. The policy issue is which future standard will create
the greatest benefit for the cost. Determining the most cost-effective new
standard is not difficult, using the procedures shown above for a single pollu-
tant in a single industry, it is possible to create a list of ICE like those
shown belowj
Industry
A
B
C
D
E
F
Marginal Cost-effectivens
of Possible Standards
2.0
2.5
2.3
3.0
4.0
.5
Percent of Total Pollution
Emitted
12
12
12
60
3
1
-------
- 44-
tJslnq t hnse values, EPA could develop a MCE-baned priority ached a io for
implementation of regulations. However, in many ways, this would not be pru-
dent. For example, Industry F whose MCE is .5 is responsible for only 11 of
all the pollution of this type, Industry D causes 60* oE this type of pollu-
tion, 'Even though Industry F is the most cost-effective target, it would not
accomplish much overall to tighten Industry F's standards alone.
Another example of the need to combine MCE with other data ishown in t in-
case where EPA knows that, if it can remove X-amount from the environment,
then the environment will be balanced with respect to this pollutant. In the
above example, if the amount that could be achieved by moving from the currant
standards to new standards in only some of the industries is m excess of X,
then the agency might want to impose the standards in order of MCE. However,
this approach creates several industry equity arguments, because as shown j n
the example, it might he possible to cleanup this pollutant rn a natisfactot y
level nationwide and could be achieved without tightening the standard on the
biggest polluter — Industry D!
This example suggests that EPA must incorporate equity consi dent ions inco
its use of MCE analysis. Additionally,, separate decision rules (e.g. , all
sources must achieve a minimum standard) may be required.
C. Comparing Amounts of Pollutant Removed Over Varying Intervals. In
most instances, we will not have continuous functions for unit processes.
Instead, we will have a series of d istl net points determined by the level o*i
abatement that specific unit processes are capable of achieving, when com-
puting MCE ratios for different intervals, the results may not be directly
comparable. For example, we may want *:o compare the MCE of going from 30 to
60 percent removal for BOD in one industry to going from 80 to 95 percent
removal of the same pollutant in another industry. According to MCE theory,
because neither the costs nor the measure of effectiveness are being held con-
stant, the ratios should not be comparable. But if effectiveness weighting
were used and a kilo removed of BOD from one industry equaled a kilo removed
from the other, the ratios would be comparable. Thus to be efficient, EPA
should require the next kilo removed to come from that industry which has t he
lowest MCE ratio.
Where inte rvals are large, however, an "efficient* decision may not be
possible. For example, suppose that SPA's goal i s to decrease BOD effluent
by 100 kilos. It must choose between Industry A which could install unit pro-
cess two at a cost of $20 to remove 110 kilos (MCE equals ,18) and Industry B
which could install unit process Y at a cost of $10, to remove 80 kilos {MCE
equals ,1255. No obvious solution exists. It may be necessary to accept less
clean-up to achieve a better coat-effectiveness.
If continuous functions existed for the unit processes in each of these
industries, the above problems would not arise. To optimize clean-up in this
situation, it would take the mix of abatement in each industry totalling 100
kilos which minimized total costs. Without continuous functions, however,
some judgment will be necessary where the MCI of large intervals are being
compared.
-------
~ 4 5 -
D. Using Interim Standards
Where EPA's main concern is deciding which of alternative, more stringent
standards to adopt for a specified date, timing does not enter into the MCE
analysis. All that is required is to estimate the incremental cos! ; of com-
pliance for the proposed standards and to compare these costs for the incre-
mental levels of abatement that result. Bjt EPA often chooser; to adopt a less
stringent interim level of required abatement in an effort to lessen the bur-
den its regulations impose on industry. This interim standard is followed at
s>ime later date by a more stringent target standard. (This is one of the op-
tions EPA is currently evaluating for air pollution control on new coal-fired
power plants,)
The use of interim standards affects two aspects of the MCE methodology.
When employing the traditional methodology m situations where alternative
levels of abatement are being compared for implementation at the same point in
time, wf can reasonably assume a firm will select the least-cost, treatment
chain to meet each of the proposed standards.
Exhibit 8 graphically demonstrates this point. ff EPA were deciding which
of these standards {1, 2, or 3) to adopt at a particular time, it would be
reasonable to consider only the I east-cost unit process for each -standard.
Thus, the numerator of the cost-effectiveness ratio is based on compliance
costs using an estimate of points A, B and C on treatment chains 1 and 2.
Where the use of interim standards is being considered, examining only the
least-cost treatment chain to comply with the alternative standards imposed it
different points in time would be misleading. With the likelihood of shifting
overtime from a given standard to a more stringent one, it may be less costly
for a firm to select a unit process which actually costs more than another at
the initial period. Turning again to Exhibit 8, we observe that using treat-
ment chain V at 3.5 (point 8) is the least cost way to comply with Standard
2. But the MCE analysis would suggest thet treatment chain 1 be emnIoy*:"1. In
moving from Standard 1 to Standard 2, the marginal cost of maintaining treat-
ment chain 1 is "3 (4 minus 1). Instead, if we were to shift to treatment,
chain 2 to meet standard 2 , the marginal cost would be 3.5 (3.5 minus 0).
More accurately, the marginal cost of shifting from treatment chain 1 to 2 to
comply with standard 2 would be 3.5 units less any salvage value from abandon-
ing treatment chain 1. But pollution control equipment is generally plant
specific and would have minimal salvage value. For this i»ason, future calcu-
lations will omit this consideration.
Thu f rom a marginal cost standpoint in situations where two or mote
time periods are being considered, it is essential to examine the full range
of possible unit processes to comply with each standard. The burden this
requirement iiiposos on EPA will be reduced to the extent that the unit pro-
cesses being considered have a limited range of applicability and therefore,
are incapable of reaching certain standards. Exhibit 8 illustrates this
point. Treatment chain 2 can be used to canf>ly with abatement levels 1 and 2,
In it as it is currently designed, cannot remove an adequate amount of po! lotion
to coBfjiy with standard 3,
-------
- 46-
EXHIBIT 8
TOTAL COST OF CONTROL OPTIONS OVER TIME
1979 1982 1985
STANDARD/TIME
In addition to necessitating that we examine the costs of compliance for
each unit process and not just the least cost one, the issue of timing pre-
sents a second interesting implication for the development of an MCE method-
ology. By focusing attention on marginal costs, we may in fact, through the
use of interim standards, be adopting a regulatory program where total costs
are higher than they need be. Exhibit 9 illustrates this point. The two pos-
sibilities are to either require standard 1 and at a later date impose stan-
dard 2, or immediately require compliance with standard 2. The total costs
would be:
EXHIBIT 9
TOTAL COSTS FOR MEETING INTERIM STANDARDS
Option Treatment Chain Total Costs
Standard I, then 2 1 4,0
Standard 2, immediately 2 3.5
If EPA's main concern is to minimize the costs its regulations impose on in-
dustry, Exhibit 9 indicates that this can best be accomplished by imposing the
most stringent level of abatement attainable using an available treatment
chain. This analysis suggests that in situations where phasing-in more
stringent standards is being considered, EPA may want to consider both the
total and marginal cost implications of its actions.
E. Comparing MCE of Alternative Dates for Implementation.
The previous section examined the complications arising from timing issues
a3 they effect the numerator (i.e. , the costs of the MCE methodology) .
Although timing is straightforward as it relates to costs, it severely compli-
cates attempts to measure effectiveness. This problem arises when the MCE
-------
- 4
methodology is used to compare e. proposed standard which is to take effect
immediately to one to take effect at some future date. This policy question
most frequently arises when EPA consider:; requests seeking extensions of
abatement deadlines.
Comparing costs in this situation is relatively easy. Although there may
be some debate about the appropriate discount rate when applied to the esti-
mated future costs, these costs can be directly compared to present invest-
ments .
But no such clear-cut manipulation exists for comparing the effectiveness
of the same standards imposed at different points in time. If we were to ig-
nore this problem, the MCE of a standard imposed today would be exactly the
same as that of the same standard if its implementation were delayed for a
period of time. In Exhibit 10, the problem is shown in numerical terms.
EXHIBCT 10
MCE OF DELAYING IMPLEMENTATION
Year
1
Annua 1
Costs
100
100
Annua 1
Pollutant
Removed
50
50
Number of
Years in
Effect
30
25
PV of Total
Costs
1,037
998
Discounted MCE
Effectiveness* Ratio
50x30 = 50 20.74
30
50x25 = 41.67 23.95
30
* Assumes a 10% discount rate.
In this example, we assume the choice is between requiring 50 kilos of
abatement new or delaying implementation for four years. In either case, the
annualized costs would be 100. We first calculate the costs by determining
the present value of the annualized costs over a 30-year timeframe. To deter-
mine effectiveness, we would divide the total amount of pollutant removed for
each option over 30 years and divide this by the 30-year timeframe. The MCE
ratio for these options suggests immediate implementation (20.74 compared to
23.94) would be the best policy for EPA to pursue.
STEP 7: AGGREGATING ENTITIES
Marginal cost-effectiveness analysis, although most applicable at the
entity level, can be conceptua1ly extended to apply to industry segments and
industries. Further, multiple industry analyses within regions or nationally
might be conducted though the data base requirements are potentially massive.
Nevertheless, many macroeconmic policy questions of interest, to EPA deci-
sionsraakers cannot be appropriately assessed unless model plant data regarding
marginal cost-effectiveness are suitably aggregated.
-------
- 48-
Policv Issues and Aggregation
EPA wants to answer several types of policy questions that could be accom-
plished with the proposed methodology, including the following:
• What is the marginal cost-effectiveness of cleaning up one pollutant
to the same degree in d iffereat industries?
• What is the marginal cost-effactiveness curve for cleaning up one
pollutant in all industries?
• What is the marginal cost-effectiveness of cleaning up all pollutants
in one region of the country?
• What is the marginal cost-effectiveness of cleaning up one pollutant
across all industries in one region?
Types of Aqgregation
" " - -* —* j i i ¦¦
To respond to these policy questions requires that model plant data be
aggregated along any of three dimensions:
Q pollutant
© industry
® geography
In addition to the data created in previous steps, identification of all
present and planned entities and segments in an industry by size, age and
1ocation is also necessary. Most of this data is readily available from the
Commerce Department.
With the above information available, it is possible to aggregate along
any of the three dimensions, i.e., pollutant, industry, and geography, we
discuss each of these further after describing them briefly here.
Pollutant. A relevant policy question is the extent to which different
industries must remove a particular pollutant at a specified marginal
cost. In general, when exploring this question it is necessary to 1)
identify all of the industries in which the pollutant of interest is pre-
sent, 2) construct and analyze a model plant or several plants (depending
on production processes In the industry), 3) identify the total number of
plants (of each type modeled) in each industry and 4) sum the abatement at
each specified marginal cost across all industries. Marginal cost-effec-
tiveness could be compared for several different pollutants to respond to
another policy question, i.e., "What are the most/least cost-effective
standards?"
Industry. Aggregation by industry is essential for EPA to know how one
industry's burden compares to others across all pollutants and for indivi-
dual pollutants. Furthermore, industry aggregation discloses those indus-
tries where the most effectiveness is achieved per dollar spent. In this
case the necessary elements are the model plants, the number of plants in
the industry, and the marginal cos--effectiveness curves.
