DRAFT:
                  Assessing the Benefits of Policies
                 Designed to Reduce Acid Deposition:
                A Decision Analytic Benefit-Cost Framework
energy • engineering • economics • fii
anciaT analysis • environment -minei
Is - water • air -energy - engineering • e
niomics • financial analysis • environ
>nt -minerals -water • air -energy -env

   P.O. DRAWER O, BOULDER, COLORADO 80306 (303) 449-5515

-------
                                                              950D82001
                     Energy and Resource Consultants, Inc.	
                                               DRAFT:
                                    Assessing the Benefits of Policies
                                   Designed to Reduce Acid Deposition:
                               A Decision Analytic Benefit-Cost Framework
Disclaimer:                                        ;          ,


This is a draft report and has not been formally reviewed by the Environmental Protec-
tion Agency.  This is a contractor report. The analysis and conclusions are those of the
contractor and do not, necessarily, reflect the views of the Environmental Protection
Agency.

-------
                    Energy and Resource Consultants, Inc.
                    ASSESSING THE BENEFITS OF POLICIES
                   DESIGNED TO REDUCE ACID DEPOSITION:
              A DECISION ANALYTIC BENEFIT-COST FRAMEWORK
                                Prepared for:

                          Economic Analysis Division
                                   of the
                  United States Environmental Protection Agency
                                    By:

                              Daniel M. Violette
                             Donald C. Peterson
                     Energy and Resource Consultants, Inc.
                               P.O. Drawer O
                             Boulder, CO  80306
                                April 21, 1982
Our Reference: ACIDR

-------
                     Energy and Resource Consultants, Inc.
                               TABLE OF CONTENTS
                                                                             Page

       Preface	      i
1.0    Problem Statement	    1-1

       1.1   Issues	    1-1

       1.2   Types of Policy Options and Their Information Requirements	    1-2

       1.3   The Role of Benefit-Cost Studies	    1-3


2.0    Probabilistic Benefit-Cost Analysis	    2-1

       2.1   Uncertainty In the Estimates of Damages and Benefits	    2-1

       2.2   Evaluating Whether the Available Information is Adequate for
            Setting Policies	    2-5


3.0    One Approach to Probabilistic Benefit-Cost Analysis	    3-1

       3.1   Steps in the Analysis	    3-1

       3.2   Sample Calculations	    3-3

            3.2.1   The Value of Additional Information	  3-13

            3.2.2   The Timing of the Decision	  3-18

                   3.2.2.1    Irreversible Capital Investment in
                             Pollution Control Equipment	  3-30

                   3.2.2.2    Sensitivity to the Analysis of Different
                             Probability Estimates	  3-3*

       3.3   Discussion        	  3-36


4.0    Probabilistic Damage Functions and Estimates	    4-1

       4.1   Elicitation of Probabilistic Damage Functions	    4-1

       4.2   Discussion of the Elicitation Procedure	    4-9

       4.3   A Sample Application: Damages to Forests from
                   Acid Deposition	  4-11

       4.4   Summary         	  4-30

-------
                     Energy and Resource Consultants, Inc.
5.0   Bibliography           	   5-1


Appendix A - Review of the Literature on the Economic Benefits
              of Acid Deposition

Appendix B - Information Needs and Available Information for Acid
              Deposition Benefits Studies

-------
                    Energy and Resource Consultants, Inc.
                                     TABLES
Table 3-1     A Probabilistic Approach	  3-2

Table 3-2     Basic Data for Sample Calculations	  3-4

Table 3-3     Total Costs if Strategy 1	  3-9

Table 3-'*     Total Costs if Strategy 2	  3-10

Table 3-5     Total Costs if Strategy 3	  3-11

Table 3-6     Loss Function if Strategy 2 is Chosen But is Incorrect	  3-16

Table 3-7a   Flow of Costs and Benefits for Strategy 1 with a Five
             Year Research Plan - Outcome:  Low Damages from Acid
             Deposition -  	  3-21

Table 3-7b   Flow of Costs and Benefits for Strategy 1 with a Five
             Year Research Plan - Outcome:  Moderate Damages from
             Acid Deposition -	  3-22

Table 3-7c   Flow of Costs and Benefits for Strategy 1 with a Five
             Year Research Plan - Outcome:  High Damage from Acid
             Deposition -  	  3-23

Table 3-8a   Flow of Costs and Benefits from Strategy 2 with a Five
             Year Research Plan - Outcome:  Moderate Damages from
             Acid Deposition -	  3-25

Table 3-8b   Flow of Costs and Benefits for Strategy 2 with a Five
             Year Research Plan - Outcome:  Moderate Damages
             from Acid Deposition - 	  3-26

Table 3-8c   Flow of Costs and Benefits for Strategy 2 with a Five
             Year Research Plan - Outcome:  High Damages from Acid
             Deposition -  	  3-27

Table 3-9     Total Costs for Alternative Strategies and Research Plans	  3-28

Table 3-10   Net Benefits for Alternative Strategies and Research Plans          3-29

Table 3-11   Total Costs for the Alternative Strategies and Research
             Plans Assuming Irreversible Capital Investment in Pollution
             Control Equipment	  3-32

Table 3-12   Net Benefits for the  Alternative Strategies and Research
             Plans Assuming Irreversible Capital Investment in Pollution
             Control Equipment	  3-33

-------
                     Energy and Resource Consultants, Inc.
Table 4-1

Table 4-la

Table 4-lb

Table 4-2


Table 4-3

Table 4-4



Table 4-5



Table 4-6



Table 4-7
                                                               Page

Elicitation Table	    4-5

Elicitation Table (partially completed)	    4-7

Elicitation Table (completed)	    4-8

Damage to Forests from Acid Deposition:  Steps in the
Analysis	   4-12
Percent Decline in Biomass for Red Spruce Due to Acid Rain
4-15
Net Volume of Growing Stock on Commercial Timberland in
New England, New York and Pennsylvania: Current Exposure
to Acid Precipitation	   4-18

Net Volume of Growing Stock on Commercial Timberland in
New England, New York, and Pennsylvania: Exposures to
Rainfall pH Given a  15% Increase in Anthropogenic H+	   4-23

Net Volume of Growing Stock on Commercial Timberland in New
England, New York,  and Pennsylvania: Exposures to Rainfall
pH Given a 25% Reduction in Anthropogenic H+	   4-24

Net Volume of Growing Stock on Commercial Timberland in
New England, New York, and Pennsylvaina: Exposures to
Rainfall pH Given a  50% Reduction in Anthropogenic H+	   4-25
Table 4-8    Cumulative Probability Distributions of Forest Damages
            for the Four Scenarios	  4-26

-------
- Energy and Resource Consultants, Inc. -



                                     FIGURES

                                                                            Page

 Figure 3-1   Policy Crossover Points ......................................   3-7

 Figure 3-2   Total and Marginal Costs of Emissions Control ..................  3-12

 Figure 3-3   Cumulative Probability Distribution for the Damages
             of Acid Deposition ..........................................
 Figure 4-1   Estimated Damage Functions for Rainbow Trout .................   4-4

 Figure 4-2   Cumulative Probability Distribution for Damage to Red
             Spruce Exposed to Rainfall with an Annual Average  -
             pH of 4.5 [[[  4-14

 Figure 4-3   Marginal Probability Distribution for Damages to Red
             Spruce Exposed to Rainfall With an Annual Average
             pH of 4.5 [[[  4-17

 Figure 4-4   Baseline Distribution of Damages to Forests in New
             England, New York and Pennsylvania ...........................  4-20

 Figure 4-5   Distribution of Damages to Both Hardwoods and Softwoods
             in New England, New York and Pennsylvania with a 15%
             Increase in the Concentration of Anthropogenic Hygrogen
             Ions  [[[  4-27

 Figure 4-6   Distribution of Damages to Both Hardwoods and Softwoods
             in New England, New York and Pennsylvania with a 25%
             Reduction in the Concentration of Anthropgenic Hydrogen
             Ions  [[[  4-28

 Figure 4-7   Distribution of Damages to Both Hardwoods and Softwoods

-------
                      Energy and Resource Consultants, Inc.