-------
49
Geography. To understand the impact of pollution control on different -concm
and environmental regions, aggregation by geography is important. Geography
is important because of the varying levels of pollution found in the waters
•and atmosphere in different c eg ions of the country. The ele- meats necessary
to perform this kind of analysis are: 1) model entities by relevant segment
foi each industry, 2) the number of entities by segment in the region and 5}
the marginal cost-effectiveness curves for the entities.
As explained i n Chapter 2, aggregation is eguiva lent to j^umm^ng the appl i -
cable entity-level MCE curves over a constant range of marginal costs. This
results in an expanded effectiveness range {horizontal, axis) for the given
pollutants being aggregated. Such an expansion procedure is not unlike an
aggregate supply curve of abatement for the specified aggregation case, e.g.,
segment., industry,, region. However, all applicable entities must be included
and weighted by the number of equivalent model entities to accomplish the
aggregation.
In actuality, two main problems will typically exist even when MCE analy-
sis as been performed for all applicable entities to be aggregated. First,
the range of marginal costs estimated for each entity may not be common among
all entities because their data bases are different. In effect, tn» MCE curve
for each entity should be extrapolated (potentially for both higher and lower
marginal costs) to a common marginal cost range, e.g., from the lowest to the
highest observed MCE1s among the entities being aggregated.
Second, aggregation of MCE curves of different pollutants — which is pos-
sible only with the introduction of effectiveness weights -- piesent.s the pro-
blem of judgmentally obtaining such weights that are exogenous to this analy-
sis,. Furthermore, unless the MCE ratios of the separate pollutant are compar-
able, MCE extrapolation problems as presented above may be compounded. Al-
though, alternatively, and appropriately, the standard physical units of each
pollutant should be scaled in such a manner that the mean MCE ratios analyzed
are normalized. That Is, the different, pollutants * observed MCE ratios should
first he made to have comparable ranges of marginal cost;, by scaling the phy-
sical units of abatement so that standard units have similar marginal costs,
e.g., if MCE, = $10 for pollutants at its means abatement of 100 units (kg),
and MCEj = §20 for pollutants at its mean abatement of 40 units (g), then
MGE^ = m::e2 = 10 if the standard unit Cor pollutant 2 is changed to be
H/2g) rather than (g) .
Following this conversion, effectiveness weights per standard units must then
be applied to create a comparable horizontal axis. Returning to the
extrapolation problem as described above, certain theoretical principles
should be followed when extrapolations are made (often judgemental- 1 y r it.her
than statistically because of limited! observations). Namely, MCE curves
theoretically are high as abatement approaches 100 percent, e.g., asyra- totic
to the line extended vertically upward from 100 percent abatement foe each
entity. Also, as abatements are lowered, the MCE curve theoretically has a
non-zero minimum point {due to fixed costs) . Hence, extrapolat ions "down
ward" should not approach zero marginal costs as might b*1 implied by limited
-------
- 50-
data. Rnooqnlzing these principles for each entity's MCE curves over the
relevant MC cange is critical when aggregations are to be made.
~ * »
The folLowing two chapters present applications of this methodology to the
textile and coal-fired power plant industries.
-------
- 51 -
4. TEXTILE INDUSTRY: A CASE STUDY
This application of marginal cost ef f ecti veness to the textile industry t o-
'¦uses on water. The data used in the examples were derived from the recent
Effluent Guidelines Document entitled, Technical Study Report BATEA-NSPS-PESE-
PSNS, Textile Mills1 Point Source Category!- and' associatedunf eoorted data from
the study. This industry was chosen, in part, due to the large amount of data
available. However, as will be subsequently noted, even the large amount of
available data was found to be limited for a rigorous MCE analysis, in terms of
logical alternatives generated and accompanying engineer ing descriptions and
explanations of the unit processes and technologies involved.
The serious limitations, have restricted the pilot, study and we cautmn th>-
reader that the MCK applications and results should be used more as illustrative
of the methodological and problems encountered rather than indicative of conclu-
sions regarding abatement levels,
I. IDENTIFICATION ANALYSIS
The Clean Water Act^ requires existing industrial dischargers to waters to
achieve effluent limitations requiring the application of the best practicable
control technology currently available (3PT) by July I, 1977 . By July. 1, 1984,
these same dischargers are required to achieve effluent limiations requiring i hf
application of the best available techno logy economically achieveable (BATEA)
and the best conventional pollutant control :echnology (BCT). Additionally, new
industrial dischargers are required to comply with New Source Performance Stan-
dards (USPS) under Section 306 of the Clean Water Act (the Act) , and new and
existing industrial dischargers to Publicly Owned Treatment Works CPOTW's) are
subject to Pretreatment Standards (PSES for existing sources and PSNS for new
sources).
The textile industry3 1S composed of over 1,100 individual textile mil is
engaged in manufacturing processes which, in one form or another, generate
wastewaters and are thus subject to these abatement regulations. Approximately
80 percent of these mills discharge into publically owned treatment works (POTW)
and are classified as Indirect dischargers; the remaining 20 percent discharge
directly into receiving waters and are classified as direct dischargers. Al-
though indirect dischargers will be subject to certain pretreatment standards,
the segment of interest for this analysis is existing direct dischargers,
^Sverdrup and Parcel and Associates, Inc., Technical Study Report
BATKA-NSPS-PSES-PSNSf "Textile Mills Point Source Category", prepared undei
contract Nos. 68-01-32 89 and 68-01-3884, Nov 1978.
2The Federal Water Pollution Control Act Amendments of 1972, as amended
by P. L. 95-217, the Clean Water Act of 1377,
^The textile industry consists of establishments which typically
create and/or process textile related materials for further processing into
apparel, home furnishings, or industrial goods.
-------
, - 52 -
2_._ DEFINE ENTITY
Direct, dischargees in the textile industry are comprised of a diverse group
of establishments varying in size, process, and product. The general character-
istics of the industry establishments range from small family-owned mills utili-
zing traditional manufacturing and.managerial practices to large multi-mill cor-
porations who rely on the latest managerial practices and sophisticated proces-
ses available.
The most common structural depiction of the industry is the standard indus-
trial classification {SIC} system used by the U.S. Bureau of the Census. How-
ever, for purposes of an analysis of effluent controls, a classification system
based on manufacturing or process functions performed at the facility is more
appropriate. This is because wastewater characteristics are predominately de-
pendent upon the process functions performed at a facility. Such a classifica-
tion industry establishment to be grouped into categories with similar waste-
water characteristics.
Based on extensive industry analysis, in which SIC, functional, and other
characteristics were used, -- direct dischargers can be defined according to the
entities shown in Exhibit 11. These entity descr iptions, in this instance, were
adopted from the Effluent Guideline Document^ an<} the associated economic im-
pact analysis.'' in many instances previous studies will provide entity defi-
nitions, although they should be evaluated with regard to their representative-
ness .
Twenty-six entities were defined to represent this industry. The number of
entities in practice will depend upon the extent of variation in the population
of interest, availability of data, time and budget constraints. Generally, the
larger the number of model plants, the greater the accuracy of the analysis.
In addition to the entity descriptions, the number of plants and/or other
weight measures should be reported as shown in Exhibit 11-
3. ESTABLISH DATA BASE
3.1 I dentify Key Pollutants
As previously noted the data base for the analysis was taken from a recent
Effluent Guideline Document. A large number of pollutant parameters are found
in the textile industry. Out of the large number of parameter and pollutant
constituents, seven were selected and reported. These include conventionals:
biochemical oxygen demand (BOD), chemical oxygen demand (COD); total suspended
solids (TSS); oil and grease (0 & G); two nonconventional pollutants - total
phenols and sulfide; and one priority pollutant - chromium, which will be in-
cluded with the nonconventional group as a natter of ease.
^Sverdrup and Parcels, op. cit,
^Development Planning and Research Associates, Inc., Economic Impact
Analysis of Proposed Effluent Limitations Guidelines For the Textile Industry,
EPA Contract No. 68-01-4632, in preparation.
-------
No. Segment
W0ol Scourir.g
Woo\ Finishing
Woven Fabric Finishing
Woven Fabric Finishing
Woven Fabric Finishing
Knit Fabric Finishing
Knit Fabric Finishing
Hosiery Finishing
Carpet Finishing
Stock and Yarn Finishing
25, SonwovM[i Fabric
Ft aish ing
26. Nonwrivon Fabric
F in i i:i£
Proc es s
exhibit ; i
DEFINITION OF MODEL PLANTS FOR DIRECT DISCHARGERS. TEXTILE INDUSTRY
Site
Code
Simple
Complex
WF
WFFS
WFFC
Complex plus
designing WFFCD
S im p 1 p
Complex
W.ivor
Fe 11
KFFS
KFFS
KF
CP
SYF
NFtU'
NF F F
Si ze
Sma 11
Modi um
I. arp?
Smal 1
Med i un
Large
Sraa 11
Smal 1
Metli-.ia
Large
Small
Meditrai
Smal 1
Staa ! I
Med iire
Sma 1 i
Mediam
Stoii 11
Hed ium
Larg«
i. I
Med i um
Lar ge
X-L.tt 7
10
20
20
2.0
• E *irnar-1'! f r -*1 tv - & *.• d i 1
-------
- 54 -
This analysis was based available data, Howeve*:, in other studies, it will
be beneficial to carry out detailed technical analyses of waste character istics
and relationships.
3.2 identify Unit Process and Treatment Chains
The next step is to determine the range of unit processes which will provide
abatement. The unit processes reported in the Effluent Guideline Document on
Textiles, were used herein aa shown in Exhibit 12. This Exhibit also presents
the treatment chains, that is the basic cornbinat ions of an it processes. Avail-
able alternatives may or may not be sufficient for rigorous MCE analysis, depen-
ding upon the number and the logic of the resulting abatement levels.
In this application, an analysis of the unit processes and resulting treat-
ment. chains war, not performed, as it was outside the purpose of the effort.
However, as will be discussed, the treatment chains and resulting abatement
levels were limited, indicating the need for presenting additional unit proces-
ses or treatment chains.
Both water and sludge were included in the data base; sludge characteristics
were not separately reported. Although it will depend upon the nature of the
policy question, the analyst should include all media and pollutants that will
be affected by the policies being evaluated.
3.3 Relationship Among Unit Processes and Pollutants.
A critical factor in this step is to identify the unit process treatment
cha in-pnl Lutant relationships, starting at the unit process level. The textile
data were deficient in this regard; which limited the ability to create addi-
tional treatment: chains and to identify unit processes with specific pollutant
parameters. Additionally, we noted that the sequence of treatment chains may
influence abatement levels. For example in this textile case we found by deduc-
t ive analysis, that treatment chain C — chemical, coagulation (1), sedimentation
(2) and mu11 i-med t a filtration {3} — resulted in a different level of abate-
ment, depending on whether it was compared against multimedia filtration or
against chemical coagulation. Other words unit processes may not be strictly
additive. Nonadd i t i ve relationships should be explained and reflected in speci-
fying the associated abatemeit levels.