                                     PREFACE
The purpose of this project was to consider different approaches for conducting benefit-
cost studies in the evaluation of policy options for reducing acid deposition.  There are
several basic  steps in the performance of  this type of baseline assessment.  First, the
issues that pose critical questions for the development of the appropriate control strate-
gies must be identified.  Second,  the data that are available to evaluate the alternative
policy  options must be assessed,  and  important data gaps identified.  Finally, the dif-
ferent approaches available to perform an  integrated benefit-cost  assessment should be
evaluated.  One objective  of this project is to consider new methods to  perform such
assessments.

The design of projects  to assess the benefits of strategies to mitigate acid deposition is
particularly difficult.  The complexity of the affected aquatic and  terrestrial systems
have made it  difficult to identify the effects of acid deposition with any degree of cer-
tainty.  For example, it is  hard to isolate the effects of acid deposition from the num-
erous other factors  that could be adversely affecting a particular  tree, fish species, or
other environmental attribute. In addition, analyses of acidic precipitation  extend across
many scientific disciplines  and areas of research. This breadth poses problems of inte-
grating information generated by different  research disciplines and requires that a bene-
fits assessment be truly interdisciplinary.  One of the most difficult  integration  tasks is
between the  output of  the  atmospheric  transport  models and  the  research  on the
effects.  For example, long range transport models predict wet and dry sulfate deposition
in terms of kilograms per hectare  per year (kg/ha'yr),  while many researchers of aquatics
and  terrestrial effects  use  micro equivalents  of  hydrogen  ions  per  liter of  rain
(H+ Aeq./l) as the indicator of deposition in their damage assessments.  The conversion
from one value to the other is  not easy and  requires estimates of the seasonally adjusted
volumes of rain, the H* contributions from other sources (e.g., NOX  emissions and natural
sources), as well as  an  interpretation  of whether the indicator (in  this case H+ ion con-
centration) is actually  meant  to  represent other factors as well.   Also,  it  is  not un-
common for scientists  to use one specific factor as  an indicator of  stress which is ac-
*For example, the Henrikson Nomograph is a model used for  estimating the change In
lake pH as deposition changes. The input to the calculation is  A eq of H per liter.

-------
                      Energy and Resource Consultants, Inc.
tually an index of a larger number of variables.  As a result, policy makers must be cau-
tious when interpreting  scientific testimony which is  focused only upon one factor.
These  considerations  indicate that while  different researchers have  been performing
excellent work within their own areas, the necessary integration of their work has been
very difficult to achieve.  Advance planning, however, can help to curb these difficulties
through the specification of policy relevant research programs.

An extensive survey of scientific literature on acid deposition and discussions  with indi-
viduals  involved in ongoing research was undertaken to examine  the availability of in-
formation  for benefits studies.  In addition to assessing the available information, this
interaction with researchers in the different disciplines presented an opportunity to dis-
cuss the design of benefits studies with researchers generating the basic data.

Several recommendations have been developed during this project. One recommendation
is  to start  designing an integrated benefits or policy assessment methodology immediate-
ly.  This  would accomplish several objectives.  By designing an integrated assessment
methodology prior to the completion of many of the scientific studies being performed as
part of  the National Acid Precipitation Assessment Plan, it would help to assure that all
the linkages and information required for policy evaluation are actually being developed,
and that they will, in fact, be integratable.  Identifying the items  required by an evalua-
tion or benefits model, but not currently available is one method of identifying  data gaps
and assigning research priorities.  Developing an integrated assessment model in some
detail while much of the scientific research is ongoing will help to assure that the out-
puts of, say,  the  atmospheric models will,  in fact, be the inputs needed  to assess the
impacts of acid deposition on aquatic and terrestrial ecosystems as well as in other dam-
age categories.

A  second  recommendation is  to develop an explicit  procedure for  characterizing the
uncertainty in the analysis.   Any  assessment of  strategies to mitigate acid deposition
must necessarily deal with many uncertainties. The way  in which these uncertainites are
handled will be important in addressing several critical policy questions and will, in large
part, determine the usefulness of the results.  A  suggested method for  incorporating
uncertainty is presented along with an example of  its application.

Finally, one of the purposes of benefit-cost studies is to provide a link between the policy
makers  and the scientists. Attention must be  given to the manner in  which the informa-
                                         ii

-------
                      Energy and Resource Consultants, Inc.
tion is transmitted to policy makers. The results of benefit-cost studies should provide
current information to policy  makers on the critical issues as well as incorporating, to
whatever extent possible, the current state of the  scientific community's opinions in
highly uncertain areas.
                                         iii

-------
                       Energy and Resource Consultants, Inc.

                             1.0 PROBLEM STATEMENT
This section will discuss several issues that are central to the debate on what actions, if
any, should be taken to reduce acid deposition.  In addition, it will  cover, in general
terms, the available policy options and the information needed to choose among those
policies.
 1.1 Issues

 There are a number of contrasting views held by parties who have studied the acid depo-
 sition problem. Several of these are presented below:

 One View:

     o     Substantial damages to the environment are presently occurring and immed-
           iate action is required to prevent potentially irreversible damages.

 Opposing View:

     o     Adequate time is available to collect and analyze new information.

 One View:

     o     The currently available information is adequate for the formation of control
           strategies.

 Opposing View:

     o     The limited  information on the effects of acid deposition on ecosystems, as
           well as uncertainty  regarding the sources of the pollutants makes immediate
           action undesirable.

These statements indicate that one of the focal points of the debate is the timing of the
decision.  To put it another way, when is the available information sufficient to warrant
                                        1-1

-------
                       Energy and Resource Consultants, Inc.
a decision to either take action to reduce acid deposition or to conclusively decide that
additional controls are unnecessary?  Because the complexities inherent in assessing the
effects of acid deposition essentially  preclude the precise documentation of the environ-
mental damages,  a decision,  whenever it  is  made,  will be based  on uncertain in-
formation.  The question then becomes, "At what point is the uncertainty so great that
decisions to act should be delayed while additional information is gathered and, converse-
ly, when is there enough certainty to act?"

The implications seem clear.  A benefits study designed to help resolve these issues must
include a mechanism  for  dimensioning the  uncertainty  around the estimated effects.
Additionally, a study design that can examine the costs  and benefits of delaying action to
allow for the collection of additional information would be useful.
1.2 Types of Policy Options and Their Information Requirements

Given the current status of  acid deposition research, policy makers have three general
courses of action. First, they could decide  to immediately implement controls to reduce
acid deposition.  Second, a decision could be made to do nothing at this time, but to en-
gage in research designed to  provide  better  information  and reduce the uncertainty.
Third, a compromise strategy could be chosen that would take limited action to reduce
acid deposition  in conjunction with a research program designed to provide additional
Information.

Given these three general policy options, the analyst must ask what information will best
assist the policy maker in choosing among these options? Useful information would in-
clude:

     o     Estimates of the current levels of acid deposition.

     o     Estimates of how damages are  likely  to  change  as  acid deposition levels
           change.

     o     Estimates of the probability of different occurrences, and the penalties of
           being wrong.
                                        1-2

-------
                      Energy and Resource Consultants, Inc.
1.3 The Role of Benefit-Cost Studies

The appropriate role of benefit-cost studies in setting and evaluating environmental poli-
cies has often been controversial.  To partially side-step this controversy, the term bene-
fit-cost study  will be broadly interpreted.  In this paper, a benefit-cost study is one that
details  the different  beneficial and detrimental impacts without, necessarily, putting a
monetary value on every attribute.  Some benefits can be valued in dollars while others
can simply be  itemized and compared to costs in a subjective manner.  The need for this
type of benefit-cost analysis would seem unambigious.  It is simply an itemization of the
consequences of a policy in a manner that can be compared to the costs. No attempt is
made to provide  a comprehensive  discussion of these damage categories or policies.
Instead, the focus is on examining the types of  scientific information available and how
benefits studies can be designed to use this information.
                                        1-3

-------
                  —  Energy and Resource Consultants, Inc.	

                   2.0 PROBABILISTIC BENEFIT-COST ANALYSIS
This section presents an argument for the use of probability distributions as an integral
part of the calculation of costs and benefits.  The application of probabilistic methods to
benefit-cost studies of environmental policies is not typically  done; however, there have
been  several  exemplary  applications  (see National Academy of Sciences (1975b) and
Morgan et al.  (1979)).  Conversations with researchers conducting studies on atmospheric
transport of pollutants, terrestrial impacts, and aquatic impacts have indicated concern
regarding the uncertainty  present  in any estimates  of damages  caused  by acid
deposition.  However, no formalized approach for  determining the  dimensions of this
uncertainty  is currently being used.  Assuming a benefit-cost framework, the explicit
incorporation  of uncertainty into benefit-cost estimates would have the following advan-
tages:

      o   It  would help decision makers evaluate alternative policies with uncer-
          tain implications.
      o   It  would eliminate the use of point estimates which represent a "best"
          estimate. The use of point estimates often results in misplaced confi-
          dence in the accuracy of estimates by policy  makers who are not fully
          versed in the scientific complexities of the problem.
      o   It  would allow for the estimation of the value  of  gathering additional
          information.
      o   It  would provide  the information necessary to estimate  the costs and
          benefits of delaying action  to control acid deposition in order to collect
          additional information.
2.1 Uncertainty in the Estimates of Damages and Benefits

There has been an ongoing debate regarding the environmental effects of acid deposition
and the benefits to be gained by strategies to reduce anthropogenic sources of acid depo-
sition.  One of the common themes in this debate is that the complexity of ecosystems,
as well as the limited knowledge of those systems, makes any current assessment of dam-
ages from acid deposition spurious and effectively precludes the use of these estimates in
policy formation.  This position has many proponents and is the keystone of the utility
industry's position on acid deposition. A report prepared by Battelle Laboratories  (1980)
                                        2-1

-------
	Energy and Resource Consultants, Inc.	

 concluded that there is a need for better and more reliable information on acid precipi-
 tation before environmental decisions can be made on a rational basis.   The Utility Air
 Regulatory Group (1981), in a communication to the Administrator of the U.S. EPA, Ms.
 Anne Gorsuch,  concluded that ". . . in light of the troubling uncertainties . .. regulatory
 controls cannot be  intelligently evaluated  until the critical chemical, ecological and
 economic questions  have been resolved."  The  Utility  Air Regulatory Group goes on to
 state that "where multibillion dollar  costs are involved, there must be a high degree of
 certainty that  significant benefits will result from the imposition of such costs.  I think
                                                          JLJL
 everyone will agree that no such certainty presently exists."

 These statements indicate an  extensive concern with the uncertainty surrounding the
 benefits from  acid  deposition mitigation  strategies.   They  also imply that a benefits
 study should explicitly incorporate uncertainty  into the methodology  if it is to be cred-
 ible to,  at least, many of the interested parties.  Much of the  debate on the uncertainties
 of acid  deposition damages seems ill focused. The simple existence of uncertainty does
 not pose a problem.  The problem stems from a need to quantify and dimension the un-
 certainty.  As  yet,  there has  been no systematic  attempt to dimension the uncertain
 outcomes.

 If the uncertainty associated with the benefits can be dimensioned, then benefits studies
 can yield useful policy implications.  For example, Crocker (1981) provides an estimate
 of economic damages from acid precipitation of five billion dollars annually.  If it were
 possible to deduce that this estimate had a .5 probability of being correct, but also a .5
 probability of there  being no damages from acid precipitation, then the  decision to im-
 plement pollution control measures becomes one of deciding how much society is willing
 to pay to avoid a  .5 probability of a  five billion dollar loss.  If society is risk neutral, a
 cost of control  of up to $2.5 billion would be acceptable. A more realistic example would
 be comprised of a number of different benefits  outcomes, each associated with a proba-
 bility of occurrence.  If the information is available, a continuous probability distribution
 of damages from acid deposition can be used. Given information on the appropriate out-
 comes and  their probabilities, a best policy option can be selected.   The problem is to
 derive the probabilities and outcomes.
 * Executive Summary, Battelle Laboratories (1980).
 ** Utility Air Regulatory Group (1981, p. 21).
                                         2-2

-------
                      Energy and Resource Consultants, Inc.
The conventional approach used in benefit studies is to discuss the weaknesses and uncer-
tainties in the data, but then to choose best estimates and proceed with the calculations
as if they were certain values. When uncertainty is incorporated into the analyses, two
approaches are  generally used.  If the damage functions or  monetary values have been
derived statistically from historical data, then confidence intervals around the estimated
parameters  have been used as a measure of the uncertainty. While useful measures of
uncertainty,  these  confidence intervals are based on a  number of  underlying statistical
assumptions  that may or may not  be  true.  Furthermore, when damage functions are
based on laboratory or controlled field studies, confidence intervals  can be difficult, or in
some cases, impossible to calculate.

The second approach for handling uncertainty in benefit-cost studies has been to try to
establish  a lower bound  for the estimate of benefits. This is done by selecting the pa-
rameters  used  in the benefits calculations conservatively, i.e., whenever uncertainty is
present the researcher tries to err on the low side.  By doing this, the researcher hopes to
guarantee, with a high probability, that the actual benefit estimates are greater than or
equal to the  estimated levels. This procedure is useful  in establishing the dominance of
selected policies. If a conservatively estimated, lower bound of the  benefits still exceeds
the costs  of  a control policy, then one can be quite sure that the actual, unknown bene-
fits will exceed costs.   However, there is  no actual dimensioning  of the uncertainty in
this approach.

If  the  range  of uncertain outcomes of strategies to control  acid deposition  is to  be di-
mensioned, it is necesary to base the estimates of the probability distributions on subjec-
tive evaluations by experts.   Two questions then arise.  First, can subjective probability
distributions  be obtained, and second, how accurate will the probability assessments be?

Addressing the second question first, it would seem that the use of subjective probability
distributions  will necessarily be an improvement over a  decision making framework that
uses only a  single "best estimate" of damages.  In choosing  a single best  estimate, the
researcher or policy maker must be selecting this estimate from some underlying, intui-
tive  distribution.  If there is not an underlying distribution, then  there is no basis for
claiming that any particular  estimate  is better than any other.  By not expressing this
                                      2-3

-------
	Energy and Resource Consultants, Inc.	

 intuitive distribution, the researcher is not utilizing all of the information at his disposal.
 In addition, he may be obscuring information important to the correct interpretation of
 his results. As a researcher conducts experiments or performs studies, numerous choices
 and tradeoffs are made in the design and scope of the  project. Often, these choices are
 based on subjective probability assessments of the most likely structure of a causal rela-
 tionship.  Scientific research, as well as environmental decisionmaking, requires that
 numerous  subjective  probability assessments  be made during  the normal  course of
 events.  Making these underlying, intuitive probability assessments explicit should only
 increase the amount and quality of information available to decisionmakers.

 Another advantage to the use of explicit probability distribution is the additional qualita-
 tive understanding that can be gained through the process of quantifying important deci-
 sion parameters.  Examples of the use of subjective  judgements in a structured  model
 have been evaluated in controlled laboratory studies.  Many of these studies have shown
 substantial benefit  from the  use of judgement based models when compared with the
 more  common and untutored  intuitive decision  processes.**  These approaches require
 the estimation of  parameters for which  data are not available.  By  having  to express
 estimates  in unambiguous quantitative terms, researchers and policymakers are forced to
 give more  thought to parameters potentially important to the decision problem.