3.4 Cost Estimates
The other critical data component is cost estimates. In this case, total
annual costs were used as reported.3 investment costs were repotted by unit
processes, but operating and maintenance costs were only reported by element
within a treatment chain. Consequently, the ability to assign costs to unit
processes was severely hampered. Both the unit process pollutant abatement and
unit process cost relationships are necessary for rigorous MCE analysis.
^Annua 1 capital costs were estimated in the reported data as interest on
total investment plus depreciation. Because of limitations of the data base,
reestimati ng annual capital costs by the discounted cash flow method was not
considered to be warranted.
-------
EXHIBIT 12
ALTERNATIVE END-OF-PIPE TREATMENT TECHNOLOGIES
EXISTING SOURCES^
No. No;;. Treatment Chain
B 1,2 Chemical coagulation (1) and sedimentation (2)
C 3 Multi-media filtration (3)
0 1,2,3 Chemical coagulation (1),sedimentation (2), and
multi-media filtraiion (3)
E 3 4 Multi-media filtration {3} and granular activated carbon
(4)
F 1,2,3,4 Chemical coagulation (1), sediment at ion (2), mu 11 L-m<-dia
filtration (3), and granular activated carbon (4)
G 5 Ozonation (5)
H 1, 1', 5 Chemical coagulation (2), sedimentation (2) , and
ozonation (5)
J 3,5 Multi-media filtration (3) and ozonation (5j
1 1,2,3,5 Chemical coagulation (1), sedimentation (2), multi-media
filtration (3), ard ozonation <5)
M7 1,6 Chemica 1 coagulation (1) and dissolved air flotation >6}
N 1,6,3,4 Chemical coagulation (1), dissolved air dotation (6),
multi-media filtration (3), and granular activated
carbon (4|
P 1,6,5 Chemical coagulation (1), dissolved air flotation (6),
and ozonation (5)
feBPT consisting of screening, extended aeration activated sludge,
sedimentation and solids recycle to creation basin assumed to be in place.
7 A! ternat. ives M, N, and P apply to wo 11 scouting.
-------
5b -
4. CREATE POLLUTION CONTROL COST AND ABATEMENT
As noted in the preceed1ng discussion, it is important to create the neces-
sary data base. The limited data available in this allows only those combina-
tions reported to be used. Had additional unit process pollutant information
been developed or available, additional treatment chains might have been devel-
oped. Due to the large number of entities involved in this population of exis-
ting direct dischargers, a table for only one entity — a medium sized complex
plus desizing woven fabric finishing mill (No 12 in Exhibit 1) is presented as
Exhibit 13, Normally, Tables would have to be prepared for each entity.
The lower portion of the table was ordered from the reported data set as
shown in the top nine rows of the Table. Though analysis we ordered each set of
treatment chains such that the abatement of each pollutant in a poliutant vector
remained equal or increased and that total annual costs were increasing, pursu-
ant to the principles of marginal analysis. This produced the nine unique sets
of logical paths.
The indicated cost separation was done by examining the marginal effective-
ness of each treatment path vector. From this analysis we could separate cer-
tain costs to conventional pollutants and in five instances, directly to COD.
Had we additional technical information, additional cost separation might have
been. We suspect that a considerable degree of incidental removal is in fact
occurring with some unit processes. If this is true, we would recommend that no
costs be assigned to the incidentally impacted pollutants.^ Hence there may
be a greater degree of separation than shown in Exhibit 13.
5. PERFORM INITIAL ANALYSIS
The creation of the data table resulted in several readily observations,
particularly with regard to the paucity of separable costs and the relatively
few and large effectiveness intervals. This suggests that joint cost allocation
and aggregate are significant issues in water. This is in contrast to the air
media presented in Chapter 5, which had few joint costs.
With the paucity of separable costs and technical background regarding inci-
dental removals (that is, unit process-treatment chain-pollutant relatlonshlps),
cost allocations was limited to target pollutant group or equally among impacted
pollutant groups. Because of these 1 imitations,10 the answers to policy ques-
tions regarding specific pollutants or pollutant parameters would of limited
value and were not addressed in this illustration.
This illustration does present a situation that may typical when existing,
available data are being used for MCE analysis. The following discussion will
provide possible short cut.
For each of the ordered sets of d; ta, the incremental effectiveness and cost
tables were computed for each ordered data set. Then, the MCE ratios were com-
puted for each i ncremental, by dividinc the incremental cost by the cor respond ing
9This should not be considered as a hard and fast rule, particularly for
priority pollutants.
1°In practice, some additional but limited, techlrtcal background work
might prove of value and should be explored.
-------
EXHIBIT ij. TEXTILE COST/ABATEMENT TABLE'
fiU* •
fTT3
oosvUTlQii&L
m __ c v ,
&kt,£®d tiklt t*i v?i t A*«t«4 "ft* ^ r t«
fJT3oor r^fcJTjT™"iii."o«\ ""rSii/*5" '
«
Z5 ft
i?,t
IN.1
2»,4
18.f
25.6
a.c
:* =>
1.
%,'i
8.4
02.*
.
M.I
i?.:
6 ¦,
14 1
8. .4
105*
tat, 4
&§.)
i >.:
R .
•
3®.*
4.
362.1
85.5
M.3
17.2
n.4
0,i|
f
3*. 5
M?.l
59.?
6«.1
i »,?
O.-i
-
0.0
42.
2M.6
UB.i
0.0
85. ^
0.^5
iS.i.
25,5
1?.:
3i4.&
iU.o
M.9
?V6
a.o
1% .«*
f
jfci
8.4
M2.3
US.4
6«, J
6,4
6.9
¦J.-
*
>*,)
8-4
JI5.1
132.ft
8,4
6,9
B.4
,..
1
34. 1
1,4
102. ft
m,i
46.3
U,I
6.9
8 ..
n
*,l J
R J
# .4
205.3
311,4
IM
M.?
1?.?
6.9
a 4
9
38 5
4.1
*7.1
*,9
««.3
i?,J
o «,
„¦
1
)4. \
8.4
102,&
135.1
M.3
I?.,
s ^
C
TS 5
4.
142.3
§§.3
61.3
17?
15.4
o >>
f
1.4 .1 .1
J« *
*
$5? «
383
6®. J
17.2
n.4
o n
1
1./
25.4
17.i
im. s
^S, 9
25-6
0 0
0
I.l.i
14,3
i.*
i@§.s
m.4
«,5
11.2
6 ?
8.4
r
1.7,1.4
3a. 5
4.?
147.1
51,f
M.3
17.2
15,*
0 o
c
*
0.0
41.
m *
IMA
0.0
•5.^
0 = 0
11,*
j
V?
34 »
1.4
*2.1
145.4
66,3
1.4
6 9
§,<*
t
VI.1.1
14aJ
§.-
515.1
102.«
IM
iS.I
8.4
4.*
1.4
c
51
o.o
42. T
*».*
i«i.i
0,&
0.0
n.4
n
3,1 ,2
5
n.j
lit.4
Hi,4
15.6
0,®
15 4
*
5*ir3.l
14 - 3
a.4
323, 1
102.4
M.l
8.4
6,1
8.4
c
)
*4. *
3 -
102.4
m.i
3
I 7 ,2
b.f
8 .
0
Kl,l
TAl I
8 4
205.3
222.4
3S@
a.s
17.3
«l «
1.4
*
1,2.5
14.}
8.4
325,1
102.6
4f. 1
*.«
(S,9
« '*
»
!« 1
:n €
17 .
IM.i
IW.t
25,6
O.O
n .
0
*. . J. 1
14, J
205.1
122.4
44.3
i?.2
a .
1
i. 7, ». %
34. i
8-4
SS5.3
502.5
4S.1
8.4
6 #
g •
a
1.„ J
25.6
lf,l
IW.I
2H.6
W-*
2S 6
0.0
*«. s
;>¦
14.. 1
8.4
H>5.1
122.4
68 3
H.J
6. "~
A u
<
3i. 1
ft.*
121-1
102.6
68.3
4 .4
». ¦t
&
2 , ?
.'^ K »t
; ?.
i*a. t
2**,€
58.?
"n - 4
0 0
I4. .
H
1 .2
;5 ,6
17./
U6.4
111.4
li.t
J1-S
0.0
: *> «
K
34. 1
«>
123. 1
1^2.6
6A.3
6 6
6.9
a u
i
H, I
1.4
192.4
IIS. i
44.3
17,3
* 9
? .»
8
i, i
J
84.<«
2§2. 1
III.4
&S.3
U.I
4 1
8.4
K
1.5.1 .'
14. J
1.4
*».!
i02.&
it*
6i.5
£ i.
6,t
8 «
. SSii *t«2f-"2tr|5" Co»c gKt^^SiSI""gj Si^^lti5~*CM"t"
FJk<7»T T»i75Bt iEk»/fi 111 ,mo i ftaj/yt Ti iTSBo) iSki/J' 'I iTSW:
' i«
j
'as *¦«>
_
? >4.;
«46-
-
• ne,. -I
Al
S' '
*
'M +
>,00
i
lH-9 «
n
. 7Q. •»
16.ft
Vk*. i
114
£
54 . 7
.
,
1
<»&,
*
S0t>. «
.0
ts 1
2? >
4
>•<}',»
Ij4
n
%fO
0
A' >
*
m. *
1
«
! 70.0
4 U
4
: >6.6
? W
%
313. .*
400
*
IH.fe
434,
*>
1I6.&
;
?'£.
~N6 .'
¦
" W *•
^ * j
b
'^1 ' H
V*
«
,'S A. 2
i.
i *.
,
.'9' «
H'.
ft
'%4
* i -
' 1»F '
.W ^ "
l*.'1 »
u !*.
1 lb *,
n« -
114
<
.'54.
3*
*»
; H •>
S0?4
4k u
—*aat*J.cg!i
lev tfasmpm
0 0
-0«»0
0.0
.051
0.0 Ki
0.0
2*01
74 1
n
>41
0,0
0.0
.051
o.o i.f
c.o
2.01
• 4'-
U6
0 c
.SftO
0.0
051
0,0 l.t
f .0
2.0 t
13-
134
5-?
.025
.026
0.0 l.t
.08
l.fj
^3n
>30
OM
,0-i|
an
.026
0 0 1.9
.03
l.fl
' 7
-0W3
0.0
0.0
MM
IA .3
-M
.35
!0h
\oa
060
o.o
0,0
.05 >
1.6 .3
i .66
.35
SOS
.040
0.0
0.0
,051
1.6 .3
urn
.33
40 '
'.C "
.040
0.0
0.0
.051
1.6 .1
i.66
.15
¦>9*.
".9 -
U, 0
.040
5.0
.051
0.0 l.f
0.0
2.01
0
i 4^>
0,0
.060
0.0
.051
0.0 l.t
3.0
1,01
14^
in
.om
.25
'>76
0.0 l.f
.03
i.91
n
0
' ¦:¦;
' i .r
o.a
.060
0.0
.051
o.o i.i
0.0
2.01
rt
14*
S44-
- 0\1
.0^
.QiS
026
0.0 J .9
.®§
I.*3
0
w
» *.
5 MJ
0\7
.00®
.0 25
o:6
0,0 1,9
.0€
J.fl
IS.'
¦?
n ?
O.d
0-0
,051
0,0 1.9
0.0
2*01
Q
24!