 There is some concern about whether these subjective probability distributions can  be ob-
 tained for the important decision parameters. Many researchers and most people in gen-
 eral, are uncomfortable with  making  probability assessments when little is known about
 the population.  In this case, the population is  the  set of acid precipitation induced
 damages.  Different methods  can be  used to develop estimates of these distributions.
 Two of  the best known are the fractile approach for the assessment of a judgemental
 distribution (H.  F. Raiffa,  1970), and probability encoding (C. F. Spetzler and von Hoi-
 stein, 1975).  Scientists are occasionally hesitant to express their subjective probability
 assessments when  they  pertain to  research   either  they or  their  colleagues  are
   Keeney and Raiffa (1976, p. 36*0 discuss the advantages in qualitative understanding
 that  were gained through the construction of joint  probability  functions for potential
 outcomes in an air pollution control decision problem.
 **See S.  H.  Mclntyre (1982) for a discussion of the use of these models making marketing
 decisions.

-------
                      Energy and Resource Consultants, Inc.
conducting.  The use of techniques that maintain anonymity should  help reduce this re-
luctance.  Another concern  sometimes expressed by scientists is that this procedure  is
not rigorously scientific. However, the arguments that support the  direct incorporation
of uncertainty into the decisionmaking process would hopefully be convincing enough to
enlist the cooperation of the researchers in the field.  In fact, conversations with indi-
viduals  currently conducting research on the effects of acid deposition have indicated a
lot of interest  in these approaches,  and  an expressed  interest in participating in these
studies.

In summary, the encoding of information in probability distributions is a promising ap-
proach for handling uncertainty in the scientific and economic information on acid depo-
sition damages.   This requires subjective judgements be made, but subjective probability
assessments are implicit even in the generation of  single valued "best estimates." This
procedure allows these judgements to be made openly,  and also subjects them to review.
In addition, explicitly incorporating uncertainty into the analysis brings more information
to bear  on the  problem  at hand.  Finally, any decision to implement a policy to control
acid deposition will be based to a large extent on subjective judgements, (i.e.,  a subjec-
tive evaluation  of the  odds  of being correct  and  the penalties associated with being
wrong).   There are two  choices, either the policy makers are left on their own in  trying
to perform this implicit decision calculus, or experts in their respective fields can assist
decision makers by providing information on these probabilities.
2.2 Evaluating Whether the Available Information is Adequate for Setting Policies

Assuming that acidic deposition causes environmental change, a major issue in the debate
over the merits of policy action to mitigate acidic precipitation concerns the timing of
the decision. There are two important questions.  First, "when is there enough informa-
tion available to make a decision to act?" Secondly, "what is the optimal time frame for
the implementation of the  policy?"  Environmental actions taken at different points in
time can  have  different costs  and  benefits. There are several reasons for this time de-
pendence.  There can be benefits associated with delaying a decision in order to perform
additional research and develop better data on which to make a more appropriate deci-
sion.  A second reason is that the stream of costs and benefits will likely vary over time;
for example, better pollution control technologies may be available in the future.  Post-
poning a decision  to require controls until the less expensive  technologies are  available
                                         2-5

-------
	 Energy and Resource Consultants, Inc.	

 would reduce the  costs of control, and possibly increase the net benefits of environ-
 mental improvements. On the other hand, if immediate control were critical to the pre-
 servation of important environmental benefits, the decision not to implement controls
 could decrease the net benefits.  Both of these issues are important to the current debate
 over the appropriate public policy towards mitigating acid precipitation.

 Industry trade associations as well as many state  governmental organizations have ex-
 pressed a concern  that there is an unwarranted rush to  pass judgement on the sources of
 acid precipitation  and its environmental damages.  Their position is based on two fac-
 tors.  First, they expect that the  costs of controlling SO2 and NOX will be lower in the
 future, and second, they feel that adequate time is available to collect and analyze new
 information on  acid precipitation  in order to make a better public policy decision.   In
 other words, their  position is that there is no immediate, severe threat to the environ-
 ment and, therefore, the benefits to  delaying the decision in order to gather additional
 information to reduce the uncertainty outweigh the costs of the delay, i.e., any environ-
 mental damages that may occur  during the delay.  Others disagree with this position
 expressing the opinion that immediate action to protect  the environment is required.

 This question on the appropriate timing of the decision on acid precipitation has been a
 focal point of the debate.  Perfect information on the effects and causes of acid pre-
 cipitation will never exist, but additional information designed to reduce the uncertainty
 surrounding key decision parameters may be of great value in reaching a "correct" policy
 decision. In decision analysis, procedures have been developed for assessing the value of
 increased information.  This value can then be compared to the costs of gathering the
 additional  information and the risks to the environment from delaying action.  This pro-
 cedure can also be used to identify the specific areas where additional information will
 be of most value.   Given  the debate  regarding the appropriate timing of the decision, a
 systematic examination of the benefits and costs of additional information may be war-
 ranted.
 * Edison Electric Institute (1981).

 **  A recently completed  study  by  the  National  Academy of Sciences (1981) takes this
 position.
                                         2-6

-------
	 Energy and Resource Consultants, Inc.	

 The basic procedures used to value additional information are well known and can be
 found in Hirshleifer and Riley (1979), Raiffa (1970), and Schailfer (1969).  The ability to
 use these  methods is  a direct consequence of using probability distributions to express
 uncertainty as dicussed in section 2.1.  An example  illustrating the use  of subjective
 probability distributions to estimate the value of a research program can be found in an
 analysis conducted by North and Merkhofer for  the National Academy of Sciences (See
 NAS 1975b).  They calculated the value of  resolving the uncertainty  in the  human health
 damages resulting from SO2 in choosing the correct control strategy  for coal fired power
 plants.

 The value of information can be expressed  as the difference between the expected bene-
 fits of a decision based on current information, and the expected benefits of a decision
 based on new information.  The  usual first step is to calculate  an  upper  bound on the
 value of additional information, i.e., the value of perfect information. Of course, per-
 fect information  can  never be obtained, but the calculation of the value of perfect in-
 formation provides a useful upper bound to the value that can be obtained from any addi-
 tional information.  The computation of this upper bound is easy, given that probability
 distributions are used  to dimension the uncertainty.*

 For example, consider the following simple example of how the  value of perfect infor-
 mation can be calculated for a case where there are discrete outcomes. Suppose that the
 damages from acid deposition are estimated to be $10 million.  The uncertainty surround-
 ing the calculation of these damages is such that the probability of these damages oc-
 curring is  estimated at .6, while the probability of there being zero damages is .4. Also,
 the cost  of implementing  controls to eliminate the damages from  acid deposition are
 estimated to be $3 million. The problem can be summarized as follows:

      o     probability = .6 that acid deposition damages are $10 million
      o     probability = .4 that acid deposition damages are zero
      o     cost of controlling acid deposition is $3 million
   See National Academy of Sciences (1975a, p.  185) and National Academy of Sciences
 (1975b, p. 630).
                                         2-7

-------
	Energy and Resource Consultants, Inc.	

 The expected value from the control of acid depostion is .6 x $10 million or $6 million,
 while the costs of control are only $3 million.  Based only on this information the deci-
 sion  would be to implement the controls. However, there is a significant likelihood, i.e.
 *fO percent, that this course of action is incorrect resulting in a loss of $3 million dol-
 lars.  The value of the additional information that would resolve this uncertainty can be
 calculated.  This value  is the probability of being wrong multiplied by the penalty of
 being wrong. In this case, the penalty associated with being wrong is $3 million, dollars
 and the probability of being wrong is .*; therefore, the value of perfect information is (.4
 x $3 million) or $1.2 million.  Since perfect information  will never be available, this pro-
 vides an upper bound to the value of additional information for the decisionmaker.

 If instead, the problem was characterized by the following parameters:

      o     probability  = .8 that acid deposition damages are $10 million;
      o     probability  = .2 that acid deposition damages are zero;
      o     cost of controlling acid deposition is $3 million;

 The decision maker is  now more certain  that the correct strategy is to implement con-
 trols, and the value of  additional information that would eliminate all uncertainty is only
 (.2 x $3 million) or $.6  million. This is one half the value of information calculated in the
 preceeding example. Thus, as the certainty  of the correct choice increases, the value of
 additional information  decreases.