24 1
o.n
.060
0.0
.011
0.0 1.9
0,0
2.01
0
354
11*
.4**:
9«
.015
.^16
O.Q 1.9
,
0.0
0*3
1 ,4 .5
t .66
,33
0
0
•.
0.0 l.f
o.o
2.01
k«A
[4^
'
'i...
iW
0 0
0.0
.0*1
1.6 .3
I.M
.15
0
»«
%«»•-
,04*
ll.u
jJM
n.O 1.9
0.0
2.01
n
24 '
a«io
a. «j
n o
i
t .* ,3
1.44
.15
'¦¦fc
Vi*
u.->
fl.O
051
1.6 .3
i .66
,55
a
1* •
< i. 11
Ml
0.0
.cni
u ,0 1,9
0.0
2.0i
;
,
6 '
n.fl
0.0
-051
1,6 „3
l. 6&
,35
o
¦5.0
0.0
0^1
1 .6 .}
i ,«6
,15
us
..,'T '
5 .o5o
2*3
334
if* tirfW for «m« •( »NHftt«l«n
-------
- 58
incremental effectiveness for each set of cost estimates (i.e., target groach
effectively provides an upper limit of the ratio, because any cost allocation
would serve to reduce the incremental costs and hence, the MCE ratio.
The MCE ratio resulting from these computations are summarized in Exhibit
15. Additionally, this Exhibit carries data information regarding total abate-
ment and the incremental impacts on nonconvent i ona1s as reference points for
ana lysis of th^ results.
Examination of the Exhibit indicates that the range of MCE ratios is $.64 to
57,11. However, the corqparabi 1 ity of the sets is inf>erfect because nonconven-
tionals are accounted for in different ' increments. Kocoqniz inq this 1 imi (ra-
ti on and in the absence of additional information, the MCE ratios can be used to
estimate a marginal cost curve. One approach would be a statistical fitting of
a function based on the data points shown in Exhibit 15. However, due to the
gross nature of the data, we believe that graphical analysis wjuid be moo* ap-
propriate as a first approximation with limited data so that extreme pointy
could be adjusted in or out depending on understanding of the control technolo-
gies and cost theory.
Exhibit 18 shows the plotting of the MCE ratios at this raid-pointl^ 0f j-he
respective intervals. For this estimate, all data points were plotted. Other
plots could be made using allocated costs, omitting those points involving al-
located costs or other adjustments as expertise permitted. An overall examina-
tion of this Exhibit indicates that the marginal cost for the conventionale in-
crease relatively rapidly. Precise MCE ratios would not be warranted, although
inferences could be drawn. For example, lI a threshold of 5.75 to $1.00 was
considered, abatement levels are 100 to 12S kkg's (about 20 pet cent.) per y<">ar
would appear to be indicated. Moving the threshold upward would indicate higher
abatement levels. Due to the limited data points and the nature of the varying
widths of the increments, inferences about abatement levels above 75 percent
would tie tenuous. If the threshold criteria were sui f ic i ent ly large, say above
$3.00, the generation of additional deta points might be warranted, if tin* af-
fected pollutants were considered important.
"If nonconventionals, were in fact the target, the increments would
have to be reordered i. £ treatment chain C were required (and not in place) to
achieve treatment chain D.
1 11 was also observed in the data that the control of oil and grease
appears to create abbe rat ions and technical analysis would be required to
analyze the ootnponents of the conventional group.
*'The mid-point is considered to be a better approximation than the end
point, as the end point will consistently underestimate the MCE ratios in the
rising portion of the marginal cost curve. The extent of the mid-point bias
will depend on the shape of the "real' marginal cost curves.
-------
EH IB If 14
Connputation of Marginal Co«t-Effeetivene«a iacieg for Convent tonals
Under Two Cost Allocation Procedures
ABATEMENT
AHWJ4L COST
MCE RATIOS
Treatment
Conventional
Non-Convent ion#]
Incremental
Target
Allocation
E<|ua
Allocation
Chain
Total
Incremental
Total
Incremental
Total
Total
Separable Non-Separable
Conventional
Son-Convent iona 1
(kkg/
yS
(Si,000)
7$/kkg/y)
C
212.1
212. 1
0.00
0.00
146
!4ft
146
.69
.69
E
46* .«
252,3
o,oa
0.00
no
)fu»
384
1.52
.J6
P
490,0
25.6
0.08
0.00
~ _
182
7.11
7.11
u»
vO
-------
EXHIBIT 15
A Summar-v of MCF Ratios (or Convent ionn! <
ABATEMENT
MCE RATIO
Ti p,irf
Incrcaent_
10h."0
26 3 ,4
40.' .4
Targer
. f> 9
I .83
:. is
Increments! Abatement of Non-Convent iona:s
Phi»n«il 5
0.0
0.0
0"i2
'"hfaffii -ifr.
0.0
o.o
O 1 c,
Sul firfe
~ 0.0
0,0
0.0
3t* .
0.1
0.0
.3 77
A,
5,
6.
21:. i
464,4
490.0
1 <(~>. 11
330.2
477 ,2
,¥i
1.12
a, ct
.055
r> .0
0,0
.0."i
•1.1
c.o
0.0
0.0
n, o
.077
0.0
7.
8.
9.
772.h
314.8
490.0
Hh.3
293.7
402,4
.89
!. 16
!, 16
0.0
0,0
,052
0.0
0.D
.01
0.0
0.0
0,0
0.0
0,0
,077
:rs
1 1
n
u
i ?>
l»3
1 ?
18
19
2u
21
; i
J4!
t-
239. r,
391 .8
434,6
239 A
400.
434. h
212.!
314.8
434. **
27 2 ,b
314,8
,.i...»
2 72. b
400.»
212,1
31-. 8
1 '.9,6
315. 7
*n,2
! 19 ,8
320.7
417.3
106.0
26 3.4
J 7 4.7
1 Ih. 3
2^3.7
3?.,"
: id , 3
33h.«
4 r,a
106.0
302,0
t>i, ;
! ,2<*
.65
4.39
1.29
1 .33
2 .hS
,fe<9
1.83
2.18
.39
2, !
.89
; .07
:. ¦;?
.69
I .45
,060
0.0
0.0
.060
0.0
0.0
0.0
0,0
.060
0.0
0.0
.Phi"
fl.il
.0b"<
0, c
. >jO
0.0
0.0
3.0
0.0
0.0
o.l
0,0
0.0
O.C'
0.0
0.i>
3,1'
n.o
0. u
1.0
i .6
0.0
0,0
1,6
0.0
0,0
0,0
0.0
S. 6
0.0
0.0
: .6
0.0
I .(J
3.0
0.0
1.6
n. *
1 .66
0.0
o.o
1.60
0.0
0,0
0.0
0.0
I .»
0. 1
O.n
. t-
0.0
1
o.o
' ,!¦*-
-------
EXHIBIT 10
MA F. Ill'A L COST-EFFECTIVENESS RATJDS FOR CON*VE^IO.\'ALS '• AS A TAPGET)
Hi MEDIUM WOVEN FABRIC FINISHING (COMPLEX PLUS DESIZI^s
Marginal
Cost
W 8. Or;
7. Oi'
6.0 C'
5.00
27
12
4
3.00
2. '
1.0(
.iO,13
f
4,16,25 * 7,19,2 2
3,20 • , • «33
io , J
23
*
28
io
26
14
*
11
100
lft
ion
300
4 on
500 5 71 kkg/year abatement.
100 Percent abatement ;above assume
existinq treatment )
-------
This same type of analysis could be done for the non-conventionals, although
in this instance, the data points are limited and our understanding of the tech-
nology was so limited that analysis was not presented.
fi. AGGREGATION
The development at an aggregate marginal cost estimate for the textile in-
dustry would first require an analysis of each of the 26 entities defined in
Exhibit. U- Huf to lack of the requisite technical und«*rstandinq and hence the
tonnusi nature of the results we chose not to present aggregation results for t:h<-
lndustry as a whole. However, to illustrate the approach the two model plant*?
Lor the wo von fabric finishing {complex plus df»s i z i nq) segments, are presented
be low.
Aggregation involves one additional requirement - namely, the establishment
of a spec i f i c curve or marginal cost effect iv«-ness and their horizontal surama -
t: ions. 1 * In practice, and particularly with a i united available data, the
analyst should est unate the aggregate MCE ntio as a range, nsing two carves
representing the entities. The results of the mode I plant analysis, estimate of
each segment's marginal cost effectiveness curve, the weights and the results
are shown in Exhibit 17, The weights, in this case, are the number of model
plants shown in Exh ibi t 11. Total abatement for each segment is the product of
the weight and model plant abatement; total abatement in the right columns is
the sun of the total abatement of segment,
Athough these results should be considered as illustrative, they were drawn
from the data sets previously discussed and the given target conventional target
approach. For a given marginal cost level the results show that each component
achieves a different degree of abatement. For example, at $1.00, the total seg-
ment shows <» 25 percent removal. However, the medium segment is at 29 percent,
and the small sized mi lis as at 26 percent, reflecting economies of size.^
It is noted that in preparing these results that tho abatement levels wero
assumed he linear {i.e., a constant scale factor). However, the reported annual
costs were found to behave differently among the treatment chains, indicating
different cost scale factors for different, components. This suggest that care*
should he given to the use of scale factors in scaling costs, even within a seg-
ment .
The relevant population of textile mills was defined as existing direct dis-
chargers. The population is composed of about 220 mills which are subject to
best available technology economically achievable, This population can be re-
presented by 26 entities composed of different processes and mill sizes.
The data base used was from recent Effluent Guidelines Document on the tex-
tile industry. Although a number of pollutants are present, seven key pollu-
tants and pollutant parameters were reported including:
1*Tho range of marginal costs used for aggregation were restricted to a
range including the most data points, thus only a portion of the curve, from
$5.00 to 54.00 was estimated.
14Based on existing influent levels and does not reflect existing levels
of abatement.
-------
EXHIBIT 1 ?
1
11 us t rat i r
n of Aggregat
ion a f Msrg
ins"; Cost-
Effect
i veness n f rnmvn t: cna 1
For Woven
Fabric Fin
i a h i n 6
SMALL
SEGMENT
MED Il!M
SEGMENT
' g i -i a i
Abated
Abated
7iri! Ahat
IlJSt
He i gh
L Pet
Mod 11
Tuta!
We i gh e
Tet
Mode! T o t j!
Pc
- IKttl!
V
~ 1 a k f v
y>
s:. do
i?
¦m
4S
7^5
10
?o
\ m
2, « 15
1. 5C
17
33
15
i.I75
10
a, ¦"> 0 »' s h0-
1,8^5
40
:. oo
11
>V-«
too
1, 700
:c
57
3?'
4,^0
52
50
1?
57
no
2,210
10
63
1*0 *\,h00
5,810
61
3.00
17
66
! In
2, 550
;o
M
} »)0 J, 400
6,450
67
3,50
17
72
165
2,805
10
73
415 4,150
6.9S5
73
i.,00
17
7?
: 75
2,925
10
76
435 0
7 , 325 •
76
LPer model plant data.