 In examining alternative policies and in deciding whether to  collect additional informa-
 tion, it is often very useful to perform these types of calculations. Without them, deci-
 sion  makers may have  no idea what values they should be placing on additional informa-
 tion.  These rough calculations can serve as benchmarks for  the value of additional in-
 formation.  Without such calculations, informal estimates of the value of additional re-
 search may  be quite inaccurate.  In particular, individuals who are not  accustomed to or
 are uncomfortable with making decisions under uncertainty often place too high a value
 on resolving uncertainty.

 Estimates can also be obtained for the value of imperfect information from research that
 does not eliminate the uncertainty, but instead provides information that can be used to
 estimate a new, more  accurate probability  distribution of damages.   These procedures
 are straightforward in  theory, but can be complex in practice. To estimate the value of
                                         2-8

-------
                      Energy and Resource Consultants, Inc.
a research experiment designed to provide additional information on the damages from
acid precipitation, one would have to develop a subjective probability distribution of the
outcomes of the research prior to the research being conducted.  An estimate of the
value of information from this experiment could be obtained by using Bayes1 theorem to
adjust the prior  probability distribution  of damages given the subjectively  estimated
distribution of research outcomes.   The difference between the benefits of the decision
based on the prior probability distribution and the benefits of the decision using the re-
vised probability distribution of damages yields the estimated value of information from
this  particular research experiment.  This type of pre-posterior  analysis of the value of
information is also advocated by Conrad (1981) to address uncertainties associated with
potential irreversible environmental damages.

An important assumption  in valuing the additional information that can be obtained from
a particular research experiment is the requirement that the researcher be able to gen-
erate a subjective distribution of the likelihood of  the different outcomes of the experi-
ment prior to the performance of the research. Estimating subjective probabilities can,
in general, be an unsettling task and  in this specific case it often proves even more dif-
ficult.  However, researchers often have an intuitive assessment of the likely outcome of
experiments and this procedure is designed to use this information. Still, a useable upper
bound on the value of information can be easily obtained by calculating the value of per-
fect information, as  shown earlier, given that the estimates of the costs of control and of
damages are expressed as probability distributions.  The calculation of this upper bound
on the value of information will be useful in setting research priorities and can, in fact,
help determine if additional research is necessary.

The  estimates obtained for the value of additional information should be compared to the
costs of obtaining that information.  The costs of obtaining the information is comprised
of two  parts — the direct costs of the research and the expected value of the environ-
mental  damages that could occur during the time period the research is taking place. A
.
  An example and discussion of valuing imperfect information can be found in chapter 7
of H. Raiffa (1970).
                                        2-9

-------
                       Energy and Resource Consultants, Inc.
decision to proceed  with a research program should weigh the likely benefits of the in-
formation generated against these costs.

Useful insights can  be gained by an analysis of the benefits  and costs of delaying the
policy decision on acid precipitation to allow additional research to be conducted. Two
pressing problems facing decision makers are the determination of the appropriate timing
of the decision and  whether the available information, although uncertain, is sufficient
for a good policy decision.  These questions can be addressed by the use of probability
distributions  to dimension the uncertainty in the estimates of damages from acid deposi-
tion.
                                        2-10

-------
                       Energy and Resource Consultants, Inc.
         3.0 ONE APPROACH TO PROBABILISTIC BENEFIT-COST ANALYSIS
 One  approach to probabilistic benefit-cost analysis will be  presented  in this section.
 There are a number of approaches that can be used, but they  all have certain basic ele-
 ments in common.

 One  of the primary considerations in the design of  this particular approach is the ease
 with which the information can be presented to decision makers.  One problem that can
 be present in probabilistic analyses is that the complexities of  the methodology can make
 it diffic"lt to communicate the resulcs to  policy makers. It is important to present the
 methods and  results in a manner that facilitates understanding and instills confidence in
 the procedures.   During  the course  of this project,  several briefings were presented to
 non-technical policy makers in an attempt  to determine the aspects of the approach that
 were difficult to communicate.  Many, if  not most, policy makers are  unfamiliar with
 probabilistic  approaches  to benefit-cost analysis and, as a result, may not have much
 confidence in the approach. Thus, it is necessary to design studies whose logic  can be as
 easily understood as possible.

 In  this chapter, care is taken  to make sure that all  the calculations are  straightforward
 and easily reproducible.  For example, rather  than using a computer algorithm to mani-
 pulate probability distributions, a piecewise linear cumulative probability distribution is
 used to allow for straightforward computation.
 3.1  Steps In the Analysis

 The basic steps that must be performed in the approach are summarized in Table 3-1.
 The first step is to identify the available policy options, i.e., the potential control strate-
 gies and their costs. This simply involves an identification of the different options avail-
 able to the policy makers. It is often useful to develop estimates of the  costs of each
 policy option prior to performing the benefits estimation.  Since the costs of environmen-
 tal control options can typically be estimated more accurately than the benefits, knowl-
 edge of the  costs of the different  strategies can be useful in reducing the  dimensions of
 the benefits estimation problem.  This can allow  researchers to take advantage of the
 asymmetry regarding the  relative  precision of  the cost estimates compared to the rela-
 tive inaccuracy with which the benefits can be estimated.

	3-1   	

-------
                      Energy and Resource Consultants, Inc.

                                    TABLE 3-1
                           A PROBABILISTIC APPROACH
Step 1 -    Identify policy options, i.e., potential control strategies and their costs.

Step  2 -   Determine the  underlying assumptions that are necessary for each  policy
           option to be optimal.

Step  3  -   Estimate  the probability  that the  assumptions necessary for each  policy
           option to be optimal are actually the true states of the world.

Step & -    Make the best (tentative) decision given the current information.

Step 5 -    Estimate the value of gathering additional information to reduce uncertainty.

Step 6 -    Decision stage;  Choose between a terminal  policy action or a research plan
           to gather additional informtion.
                                      3-2

-------
	Energy and Resource Consultants, Inc.	

 The second step is to specify the underlying assumptions that must be true for each of
 the policy options to be the correct choice. This step is designed to clearly spell out the
 implications of adopting one policy over another. If a policy is selected for implementa-
 tion, there are a number of implicit assumptions that accompany its choice  as the best
 policy.  For example, one  obvious assumption is that the benefits from the  policy out-
 weigh the costs. The nature of the underlying assumptions will be made more clear in a
 following example.   The purpose of specifying these underlying assumptions is to make
 sure that the  implications  of each policy are well understood and to establish criteria
 against which the benefits estimates can be compared.

 The third step is to estimate the probability that the assumptions necessary for each
 policy option to be the correct choice are, in fact, actually true.  The fourth step uses
 the information gathered in steps 1 through 3 to make the best decision given  the current
 information.  In step 5, the benefits and costs of acquiring additional information  to re-
 duce the uncertainty are estimated.  The final step in the analysis is to decide whether to
 implement a control  strategy or to delay in order to undertake further research.
 3.2 Sample Calculations

 An example set of calculations is presented to illustrate the different steps in the anal-
 ysis.   The numbers selected for this example are chosen for illustration only, but they
 are not too dissimilar from reported estimates.

 The basic data for this  example are shown in Table 3-2. There are three policy options
 being considered.  The first strategy is a "do nothing" strategy resulting in no change in
 emissions; therefore, the current level of anthropogenically caused acid deposition would
 be unchanged.  The second strategy is to implement pollution controls which would re-
 duce anthropogenic sulfur emissions by 20 percent.  This is assumed to cost $1 billion and
 would  reduce deposition  from  current  levels of  30 kilograms per hectare  per  year
 (kg/ha/yr) to 25 kg/ha/yr. The third strategy is to implement controls that would reduce
 emissions by  40 percent.  This is assumed to cost $3 billion  and would reduce deposition
 from the current levels of 30 kg/ha/yr to 20 kg/ha/yr.
  The basic structure of this example comes from National Academy of Sciences (1975b)
 Chapter 13.
                                         3-3

-------
                      Energy and Resource Consultants, Inc.


                                   TABLE 3-2

                      Basic Data for the Sample Calculations
CONSIDER THREE POLICY OPTIONS:
     Strategy 1 -   No change in emissions.
     Strategy 2 -   Implement pollution control which would reduce emissions by 20%.
     Strategy 3 -   Implement pollution controls which would reduce emissions by
                                     DATA
 Strategy

     1

     2

     3
 Control
  Costs
$1 billion

$3 billion
Reduction in
 Emissions

     0

    20%

    40%
Total Sulfate
 Deposition

 30 kg/ha/yr

 25 kg/ha/yr

 20 kg/ha/yr
    Change in
Sulfate Deposition

    0 kg/ha/yr

    5 kg/ha/yr

   10 kg/ha/yr
                                       3-4

-------
	Energy and Resource Consultants, Inc.	

 It is important to recognize that these numbers are chosen to facilitate a comparison of
 different strategies and for ease  of calculation. However, the numbers selected fall into
 a reasonable range of actual estimates. For example, the costs of achieving a 45 percent
 reduction in  total sulfur emissions in the eastern  31 state region  were estimated by the
 Office of Technology Assessment (OTA, 1981) to cost between $3.3 and $4.1 billion dol-
 lars.  The OTA analysis assumed that all the emissions reductions would be met by con-
 trolling electric utilities.  They also  estimated a least cost strategy of achieving the 45
 percent reduction at $3.1  billion.  Thus an estimate  of  $3 billion for an emissions re-
 duction of 40 percent does not seem inappropriate. The OTA  also estimated the costs of
 achieving a  36 percent reduction in emissions to fall between $1.7 and $2.6 billion. The
 cost estimate of $1 billion for a 20 percent reduction in emissions would seem to be a
 reasonable figure.

 Translating  reduced anthropogenic sulfur emissions  into reductions in actual sulfate
 deposition is very difficult. In this example, it is assumed that a 20 percent reduction in
 emissions results in a  16.6 percent reduction in sulfate deposition, and a 40 percent re-
 duction in emissions reduces deposition by 33 percent. The difference between the per-
 cent change  in emissions and  the change in deposition could be due to some fraction of
 sulfur  emissions resulting from  natural  sources  or  nonlinearities  in the atmospheric
 chemistry. The current transfer  matrices for the  atmospheric models being constructed
 as part of the phase in U.S. - Canadian Transboundary  Project use linear chemistry.  This
 implies a proportional  reduction  in anthopogenic emissions and deposition.  After  allow-
 ing for some fraction of deposition accruing from natural sources, the reductions in depo-
 sition resulting from the changes  in emissions assumed  in this example are, at least, plau-
 sible.

 The final assumed parameter is  the current level  of  acid deposition, i.e., 30 kg/ha/yr.
 This figure is an approximate estimate of the average annual acid deposition occurring in
 the  Eastern  U.S. region. Some areas have higher  deposition rates while others are con-
 siderably lower. This  value was  obtained by examining maps in the Interim Transboun-
 dary Report (1981) and assuming roughly equal levels of wet and dry deposition.

 The selection of the different policy  options to be considered and their costs completes
 step 1  of the 6 steps presented in Table 3-1.  The second step is to specify the assump-
 tions that are necessary for each policy option to be the best choice. These are outlined
 below:
                                         3-5

-------
                       Energy and  Resource Consultants, Inc.
      o    Strategy 1 is the correct choice when the following are true:

                total benefits from  strategy 2 are  less than $1  billion, i.e., the
                benefits of the other strategies do not exceed their costs; and,

                the social cost of sulfur deposition is less than $200 million per
                kg/ha/yr (see Figure  3-1).

      o    Strategy 2 is the correct choice when:

                total benefits of a 20 percent reduction  in emission exceeds $1
                billion (i.e., the benefits exceed costs); and,

                the social cost of sulfur  deposition is between $200 million and
                $400 million per kg/ha/yr. (see Figure 3-1).

      o    Strategy 3 is the correct choice when:

                benefits of a 40 percent reduction in emissions exceed $3 billion
                (i.e., the benefits exceed costs; and,

                the social cost of sulfur deposition is greater than $400 million per
                unit (kg/ha/yr) (see Figure 3-1).


These assumptions indicate that any  decision to implement a specific control strategy
incorporates a judgement regarding the social cost of the current level of emissions.  By
specifying the social costs per unit necessary to make each control strategy the correct

choice, this judgement is made explicit.  This allows for an  assessment of the probability

that each of these underlying assumptions Is actually the  true state of the world.  In the

case of acid precipitation, the selections between H+ ion concentration and biologic ef-

fects are principally what are responsible for one  assumption, rather than another being
correct.


These assumptions can  be depicted graphically.   Figure  3-1 shows the control strategy

that minimizes total costs for different values of social cost per unit (kg/ha/yr) of depo-

sition.  The total costs measured on the vertical axis are defined  as the costs of pollution

control plus the social  costs  from the remaining, uncontrolled emissions.  If the  social

costs of deposition are less than $200 million per unit, then strategy I would be the cor-

rect choice.  If social costs per unit fall between $200 and $400 million per unit, then
strategy 2  would be the correct choice.  If social costs are  greater than $400 million per
 This graphical depiction was used by  North and Merkhofer in National Academy of
Sciences (1975b).


                                         3-6

-------
                     Energy and Resource Consultants, Inc.
                                     FIGURE  3-1

                             Policy Crossover Points
Total Cost (9 x 10')

15-

14-


13-

12-


11-

10-

 9-


 8-

 7-


 6-


 5-


 4-


 3-

 2-


 1-
                                                                   Damages/Unlt

                                                                ($ x 106)/(V8/ha-yr)
               100
200
300
                                                         400
                                         500
                                         600
         Choose Strategy 1
       Choose Strategy 2
                     Choose Strategy 3
                          Damages per Unit of Acid Deposition
                                 ($ x 106)/(kg/ha'yr)
                                      3-7

-------
	 Energy and Resource Consultants, Inc.	

 unit, then strategy 3, the tightest  control strategy, would minimize total costs.  The
 calculations made to construct Figure 3-1 are shown in Tables 3-3, 3-*, and 3-5.

 These categories correspond to the  selection criteria used by economists, i.e., the com-
 parison of marginal benefits and marginal costs leading to maximization of net benefits.
 Given the information on the costs of the two strategies and using linear segments to
 characterize the discrete investments, the total cost curve is shown in Figure 3-2a. The
 corresponding marginal costs are shown in Figure 3-2b. The criterion for strategy 2 to be
 the correct choice is that marginal benefits exceed marginal costs.

 Figures 3-1  and  3-2 depict the underlying assumptions that must be true for each of the
 policy options to be the correct choice.  These criteria can select the best policy among
 the three under  consideration; however, there may be policies not under consideration
 that might be superior. Still, through the use of an iterative process where different sets
 of policies are examined in successive analyses, an optimal solution can be approached.

 The example used in this section reduces to the problem of estimating the probabilities
 that damages per unit fall into three categories:

      o    the probability that damages per unit of deposition fall within the inter-
           val between zero and $200 million;
      o    the probability  that damages per unit  of deposition are between $200
           million and $400 million;
      o    the probability that damages per unit of  deposition are greater than
           $400  million.

 While still a very difficult problem,  estimating the probability that damages fall within a
 selected  range is often a more tractable task than estimating the probability that dam-
 ages are equal to a specific value.

 Organizing the information in this manner has several advantages. First, the selection of
 a number of discrete policy options to  be  evaluated more closely resembles the  process
 decision makers actually undertake.  It is unusual for decision makers to view the control
 parameters as continuous variables with a goal of  selecting the level that maximizes net
 benefits.   Instead, a limited agenda of policy options is devised and the alternative poli-
 cies evaluated.
                                         3-8

-------
                     Energy and Resource Consultants, Inc.