-------
- 64 -
Conventional BOD, COD, TSS, 0 k G
Nonconventiona115 Phenols, Chromium, Sulfide
The Textile Industry rase study demonstrated that marginal co:;t.-ef feet ive-
ness can ho developed but only after significant analytical effott not normal!/
included in the industry engineering studies. Model plant information can be
aggregated to industry or regional totals provided that the study includes geo-
graphic as well as size/type of plant parameters.
These positive results are offset by the following factors. The results
actually generated are useful only as an example application of the methodology
and cannot be used for policy decisions. Moreover, the additional analyse-, re-
quired is substantial and requires a degree of sophistication not normally need-
ed in engineering studies — the methodology relies on data that ma*/ be avail-
able during the engineering analysis but not generally required to achieve study
objectives. Finally,, MCE results of the type possible exclude such salient con-
siderations as economic impacts and are therefore only one tool of. several re-
quired for sound policy decisions.
Because of the data limitations the MCE analysis was confined to one of the
26 segments: "complex plus desizing" mills of the woven fabric finishing seg-
ment. Also, the analysis was confined to one of conventional pollutants. This
stemed from the lack of technical background information for detailed cost as-
s ignment ,
Within a relevant range of the available data, we found a marginal cost
curve for conventional pollutants for woven fabric finishing composed of 1?
small and 10 large plants. This components of thin curve are:
Marginal Abatement
Cost Amount Percent
~ (55 kkg/y
1.00 2415 25
1.50 3875 40
2.00 4950 52
2.50 5810 61
3.00 6450 67
3. 50 6955 73
4.00 7325 76
At a marginal cost of $1.50, the small mills could reach a 31 percent abate-
ment level and medium mills could remove 46 percent of the conventionaIs. This
illustrates the different amounts of abatement among a subsegment for a given
marginal cost.
The overall finding of this analysis was that, even though the source docu-
ment used for data points is one of the better ones we have reviewed, a large
amount ol additional background i n tor nation is required for carrying out. t he
detailed analysis as represented by MCE analysis,
—
The inclusion of chromium was done as a natter of convention and ease of
presentation of the analysis.
-------
— 63 -
5. HEW CO&t-FIRED POWER PLANTS: A CASE STUDY
In this case study, we apply the marginal cost effectiveness methodology to
evaluate alternative new source performance standards (NSPS) Eor coal-fired
power plants. This example was selected because it is a regulation currently
being evaluated by EPA. Significant resources have recently been expended to
develop data necessary to evaluate the alternative proposals under considera-
tion. We initially believed that this data would be adequate? to mof-t the need:;
of MCR analyses. Although the costs and abatement estimates developed for use
by EPA in setting NSPS were a useful starting point for developing this methodo-
logy, we found that extensive data gaps remain.
Because of these data Limitations, we caution the reader that the applica-
tions of the MCE methodology presented m this case study are only for illustra-
tive purpose:: and should not be interpreted as meaningful analysis of NSPS for
coal-fired power plants.
]. PERFORM POPULATION ANALYSIS
This case study shows how the MCE methodology could be used to evaluate
alternative new source performance standards currently being proposed for coal-
fired power plants. The population must be defined to include those coal-fired
power plants likely to be built if specific environmental regulations are adop-
ted, Because we are only addressing issues involving the proposed standards for
coal-fired power plants, we can limit this analysis to this subset of: total
"lectnc generating facilities. The types: of policy questions likely to arise
includei
® What is the MCE of alternative, mors; stringent sulfur dioxide
standards for new coal-fired power plants?
® What i'=: the MCE of t rad ing-off particulate control tor sulfur
dioxide control in new coal-fired power plants?
Each of these questions can be addressed using the information derived from
the population definod as new coal-fired power plants.
2. PERFORM, ENTITY ANALYSIS
Defining the entity is also straightforward in this case study. Because we
are dealing with a new source performance standard, neither age nor varying en-
gineering processes are relevant considerations. We assume that all new facili-
ties employ the same boiler processes. The one exception to this would he the
type of coal fhigh or low-sulfur content) used as an input. This can be expec-
ted to vary by region; low-sulfur coal is predominantly mined and used m the
West; eastern coal typically has a high-sulfur content. There are two possible
ways to handle this difference. We could define two distinct entities — one
located in the West and burning low-sulfur coal, and the other in the East using
high-sulfur coal perform MCE analysis on each, and then aggregate to the
total industry. Alternatively, we could define a single entity and incorporate
-------
- 66 -
variations in sulfur content of coal as an alternative unit process. The method
of analysis that is adopted depends on the specific policy question being
addressed.
For the purposes of this case study* we have defined the entity to be a 500-
mogawatr power plant. This facility was selected because it is considered to ho
the optimum size to achieve economies of scale. Most new facilities will be
comprised of multiples of S00-MW units. Additionally, a single 500-MW entity is
all that was needed Ln the analysis because it is possible using a factor ->t .7
to scale the costs of scrubbers {used to control sulfur dioxide emissions) for
different size facilities,!
3. ESTABLISH DATA BASE
i. 1 : lit^nt s i'y Key Pol lutants
The Development Documents for coal—fired electric utilities and the numer-
ous studies supporting the development docunents have identified over 50 pollu-
tants which are emitted by electric utilities.2 These fifty pollutants i n-
olude conventional, nonconventicnal and priority water pollutants plus criteria
and hazardous air pollutants. Additionally, slodge is created in substantial
quantities. Of these fifty, the most serious are:
Air:
so2
Flyash (ISP)
Wa t t 'f :
Suspended and Dissolved Solids
Heat
PH
ChLorine
Oil and Grease
Trace Metals
Sludge
Two important technical relationships exist among these key pollutants which
is independent of the unit process employed to control then. One is the air/
sludge relationship. To eliminate air pollution by any process in a controlled
plant requires that sludge be created. The second is that suspended and dis-
solved solids as defined contain chlorine, oil and grease, and trace metals.
^This figure was derived from Battelie's analysis of the scrubber costs
for different size power plants. It is inadequate for MCE analysis though it
is useful ;n other t nst a nee s.
2For water pollution, see EPA, "Development Document for Ktfluent
Lim i f-.-it i ims Guidelines and New Source Performance Standards for Steam
Electric Power Generating Point Source Category™ October 1974.
-------
- 67 -
3.2: Identify Unit Processes
The second substep in creating the data base is to identify all available
unit processes which will provide control of one or more of the key pollutant.1]
In this example, several unit processes a:e available for controlling SO2
emissions, two unit processes control particulates, and one unit process is
available Cor N0X. Sludge is chemically treated and landfllLed. " Because the
data was insufficient, water pollution un t processes have not been analyzed.
Control: The most well-known uni : process to control SO2 is a flue-
gas desulfur izer (FGD), also known as a scrubber. Scrubbers use any of a var 1
ety of materials including lime, limestone and magnesium oxide to absorb the
S02. Tn this analysis we limit the discussion to lime scrubbers, which to
date have proven to be the most reliable at the lowest cost. Scrubbers can be
designed to a control, to any of a variety of abatement levels. In the model
plant, the scrubber could be correlated with the tons of SOj abated by the
equation.
Annualized cost = 24,000 ,000 +¦ 166 :c tons abated.
As mentioned in the methodology section, this function could be combined
with Functions for other unit processes, if all unit process functions had con
tinuous characteristics. However, one of the reasons for this equation so
closely matching actual data is that the tons abated are all close together.
{The 1.2, 90%, and .5 standards represent the range from 80 to 98% removal of
SO2.) Additionally continuous functions are not available for other unit pro-
cesses,
A second unit process to control SO2 emissions is physical coal cleaning
(PCC). Physical coal cleaning removes significant amounts of SCH and ash be-
fore they enter the boiler. The removed residual is called coal tailings. In
the model plant, we used PCC with a scrubber because PCC alone could not meet
current standards.^ {We had no data on PCC applied to low—sulfur coal.)
The final unit process for SO2 control is lo*r-sulfur coal (LSC) . LSC is
defined as a unit process because it achieves a lower level of abatement when
compared to high sulfur coal. LSC can achieve dramatic levels of abatement be
cause it would emit annually only 30% as much sulfur as typical high-sulfur
coal. For example, uncontrolled emissions from a 500 MW plant burning low sul
fur coal total about 21,000 tons per year while uncontrolled emissions from a
plant burning high-sulfur coal would total 78,000 tons. Uncontrolled levels o
emissions using LSC would achieve a level of abatement, close to 1.2 lbs 10^
BTTJ. LSC must be used in tandem with a scrubber to meet the proposed standard
examined in this case study.
^An interesting analysis would be to compare PCC of low-sulfur coal
(LSC) with the current standards. It may be that PCC is impractical on LSC.
-------
- 68 -
Particulates: For particulates, two unit processes have been identified.
The most common is an electrostatic precipitator (ESP). Like scrubbers, ESP:;
can be designed to control to almost any level desired. In our model plant,
all the ESPs were designed to remove more than 99% of the uncontrolled emis-
sions .
The second unit process which controls particulates is a fabric filter.
This unit process can achieve very low emission levels provided that the chem-
ical composition of the flyash is wel . matched to the cloth used.
Nitrogen Oxide: The only unit process available to control Nox is two
stage combust 1 on. This process has no cost associated with it in our example
because:
• The n populat ion" was new coa' fired power plants
• An EPA report claimed that the costs of two-stage combustion were
practically identical to the costs of conventional combustion
techniques when designed into new plants.'3
S judge: Several unit processes exist to dispose of sludge. These unit
processes are:
® pond i ng
• chemical treatment and landfill
® mine disposal
® ocean disposal
« conversion to gypsum
© conversion to sulfuric acid or sulfur
9 use as synthetic aggregate
The current costs of these unit processes reduce the practical options to
ponding and chemical treatment and landfill (CT&L).5 Because not all power
plants would be able to pond, we only examine chemical treatment and landfill
in our mode 1 plant. The cost of chemical treatment and landfill is different
for flyash than for SO^. Flyash sludge costs about 53.50 per ton; sulfur
sludge costs 512.10 per ton.6 The abatement level reached with this unit
process is difficult to quantify; however, no untreated sludge leaves the
f acil ity.
^See "Electric Utility Steam Generating Units: Background Information
for Proposed NOx Emission Standards," EPA 450/2-78-006a, pages 8-22,
^Aerospace Corporation, "Control], ng S02 Emissions from Coal-Fired
Steam Electric Generators: Solid Waste Impact," Vol, II, Technical
Discussion, pages 103-1.08.
-------
-* f) J —
Water: Because data was inadequate for MCE analysis, as compared to the
partial inadequacy of air and sludge data, we were unable to incorporate water
unit processes into this analysis.
1-3 Relationships Among Unit Processes
In the model plant, the most important relationship which exists among unit
processes is between low-sulfur coal and electrostatic precipitators (ESP). It
is much more difficult for an ESP to remove particulates from a stream of emis-
sions from a plant burning low-sulfur coal.
3•4 Relationships Among Unit Processes and Pollutants
Exhibit 18 lists the alternatives examined in out case study. Several com-
plexities of the table should be noted.
first, it should be noted that most a :r pollution control processes can be
designed to achieve more than one level of abatement. For example, FGD can
abate SO2 to any of the three levels of control examined in this case study.
In fact, there probably should be a continuous cost function for FGD use, but
the limited data available restricts this analysis to the three discrete points.
Secondly, as is more frequently the case with water,, it is possible in control-
ling SO2 to use two or more unit processes together to enhance the level of
abatement. In this example, physical coal cleaning serves to augment FGD to
reach more stringent levels of control. Likewise, FGDs and low-sulfur coal may
be used in tandem. Another important complexity is the relationship between
tons of flyash and SO2 removed, and sludge disposal costs. Generally, if the
choice of coal is known a linear tnult ivar iable function couid describe the rela-
tionship. (Sludge cost = flyash removed x A^ plus SO2 removed x A2.)
When physical coal cleaning is used, the relationship does not hold because coal
tailings can be removed less expensively, and in some instances, are removed
from the coal at the mine site and not at the power plant. In our model plant,
we assumed that the plant paid the costs of disposing of the coal tailings.
4. CREATE KEY POLLUTANT UNIT PROCESS COMBINATION TABLE
Before determining the entries in each row of the table it is first neces-
sary to estimate the relevant capital and operating and maintenance costs for
each of the treatment chains. Exhibit 19 presents this data in aggregate form.
In developing the data for the case study we were limited to existing, readily
accessible cost estimates which in many instances we believe are inaccurate. In
part, this is because for several sources of data it is impossible to determine
what cost are included as part "relevant costs" in these estimates.
Exhibit 20 is the Treatment System Table for our example. Because no joint
costs exist it does not have a nonseparable column. No water pollutant data is
shown because of the inadequacy of available data.
5. PERFORM INCREMENTAL ANALYSIS
After creating the Treatment Systems Table it is useful to review the table
for obvious relationships. Often this process will help the analyst understand
the pollution problems of the entity, and identify serious gaps in data. (Much
^Ibid. pages 118, 121.
-------
- 70 -
EXHIBIT 18
UNIT PROCESS/TREATMENT CHAIN TABLE
Unit Processes./
Treatment Chains Abbrevlat
FI Lie Gas Desulfur izer FGD
Description
Also known as a scrubber. Washes
SO
2 with an absorbent, usually
Lime or limestone. Removes to any
of a variety of levels. Does not
remove other pollutants.
Physical Coal Cleaning PCX!
Removes ash and sulfur from coal.
Often done by mines which charge
higher prices for cleaned coal,
but sometimes done by power plant
Allows for reduced (partial)
scrubbing.
Low Sulfur Coal LSC
Contains 20% as much sulfur as
typical coal, but lower heating
value requires that more be
burned. Also contains 1/18 the
ash of typical coal, but much
harder to remove the remaining
ash. Costs more than typical coal
today.
Flectro itatic PrecLpitators ESP
Fabric Filter FF
Removes ash from the stack
emissions to any of a variety of
levels. Relatively inexpensive
compared to scrubbers. Do not
affect other pollutants.
Removes ash to even lower levels
provided chemical composition
allows. Works best and at least
cost with low-sulfur coal; but
will work with high-sulfur.
Staged Combustion SC
Practically eliminates NOK.
Costs the same as other combustion
techniques except when retrofitted.
rhemteal treatment and
Landfi11
Removes acidity and toxicity from
s'ludge so that it can be land-
filled.
-------
Pollutant
Unit
Process (%
SOj
- 7 L -
EXHIBIT 19
'IXn'AL COST ESTIMATES OF POLLUTION CONTROL
UNIT PROCESSES
Coal Level of Capt t:t 1 O&K
Type Abatement Costs Costs
sulfur) (To6 BUT) (S/kw) (mi 1Is/kwh)
Annua 1ized
Costs_
(mi 11 ions)
FGD
irXjD
PtlD and
PCC
FGD
FGD
PGD
FGD and
PCC
FGD
3. 5
7.0
3.5
1.5
7,0
.8
7.0
.8
1. 2
1.2
1.2
901 removal
90% removal
901 removal
. 5
. 5
124.93
156
126.82
139.46
157.17
119.42
153.22
105.54
8.99
1 1. 68
14.38
9. 95
i;
,22
69
, 15
87
SU.15
542.97
$52.18
§36.69
544.91
$28.49
$69.26
$25.43
Par ticulates2
ESP
.8
.1
66.. 34
2.91
511-02
ESP
3.5
.1
28. 75
1. 34
5 S.!)b
ESP
.8
,05
74. 74
3.32
$12.56
ESP
3.5
. 05
32. 77
1.46
S 5.55
ESP
.8
.03
80. 71
3. 57
$13.59
ESP
3.5
.01
36.32
1. 59
S 6.02
FF
.8
.03
58.4 5
1,96
S 7.59
FF
3.5
.03
51. 8 3
1.7 2
$ 6.64
1 SO2 control costs derived from Pedeo Environmental "Par Ilculate and
I>u 1 fur Dioxide Emission Control Costa tor Large Coal-Fired Rollers" (1977) .
Annualized costs are in 1977 dollars, derived from capital and O&M costs which
are given in 1980 dollars using an annual inflation factor of 7.5%.
Annualized capital costs were straightline depreciated over 35 years.
"^Particulate control costs derived from Pedco, see footnote 1 above.
Low-sulfur coal is assumed to have an ash content of 81, high sulfur coal
14*. All cost assumptions are the ssarae a.-, above.
-------
Treat
AIR
SOx Partiru!aces
Abated Enitted Cost Treat, Abated' Emitted Coat
1.
FGD
61.6
17.1
33,15
ESP
135.2
1 ,4
5.06
2.
FCD
61 .8
17.1
33. 15
ESP
135.9
.7
5.55
3,
PCD
61.8
17. i
33. 15
ESP
136.2
.4
6.02
4.
FCD
61.8
17.1
33.15
FT
136.2
A
6.64
5.
PCD
61.8
17.1
33.15
ESP
135.2
1 ,4
5.06
6.
POT
61 .8
17,1
33.15
ESP
135,9
,7
5,55
?.
FGD
61.8
17,1
33.15
EST
136,2
,4
6,02
6.
PCD
61.8
17.1
33.15
FP
136.2
,4
6.64
9.
Fffl+PCC
71.8
7.1
52.18
ESP
135.2
1.4
5.06
10.
F»+PCC
71.8
7.1
52.18
ESP
135.9
.7
5.55
11.
FCD+PCC
71.8
7.1
52.18
ESP
136.2
.4
6,02
12.
FG5+P0C
71.8
7.1
52.18
FF
1 36. 2
,4
6.64
13.
POJ+POC
71,8
7.1
52.18
ESP
135.2
1.4
5.06
14.
FQ5+PCC
71.8
7,1
52.18
ESP
135.9
.7
5.55
15.
FCS+PCC
71.8
7.1
52.18
ESP
136.2
.4
6.02
16.
Fffl
70.9
7.9
36,69
FF
136.2
.4
6,64'
17.
FCD
70.9
7.9
36.69
ESP
135.2
1.4
5.06
18.
FGD
70.9
7.9
36.69
ESP
135.9
.7
5.55
! 9.
FCD
70.9
7.9
36.69
ESP
136.2
.4
6,02
20.
FCD
70.9
7.9
36.69
FF
136.2
.4
6.64
21.
F(31
70.9
7.9
36.69
ESP
135.2
1.4
5 .06
22.
PCS)
70.9
7.9
36,69
ESP
135.9
.4
5.55
23.
FGD
70.9
7.9
36.69
ESP
136.5
.4
6.02
24.
FCD
70.9
7.9
36.69
FF
136,2
.4
6.64
25.
FtS + LSC
19.2
2.9
28 .49
ESP
69.4
1.7
11.01
26.
F®+LBC
19.2
2.9
28.49
ESP
90.2
,9
12.55
27.
FCTS + LSC
19.2
2.9
28.49
ESP
90.6
.5
13.59
28.
Fffl+LSC
19,2
2.9
28 .49
FF
$0.6
.5
7.60
29.
FGD+LSC
19.2
2.9
28 ,49
ESS1
89.4
1,7
11.01
30.
P®+LSC
19.2
2.9
28 .49
ESP
90.2
.9
12.55
31.
FGD41SC
19.2
2.9
28,49
ESP
90.6
.5
13.59
32 .
FCD-cLSC
19.2
2.9
28.49
FF
90.6
.5
7.60
33.
FCD + LSC
14.2
7.1
25.43
ESP
89.4
1.7
11.01
34.
F®*LSC
14.2
7.1
25.43
ESP
90,2
,9
12.55
35.
FCD + IJ3C
14.2
7.1
25.43
ESP
90.6
.5
13. 59
36.
FSI+LSC
14.2
7.1
25.43
FF
90.6
,5
7.60
.17,
F® + IiSC
14 ,1
7.1
25.43
ESP
89.4
1.7
11 .01
38.
FGO+LSC
14.2
7.1
25.43
ESP
90.2
.9
12.55
39.
FOJ + LSC
14.2
7.1
25.43
ESP
90.6
.5
13. 59
40.
FCD +LSC
14.2
7,1
25.43
FF
90.6
.5
7.60
41.
PCD +PCC
150.7
7.1
69.26
ESP
135.2
1.4
5.06
42.
FCB+PCC
150.7
7 !
69.26
ESP
135,9
.9
5,55
43.
FCB*PCC
150.7
7,1
69,26
ESP
136,2
,5
6.02
44.
POD+PCC
150. 7
7. 1
69.26
FF
136.2
6.64
45.
FGD+PCC
i*5 0.
7.1
69.26
SSP
135.2
i .7
5.06
46.
pm*pcc
150. 7
7. 1
69.26
ESP
135. 9
¦ 9
5.55
47.
FCP+PCC
150.7
7.1
69 .26
ESP
136.2
,5
6,02
48.
FOT+PCC
150.7
7. 1
69,26
FF
136. 2
.4
6,64
K TONS/YR
$M
SC TONS/YR
SM
Insufficient data were available to develop entries for water.
oil and greas«, trace aet* 1», pH, and hear.
2, The onlw available ource argued tKat efficiency achieved when
9I91CH UJSi AMU AP« 1 f.ntll 1 I AOi-C.
CO&L-PIRED POWER PUNTS
MOx
Treat. Abated Emitted
o c
SLUDGE TOTAL1
Tech Abated Emitted Cost Cast
SC
.8
9
9
0
CF+L
223
0
223.0
1
7
39.91
SC
.8
9
9
0
CF+I,
223.D
223 .0
1
7
40,4 1
SC
.8
9
9
0
CF+t
223
,0
223.0
1
7
40.88
SC
,8
9
9
0
CP+t
223
0
223 .0
1
7
41 .50
SC
10.3
4
0
CF+L
223
.0
223,0
I
7
39.91
SC
10.3
4
0
CP-t-L
223
0
223 .0
1
7
40.4 1
SC
10.3
4
0
CF+t,
223
.0
223.0
I
7
40,88
SC
10.3
4
0
CF+L
223
0
223,0
1
7
41,50
SC
.8
9
9
0
CP+L
157
0
157.9
i
1
58,34
SC
.8
9
9
0
CP+L
157
0
157,9
1
1
58,6 5
SC
.8
9
9
0
CF+I.