                                   TABLE 3-3



                             Total Costs if Strategy 1



                          -- No Reductions in Emissions —



                                    is Chosen
If Social Costs,
i.e., damages per unit
of deposition, are:
($ x 106 per
kg/ha/yr)
100
200
300
WO
500
600
Then Total
Costs are:

($ x 106) =
per year
3,000
6,000
9,000
12,000 =
15,000
18,000 =



Total Social Costs +
($/unit x Loading*)
(100 x 30) +
(200 x 30) +
(300 x 30) +
CfOO x 30) +
(500 x 30) +
(600 x 30) +



Costs of Control
($ x 106)
0
0
0
0
0
0
*The loading is the deposition level (i.e., units/year) that remains after the strategy has



been implemented.
                                      3-9

-------
                      Energy and Resource Consultants, Inc. •

                                   TABLE 3-4
                             Total Costs if Strategy 2
                      — A 20 Percent Reduction in Emissions ~
                                    is Chosen
If
Social Costs,
i.e., damages per unit
of







deposition, are:
($ x 106 per
kg/ha/yr)
100
200
300
400
500
600
Then Total
Costs are:

($ x 106) =
per year
3,500
6,000
8,500
11,000
13,500
16,000



Total Social Costs +
($/unit x Loading*)
(100 x 25) +
(200 x 25) +
(300 x 25) +
(tOO x 25) +
(500 x 25) +
(600 x 25) +



Costs of Control
($ x 106)
1,000
1,000
1,000
1,000
1,000
1,000
*The loading is the deposition level (i.e., units/year) that remains after the strategy has
been implemented.
                                       3-10

-------
                      Energy and Resource Consultants, Inc.

                                   TABLE 3-5
                             Total Costs if Strategy 3
                     — A W Percent Reduction in Emissions
                                    is Chosen
If Social Costs,
i.e., damages per unit
of deposition, are:
($ x 106 per
kg/ha/yr)
100
200
300
400
500
600
Then Total
Costs are:

($ x 106) =
per year
5,000
7,000
9,000
11,000 =
13,000
15,000 =



Total Social Costs +
($/unit x Loading*)
(100 x 20) +
(200 x 20) +
(300 x 20) +
(400 x 20) +
(500 x 20) +
(600 x 20) +



Costs of Control
($ x 106)
3,000
3,000
3,000
3,000
3,000
3,000
*The  loading  is  the  deposition  level  that remains  after  the strategy has  been
implemented.
                                      3-11

-------
               — Energy and Resource Consultants, Inc.	

                                 FIGURE 3-2

                Total and Marginal Costs of Emissions Control
FIGURE 3-2a
Total Cost Curve
($ x 109)
       5 •+•
       3 --
       2 --
       1 --
FIGURE 3-2b
Marginal Costs
($ x 106)
     600 --
     500 --

     400 --

     300 --

     200
     100 --
                                                        Reduction in SO,  deposition
                                                        Reduction in SO, deposition
                                  10
15
                                   3-12

-------
	Energy and Resource Consultants, Inc.	

 A second advantage is that the explicit presentation of the underlying assumptions that
 must be true for each of the policy options to be the correct choice is a useful way of
 organizing  information,  and it helps to inform decision  makers of the implications of
 choosing one policy over another.

 To complete the example calculations, assume that the cumulative probability  distri-
 bution shown in Figure 3-3 represents the results of a probability elicitation. This cumu-
 lative probability distribution was selected  to depict a substantial amount of uncertainty
 regarding the  appropriate policy choice. The cumulative probability distribution shows
 the probabilities of each strategy being the correct choice.  From Figure 3-3 the prob-
 abilities are:

      o     probability (0 <. damages/unit< $200 million) = .35 for strategy  1 to be
            correct
      o     probability ($200 million < damage/unit* $400 million) =  .40 for strat-
            egy 2 to be correct
      o     probability ($400 milion* damages/unit) = .25 for strategy 3

 This probability distribution was selected to give each option a reasonable probability of
 being the correct choice and  to reflect the considerable range of uncertainty currently
 found in the debate on damages from acid  deposition. The .35 probability that strategy
 1, a no control strategy, is correct is very close to the probability of strategy 2 being the
 correct choice.  In addition, the .25 probability that the high control option is the correct
 choice is also significant.  As a result, these probabilities reflect considerable uncertain-
 ty about the correct policy decision.

 Given this information, the decision maker  would choose strategy 2, the most likely cor-
 rect strategy.  However, given these probabilities  there  is a significant likelihood that
 either strategy 1 or strategy 3 could be the appropriate strategy.
 3.2.1  The Value of Additional Information

 The use of probabilities allows the decision maker to calculate  an upper limit to the
 value  of information — the value of perfect information.  The value of perfect informa-
 tion is calculated by multiplying the probability of being wrong times the penalty of
                                         3-13

-------
                 Energy and Resource Consultants, Inc.
                          FIGURE 3-3

                    Cumulative Probability
                 Distribution for  the Damages
                       of Acid Deposition
Probability
    1. DO-
    CK90
    0.10-
    0.00
         0
                     500
          Choose Strategy
                 1
Choose Strategy  Choose Strategy
      2               3
600 Damages/Unit
     ($ x 106)
     kg/ha-yr
                               3-14

-------
	Energy ancj Resource Consultants, Inc.	

 being wrong.  To determine the penalty of being wrong, a loss function is constructed.
 The loss function shows the penalty of choosing strategy 2, when strategy 1 or strategy 3
 is actually the correct choice.  The loss function can be calculated from Figure 3-1. It is
 the difference between the net benefits of selected strategy (i.e., strategy  2), and the
 strategy that would have been correct if the social cost of a unit deposition falls outside
 the range  that makes strategy 2 the correct choice.

 The loss function for this example is depicted in Table 3-6. If strategy 2 is chosen under
 the assumption that damages per unit of deposition are between $200 and $400 and actual
 damages per unit turn out to be zero, then $1 billion is spent on emissions control with no
 beneficial effects at all. In this case, the penalty of choosing strategy 2 is $1 billion. If
 strategy 2 is chosen and damages per  unit are actually $100 million, then the penalty is
 $500 million.  There is some benefit that results  from the emissions controls; however,
 the benefits are not sufficient to justify the costs of implementing control strategy 2.

 The loss function in Table 3-6 provides the penalties of being wrong. Now the probabili-
 ties of being wrong must be estimated before the  value of additional information can be
 calculated. To simplify the calculations, the cumulative probability distribution in Figure
 3-3 is approximated by three linear segments over the intervals: 0 < damages/unit <, $200
 million; $200 million < damages/unit <, $400; and $400 *_ damages/unit.  The  use of a
 piecewise  linear cumulative function assumes equal probabilities  for each outcome (i.e.,
 damages per unit).  Recall  that the  probabilities  of each  strategy  being the correct
 choice are:

      o     probability  of strategy 1 being correct = .35
      o     probability  of strategy 2 being correct = .40
      o     probability  of strategy 3 being correct = .25

 The probability of choosing strategy 2 but having  strategy 1 being correct is  simply .35,
 the likelihood of strategy 1 being correct.  Similarly, the probability of choosing strategy
 2 when strategy 3 is actually correct is .25.

 Since the  values of the loss function for outcomes of damages per  unit between $0 and
 $200 million as well  as  between  $400 million and $600 million are equally likely, the
                                         3-15

-------
           Energy and Resource Consultants, Inc.