157
0
157.9
1
1
58.83
SC
.8
9
9
0
CF+L
157
0
157.9
1
1
59 ,92
sc
10,3
4
0
GF+L
157
0
157.9
1
1
58.34
SC
10.3
4
0
CF+L
157
0
157.9
1
1
58.65
SC
10,3
4
0
CF+L
157
0
157,9
t
1
58.83
SC
10.3
4
0
CF+I.
157
0
157.9
1
1
55. <>2
SC
.8
9
9
0
CF+L
237
0
237.0
1
5
43.25
SC
.8
9
9
0
CF+L
237
0
237.0
1
5
43.74
SC
.8
9
9
0
CF+I.
237
0
237 .0
1
5
44.21
sc
.8
9
9
0
CF+L
237
0
237 .0
1
5
44.83
SC
10.]
4
0
CF+L
237
0
237 .0
1
5
43. 25
SC
10. 3
4
0
cf+l
237
0
237 .0
1
5
43.74
SC
10.3
4
0
CF+L
237
0
2 37 .0
I
5
44.21
SC
10.3
,4
0
CF+L
237
0
237 .0
1
5
44.83
SC
2.9
9
9
0
CF+L
112
0
112.0
0
7
40.2
sr.
2.9
9
9
0
cr+L
112
0
112.0
0
7
41,74
SC
2,9
9
9
0
CF+L
112
0
112.0
0
7
42. 78
SC
2.9
9
9
0
CF+L
!12
0
112.0
0
7
36.7 9
SC
12.4
4
0
CF+L
112
0
i 12.0
0
7
40.2
SC
12.4
4
0
CF+L
112
0
112,0
- 0
7
41.74
SC
12.4
4
0
CF+L
112
0
112.0
0
7
42, 78
SC
12.4
4
0
CF+L
) 12
0
112.0
0
7
36.79
SC
2.9
9
9
0
CF+L
105
0
105.0
0
6
37,04
SC
2.9
9
9
0
CF+L
105
0
105.0
0
6
38.58
SC
2.9
9
9
0
CP+L
105
0
105.0
0
6
39.62
SC
2.9
9
9
0
CF+L
105
0
105.0
0
6
33.63
SC
12.4
4
0
CP+L
105
0
105.0
0
6
37.04
SC
12.4
4
0
CF+L
105
0
105.0
0
6
38.58
SC
12,4
4
0
CF+L
105
0
105.0
0
6
39.62
SC
12.4
4
0
CF+L
105
0
105,0
0
6
33.63
SC
.8
9
9
0
CF+L
359
0
359.0
3
3
77.62
SC
.8
*.
-------
_ 73 -
of what falls out of the Treatment Systems Table will have been recognized in
the process of defining the key pollutants and unit processes. However, some
things may be revealed for the first time only after the system of treatment
chains has been put together in the table,)
In this case study, several points are immediately obvious. The first of
these is that SO2 control costs drive the costs of the system. In the case of
high-sulfur coal plants at a 1.2 standard for S02 and a .1 standard for TSP,
they represent about 85% of the total sys :em cost. For very dirty coal, the
ratio is even higher at nearly 90%. Additionally, removal of sludge is rela-
tively cheap amounting to less than 3% of the total system cost; and particulate
control is similarly small, accounting for between 15 and 3 5% of total system
costs depending on the type of coal being burned.
The second interesting point is the relationship of aver age cost per ton
removed of SO2 and TSP. For 3.5% coal the average :ost per ton removed at the
1.2 standard is 8 times greater than the average cost per ton removed for TSP at
the .1 standard. Because these are the current standards for these pollutants
it might be worthwhile to ask the question "Did EPA decide that removal of a ton
of SC>2 is really worth eight times as much as removal of a ton of particu-
lates?" "How did they arrive at this relationship?"1
The third point is the sharp jump in total costs between a .5 standard and a
1.2 standard for SO2 when 3.5%-sulfur coal is burned. Upon closer examina-
tion, it becomes apparent that the extra cost is related to physical coal clean-
ing. Because this jump is so large (?26M), it raises the question as to whether
it might be possible to reach the .5 standard without physical coal cleaning?
And, if so, why hasn't this alternative been analyzed?
A fourth obvious point is that a 1.2 standard for low-sulfur coal plants is
very expensive on a ton-removed basis. The reason for this is that low-sulfur
coal itself contains only 25% more sulfur than the standard allows to be emit-
ted. Tn contrast 3.5%~coa1 emits almost "ive times as much potential sulfur
emissions as the standard permits.
Unfortunately, we place little credence in these numbers as representing the
total systems cost for the model entity. In part, this is because we were un-
able to find enough quality data about wa'ier pollution costs. It is possible
that some water and air pollution problems interact and ace important to any
analysis of the costs of cleaning up the entity. Finally, we can readily see
the need for identification of more treatment systems than the 48 listed. Not-
withstanding these Limitations, we performed this analysis recognizing that the
results are intended to shew the application of the proposed methodology rather
than to draw conclusions about the proper way to regulate new coal-fired power
p 1 ants..
5.1 Identify the Policy Ouestion
—- ¦¦*-¦¦¦¦ rt _fflh
Earlier in this chapter we identified two plausible policy questions that
the methodology should be able to address. These wsre:
What is the MCE of alternative, nore stringent sulfur dioxide standards?
— What is the MCE of trading-off particulate control for sulfur-dioxide
contro!?
-------
- 74 -
In the succeeding paragraphs we wilt analyze each of these questions applying
the relevant steps identified in the methodology•
What is the MCE of alternative, mote stc ingent sulfur dioxide standards fox
new coal-fired power plants? The I i rst step in answering this question is to
identify all of the relevant Treatment Systems from the Table. Holding ail of
the other standards constant indicates rows 1, 9, 17, 25, 33 , and 4; are those
of interest for analyzing changes in the S02 standard. The total costs asso-
ciated with these systems are shown in Exhibit 21.
EXHIBIT 21
TOTAL COSTS OF THE RELEVANT TREATMENT SYSTEMS
Coa 1
Treatment
(Su1f u r
SO 7
1 Abated
SO? Emitted
Total Cost
System
Content)
Standard
'(1,000 tons)
{1,000 tons)
(Annua 1 izesl
I
3.5%
1.2
140, 7
17. L
39.91
9
3. 51
.5
150. 7
7.1
58. 34
17
3. 5%
90*
349.9
7.9
4 5
25
. 8%
90%
155.7
1. 1
40.2
33
.0%
150. 7
7.1
37.04
41
7 . 0%
. ri
150.7
7.1
77.62
The next step is to assign nonseparable costs. In this example, there are
none, so we may omit this step. Likewise, we do not have to assign weights to
pollutants because this policy issue addresses only one pollutant.
Next, the MCE ratios are computed and analyzed. In this example a problem
arises in that the alternative cuels !low-sulfur and high-sulfur) create dif-
ferent amounts of potential emissions to be abated (low-sulfur 21,3 thousand
* on is; high-sulfur 78.8 thousand tons; highest sulfur 157.8 thousand tons). By
defining the entity an burning the highest sulfur coal, and using that as the
base, it is possible to calculate the amount of abatement achieved by
switching to either of the lower sulfur coals.
Having defined the amount of abatement, the question then is, "Over what:
inter val/i ncrement are we interested in the incremental costs?" Because we
are choosing among alternative standards for new sources we assume that the
most rational increment is from an uncontrolled condition.
Kxhibit 22 shows the average cost-effectiveness for the 6 alternatives
when the increment is from the uncontrolled hasp. Based on these results it
would appear that the appropriate standard to impose would be .5% and that new
coal-fired power plants should burn low-sulfur coal.
-------
EXHIBIT 22
AVERAGE COST-EFFECTIVENESS FOR THE SIX ALTERNATIVES
A1 tematiw
Coa1 Std.
Aba tement
Aveiage
Cos t-E ff ec ti veness
IncrementaL
3.51 S
1.2
140. 7
.284
W
¦—
3-51
0.5
ISO. 7
. 387
58. 34
3. 5%
90%
149.9
i
. 289
43.25
0.8%
90%
155.7
. 258
40,
0.8%
.5
150.7
.246
37.0
7.0%
.,5
150.7
. 515
77. 62
This application of the methodology illustrates a point worth noting. By
comparing costs and effectiveness from one point (zero abatement) to three
alternative points (standards set at 1.2, 90% ar,d ,5), we are ln effect com-
puting average and not marginal cost-effectiveness. Exhibi t 23 displays the
total cost and abatement points on a Total Cost/Total Abatement graph. As is
readily apparent any effort to fit a carve to these particular points would bo
misleading. Furthermore, the marginal cost curve (the derivative of the spec-
ulatively created total costs curve) would not be useful to this decision un-
less EPA had chosen a threshold price per ton abated that represented the mar-
ginal social benefit of removing S02. Absent this price, the above analysis
represents the best tool for answering the question.
EXHIBIT 23
TOTAL COST/TOTRX> ABATEMENT GRAPH
i >)
50
40
Ki t
1
V/-
14 ^
1 b* >
I fsf I
-------
to -
What_j:; the__MCE of t r ad inq -of f particulate control foe sulfut thoxide con-
trol? This question requires comparison of the MCE curves for S02 control
with those for TSP control. For clarity we will only examine the cases vher<-
the fuel is 5.5% sulfur and 14.0% ash. These cases .ire descr ibe-1 by Row, I co
IS. For TSP control the important rowr where all pollutants except TSP iV"
held constant, are shown in Exhibit 24.
EXHIBIT 24
TSP DATA FROM ROWS i to 4
Treatment
System Abatement
Emissions
Unit
Process
1
1 35.2
1.4
5.06
2
135.9
.7
5.55
3
I 36 . 2
0.4
6.02
4
136.2
0.4
6.64
For SO
'2 control the relevant
rows
are shown in Exhibit
25,
EXHIBIT 25
S02 DATA
FROM
ROWS 1, 9 AND 17
Tcea tine nt
System Abatenent
Em i lisions
Un i t
Process
1
61,8
17.1
33.15
9
7 1,8
7,1
52. 18
17
70.9
7.9
36.69
Again we have to decide how to measure MCE. Because we have a constant
"influent" (the sulfur content of the coal is held constant), the data allows
the value to be computed in a .straightforward manner. The MCE' s For TSP .are
the MCE's of going from 0 to Row 1, from Row 1 to Row 2, and from Row 2 to Row
1. Row 4 would be eliminated fran this analysis because it is not the least
cost ly way to achieve the level, of abatement. The resulting HCR'ss ar? -,;hown
in Exhibit '2b.
EXHIBIT 26
MCE OF TSP
MCE 0-1 ,037
MCE 1-2 .700
MCE 2-3 1.56
For SO-,; the MCE's would he calculated for 0 to Row 1, Row 1 to Row 17,
Row 17 to Row 9. Exhibit 27 lists these MCE's:
-------
EXHIBIT 2?
MCE OF S02
MCE 0 - 1
MCE 1-17
MCE 17-9
. S3f>
, 389
19.4
A comparison of the values for SO^ reveals an interesting phenomena. The
marginal cost effectiveness of going from Row i to 17 (where emissions decrease
significantly} is less than the MCE of achieving the initial standard. This
would suggest thaf the more appropriate initial standard would h.sve been at row
17.
In contrast, t ht» MCE for controlling TSP rises with incce-isinql/ stringent
st andards. This is consistent with expected results.
As for trade-offs between pollutants, the data suggests that if Row 1 repre-
sents current standards, it would be more cost effective to impose Row 17*s
standard of SO2 than to impose Row 2's standard on TSP assuming equal weights
for those pollutants. However, if a ton of S02 was considered twice as bad as
a ton of TSP, then the two pollutants would have to be normalized on the verti-
cal .ixis. In thiii cdsif, the MCE of SOj would drop to .195 which would further
confirm the desirability of imposing the tougher SO9 standard rather than the
tougher TSP standard.
Finally, it is worth noting the effect of employing a threshold in this ana-
lysis. If the MCE threshold was set at .5 and equal weights are assaned, then
TSP would probably be controlled to the Row 2 level while S02 would be con-
trolled to the Row 17 level.
Because of the limitations and questionable validity ot thf available data,
no firm conclusions concerning alternative new source performance standards for
coal - fired power p]ants can be reached. This case study has, however, illustra-
ted several interesting aspects of applying the MCE methodology. Above all, it
demonstrated the methodology cannot be applied in a mechanical fashion. With
less than the desired amount data available and with specific policy questions
to address, manipulations of the basic methodology will often be required.
The case study convincingly demonstrates that the methodology does provide a
comprehensive tool for use in setting aqeicy policy. By first establishing a
data base of all possible combinations of" treatment chains with their respect 1 v->
cost and levels of abatement, it become:-, oossible to address a variety of policy
decisions. The first example of this ca;?3 study (comparing proposed standards
For SO2) showed that even though MCE may not be applicable to specific policy
questions, the basic methodology still proves useful. In this instance, it pro-
vided the data for doing the more appropriate average cost-effectiveness analy-
sis. In the second example, we introduced the use of weights in comparing the
composite MCE where the standards for two pollutants were being examined con-
currently.
SUMMARY
-------
Finally, we suspect that future applications of the raethodoloqy w < i r«s;j!.*;
in Furtlior r e E i rvitwnt of the met itodo logy. At thijunctsn^ in '.h» :nri.?nr
of MCE analysis, we have attempted mainly to develop as r-omprRhenj; i vt- fl^x:
ble an analytical device as possible. For this reason, applications of the
methodology to specific policy questions demand the use of careful judgment.
-------
*
- 79 -
6 - IM P L EKE NT AT IOM CONOKRNS
The development and testing of methods for measuring the marginal cost
eifectiveness of EPA regulations was much more complex and difficult than ori-
ginally thought, A number of factors affect the complexity of the analyses.
These, in turn, raise theoretical issues which impact on the rigor of results
obtainable. As a result:, a serious application of MCE analyser, by EPA may in-
quire collecting and analyzing a substantial amount of new data.
The implications of the complexity and theoretical issues raised in this
phase will be clearer when the Phase II analysis is complete; the purpose of
this Chapter is to summarize our major implementation concerns as of this point
in the development of. the methodology.
COMPLEXITY ISSUES
The methodology proposed is designed to answer some rather sophisticated
policy questions. It is no longer practical to consider abatement of single
pollutants alone. Moreover, EPA has come to recognize that standards in one
media impact pollution in other media: reinoving no, from smokestacks creator,
water ptohlems, and abatement of water problems creates sludge disposal pro-
blems. Because of this, the MCE methods are designed for cross-media as well as
intermedia coupar isons.
The met hods are also "future looking" and are therefore designed to consider
new abatement processes. Moreover# most existing industrial plants have some
abatement system in place which must be accounted for in the analysis. For
these reasons, the methodology uses an entity (mode] plant, mode I process, mo-
bile source) as the basic building block for both cost and abatement (effective-
ness) analyses. Because the methods are designed for analysis of regional and
national as well as industry-wide policy questions, the entity analysis must
also identify factors Cor aggregating to appropriate levels.
Data and TochnologieaI Complexitles
With thir background in mind, several issues which introduce con^lexities
into implementation can be presented. Some are associated with the amount and
multidimensional nature of required data. Others center on technological com-
plexities, The issues are;
® Number of Entities - the number of entities included In a single
lndnsti y can be large. There were 26 model plants which re-
quired analysis in the Textile Industry case study. Moreover,
each of the 26 had five or six separate and distinct production
processes that, in theory, should be analyzed for potential in-
plant/process change abatement options. Although the Textile
Industry is not homogeneous, it is probably not atypical of what
can be ex-pec ted tor other industries.
-------
• Number of Unit Processes - The number of unit processes for
abatement was small in both case studies, but the number oC com-
binations of processes (treatment chains) that required analysis
was large. Furthermore, the number of treatment chains for
which data were provided by the literature were not sufficient
Cor proper PCE; gome apparently obvious combinations were not
studied by the engineering contractors. Despite the lack of suf-
ficient data," the combining of treatment chains or systems with
entities led to large and unwieldy Treatment System Cost and
Abatement Tables in both case studies.
3 Interdependence Among Pollutants - Both case studies showed that
there were interactions among pollutants. This seems fundamen-
tal. to water pollution abatement. For example, a single organic
substance in suspension will be recorded as all three conven-
tional pollutants (BOD, COD, and TSS), and a unit process in-
sta!I'hI to remove solids will also remove other pollutants.
This causes both theoretical and practical problems in cost
assignment. Because of this, the Textile Industry case study
was Iinited to a consideration cf conventional pollutants as a
group rather than on an individual basis.
A second complexity is introduced when seme process, aimed say at TSS, re-
moves noncorvent ional or priority pollutants in suspension. K determination
must be made whether this removal is "incidental" or a principal concern.
Theoretical Complexities
Problems related to MCE theory also complicated the research effort and pose
seme complexities to implementing the methodology. These issues are:
® Well-Ordered Treatment Chains - The theory requires a nonde-
creasing cost-effectiveness function. This means that unit pro-
cesses must be combined in such a manner that the total system
cost increases and that the amount of each pollutant removed
does not decrease. This process was somewhat easier in the case
studios when individual pollutants were aggregated in conven-
tional and nonconventional levels. Nonetheless, where ICE ra-
ti 'x; wi'ff calculated, the resulting plot to obtain a MCE curve
ihowed a surprising dispersion of points. We <],-> not know whether
this was the result of poor data {inclusion of treatment chains
that were not cost effective) or poor analysis (improper mea-
surement of abatement).
• Number of ICE (MCE) Data Points - The development of an HCE ra-
tio envelope curve foe an individual entity and the ability to
aggregate to industry, region, or national totals require a rea-
sonable number of data points covering a reasonable range of MCE
values. Each data point requires a properly constructed treat-
ment chain and data points are required over an extensive effec-
t i v-riess range. Because many EFA studies have concentrated on a
particular standard (level of effectiveness}, alternatives to
that sfandard and the cost-effectLveness relationships it im-
poses ari> not always available.
-------
- 81 -
• ICE (MCE) Interval - The data available for the case studies was
not sufficient to develop a statistical approximation to the
underlying costeff cot iveness function. Instead MCI' results wero
approximated by ICE analysis. ICE is a good estimator of MCE
under certain condi tions, but the Textile study revealed a sig-
nificant problem with its use. The addition of a unit process
to a treatment chain produced a discrete iricuase in effective-
ness, often a substantial change. These intervals were not only
large but varied substantially from entity to entity. The di-
rect comparison of ICE points is, therefore, difficult. The
best solution to this problem uncovered to date involved usinq
the longest interval among entities as a basis for comparison
and aggregation- Because the accuracy of ICE as an estimator of
the true MCE degrades with interval length, this practice in
undesirable. Where possible, comparisons should he based on
best estimates of the MCE function.
INCREASED DATA AND ANALYSIS REQUIREMENTS
The issues discussed above introduce complexi tles to implementation not so
much because they cannot be resolved but because of the scope and detail of the
tasks required. We believe that significant gaps exist in the data available,
but that these gap; can be closed by more complete and eifactive systems analy-
sts at the entity level- Moreover, it seems probable that the quality of data
available in completed- engineering studies can be improved by interaction with
selected labs or contractors.
If one may generalize from the Textile Industry case study, information gaps
for the analysis of water problems are:
© Raw Influent, 3PT Effluent and BAT Effluent Cnaractcristics -
The study assumed that BTP was in place and that various unit
processes and combinations would be added to achieve a reason-
able BATFA qoa1. No information on raw influent was availab1e;
only aggregated character ist ic data for BTP and treatment chain
effluent was available. Knowledge of the effluent characteris-
tics for each unit process in each treatment chain is essential
for cost assignment. Knowledge of * he actual constituents (com-
pounds) in the effluent stream would be helpful. Knowledge of
the hydraulic loading and effluent characteristics associated
with each product ion process is essential for considering in-
p! .int/process change options.
• Treatment Chain Logic - The reasons for selecting particular
unit processes and the sequence in which they are applied should
be ful.ly explained. Although infrences can be drawn if complete
st••{>-by-step effluent data are available, the logic of assembly
will allow a sound engineering "incidental removal" criterion to
be established. Furthermore, the data indicated that sequencing
of the same unit processes influences total abatement, that is,
arranging unit process A, B, and C in one order (A -fi-C) gave a
different toal abatement from another order (B-A-C). Finally,
specifying the logic would insure that all logical combinations
-------
_ 09 _
are considered and that ail pollutants addressed ace treated
simultaneously. Curiously? the textile industry engineering
analysis did not include a treatment chain that would reduce all
pollutants and omitted one combination of unit processes (ozona-
tion with activated carbon) which seemingly would reduce all
pollutants.
• Complete Cost Analysis - Both investment and operating cost es-
timates should be prepared Cor each unit process in each treat-
ment chain tested. This is essential when cost assignment is
requi red.
© Cost-Effectiveness Analysis - A cost-effectiveness study, not a
mere reporting of cost and effectiveness is required. Each
treatment chain presented should be the least-cost method for
achieving the desired level of effectiveness.
® Cross-Media Consideration - It is important to include impacts
on other media. The sludge problem received seme attention and
sludge disposal was included as a cost element in the textile
engineering analysis, but no discussion of potentially toxic
content was included. As more and more nonconventional and per-
haps priority pollutants are removed from waste water streams,
the difficulties and costs of sludge treatment and disposal in-
crease. Similarly, an analysis of emission problems is required
particularly when the industry uses process stream in production
processes.
The Coal-Fired Power Plant case study revealed similar information problems.
More specifically, information on the interrelationships among pollutants and
unit processes was sketchy, cost and effectiveness points were Limited even when
a continuous function was available, and data was lacking on at least several
feasible combinations of unit processes.
Implications for Phase II and Beyond
The data and analysis problems cited above can be solved but at some cost in
both money and time. The problems suggest that our Phase II effort be concen-
trated on a few key industries so that truly useful results can be obtained. It
does not appear that a broad brush study of a large number of industries based
on available data would produce results much more applicable than the two case
studies completed in this Phase.
The Phase II study will, of course, clarify the issues raised here. It
seems I ikely, however, that a comprehensive application of MCE methods will re-
quire increased resources for more detailed analyses of abatement options. EPA
must therefore assess the cost effectiveness of employing MCE at some future
time when better estimates of resource requirements are available.
------- |