                         TABLE 3-6
             Loss Function if Strategy 2 is Chosen
                       But is Incorrect
  If Actual Damages                    Then the Penalty
($ x 106) Per Unit Are:                     ($ x 106) Is:
         $ 0                                $ 1,000
         50                                  750
         100                                  500
         150                                  250
      200 to 400                              0
         450                                  250
         500                                  500
         550                                  750
         600                                 1,000
                           3-16

-------
	  Energy and Resource Consultants, Inc.	

 value of perfect  information can  be calculated by using only the midpoints of the loss
 function segments for strategies 1  and 3.* The value of perfect information is then:

      (.35 x $500 million) + (.25 x $500 million) = $300 million

 This implies that, given the policy evaluation example detailed above, a decision maker
 would be willing  to pay $300 million per year to completely  resolve the uncertainty.
 Thus, for the relatively  large amount of uncertainty depicted in the cumulative probabil-
 ity distribution of Figure  3-3, the $300 million upper bound on  the value of information
 provides a benchmark value that  can be  compared to the costs of gathering additional
 information.   Another factor that affects  the value  of information in addition  to the
 probabilities is the range of damages considered plausible.  In this example, the range of
 damages is from $0 to $18 billion per year. The actual damages are assumed to fall with-
 in this range with certainty.  If this range is changed, the value of additional information
 would also be  changed.  Still, a range of  damages of $0 to $18 billion per year is quite
 broad.

 If the probabilities of each policy option being the correct choice change, then the value
 associated with additional information will change.  Two examples are:

      A.   Suppose information not previously available yields the following revised
           probabilities:
           o    probability (0 < damages/unit < $200 million) = .25
           o    probability ($200 million ^ damages/unit < $400 million) = .60
           o    probability ($400 million < damages/unit) = .15
           Then,  the value of perfect information is:
                (.25 x $500 million) + (.15 x $500 million) = $200 million
      B.   Suppose the revised probabilities  are:
           o    probability (0 ^ damages/unit <$200 million) = .15
 *Since  all outcomes for each segment of the loss function are equal likely, multiplying
 the  midpoint of  each segment times the cumulative probability of  having  the outcome
 occur in  that segment gives the  same answer as multiplying the marginal  probabilities
 times each potential outcome.
                                        3-17

-------
	 Energy and Resource Consultants,  Inc.	

            o    probability ($200 million ^damages/unit <$400 million = .80
            o    probability ($400 million < damages/unit) = .05
            Then, the value of perfect information is (.15 x $500 million) +  (.05 x
            $500 million) = $100 million

 These figures can also be used to develop rough estimates of the value of imperfect in-
 formation.  The value of a research program that would revise our original probabilities
 of (.35, .40, .25) to the probabilities (.25, .60, .15) used in example A can be calculated as
 the difference between the value of perfect information given the two sets of probabil-
 ities.  The value of perfect  information given  probabilities of (.35, .40, .25) is $300
 million per year and $200 million per  year for probabilities of (.25, .60, .15).  Thus, the
 value of research expected to decrease the uncertainty by revising the probabilities from
 (.35, .40, .25) to (.25, .60, .15)  is  $100  million. Similarly, the value of research expected
 to revise the probabilities from (.35, .40, .25) to (.15, .80, .05) is $200 million per year. If
 the outcome of the research is uncertain and, say, the  revised probabilities  are equally
 likely to turn out to be (.25, .60, .15) or (.15, .60, .05);  then, the estimated value of the
 research program yielding this imperfect information would  be (.5 x $200 million) + (.5 x
 $100 million) or $150 million.

 The simple example in this section not only  shows the types of insights that can be gen-
 erated by this approach, but through the use of plausible  numbers  provide some idea of
 the value of additional information which can be obtained.  Through this process, decision
 makers who may  have no idea of what values should be placed on additional research can
 obtain a rough estimate of what the value of additional research may be.

 3.2.2  The Timing of the Decision

 The large values  for additional information  calculated  in the preceding section indicate
 that performing additional research may be  beneficial,  but it is important to recognize
 that they say nothing about  which policy should be implemented while  the  research is
 being undertaken.  For example,  it cannot be  concluded on  the  basis of  a high value of
 additional information, that a  good strategy is to delay the implementation  of controls
 until more information  is collected.   To address this question the analysis needs to be
 extended an additional step.
                                         3-18

-------
                       Energy and Resource Consultants, Inc.
 The correct policy option to implement while research is being conducted is the policy

 that minimizes the expected discounted total costs of both pre- and post-research per-

 iods.  This is simply the selection of the combination control strategy and research plan

 that maximizes net benefits. To illustrate this,  the three policy options examined in the

 preceding section will be considered in combination with three research plans.


 The three research plans to be considered are:


      A:   To conduct no research.

      B:   To conduct a ten year research program that at its conclusion will yield
           perfect information, i.e.,  the  correct  control strategy will  be known.
           The cost of this research program is  assumed to be $15 million per year
           for a total of $150 million.

      C:   To conduct a stepped up five year research plan that will yield perfect
           information at its conclusion.  Due to  the tighter time constraints and
           resulting inefficiencies,  this research plan  is budgeted at $200 million,
           i.e., $40 million per year.

 The policy makers now have nine options:

      o    Choose strategy 1 - no control with:

                 la, no additional research
                 Ib, the ten year research  plan
                 Ic, the five year research plan.

      o    Choose strategy 2 - a 20% reduction in emissions with:

                 2a, no additional research
                 2b, the ten year research  plan
                 2c, the five year research plan.

      o    Choose strategy 3 - a 40% reduction in emissions with:

                 3a, no additional research
                 3b, the ten year research  plan
                 3c, the five year research plan.


Recall from  the previous section that the  probability of damages/unit of acid deposition

being between zero and $200 million per unit Is .35 (the low damage case).  The probabil-

ity  that damages per unit  are between $200 million and $400 million is .40 (the medium

damage case) and the  probability that damages are greater than $400 million is .25 (the

high damage case). Again, using a piecewise linear curve to approximate the continuous

cumulative probability distribution of Figure 3-3, only the midpoint values of these three
                                        3-19

-------
                       Energy and Resource Consultants, Inc.
cases and the associated probabilities are needed  to calculate the expected values for
each case.

The next step is to show how the streams of total costs will vary over time. Total costs
now include the money spent on research in addition to the costs of pollution control and
the social cost of the remaining, uncontrolled emissions, i.e., the damage resulting from
the acid deposition remaining after the controls have been implemented.

Since each of the nine policy options have three potential outcomes, there are twenty-
seven cases to be evaluated. For each case, the benefits and costs  for each year must be
calculated.  Tables 3-7a  through 3-7c show how the benefits and costs vary over time for
the choice of strategy 1 — no control — in conjunction with a five  year research plan.
Table 3-7a shows the stream of benefits and costs if the  five year research plan shows
that  the damages from  acid  deposition are low,  i.e.,  between 0  and $200 million per
unit.  Since no controls have been implemented, no change in strategy is called for at the
end of the five year research plan.  A discount rate of  five percent and a planning hori-
zon of thirty years are used to calculate the net present value of each stream.

Table 3-7b shows the  stream of benefits and  costs for strategy 1 with a  five year re-
search plan and  an  outcome of the research indicating moderate damages from acid dep-
osition (i.e., between $200 and $400 million per unit). Now a change  in strategy is called
for at the conclusion of the five year research plan.  It is now known with certainty that
strategy 2 — a 20% reduction in emissions — is the  best policy. In year 6, this strategy is
implemented incurring control costs of $1 billion per year and yielding benefits of $1.5
billion per year.

Table 3-7c shows the  stream of benefits and  costs for strategy 1 with a  five year re-
search plan and  an outcome of the research indicating high damages from acid deposition
(i.e., greater than $400 million per unit). Again, a change in strategy Is called for at the
conclusion of the  five year research plan.  Strategy 3 — a 40% reduction in emissions —
 The  selection  of discount rate and planning horizon can be  the  subject  of some
controversy. The recent Office of Management and Budget (OMB) guidelines for benefit-
cost  analysis call for a discount  rate of  10%.   However, the peculiar nature  of
environmental benefits probably calls for the use of a lower discount rate. The results
are less sensitive to the length of the planning horizon, provided it is sufficiently long.
                                        3-20

-------
TABLE 3-7a
Flow of Costs and Benefits for Strategy 1
with a Five Year Research Plan
— Outcome: Low Damages from Acid Deposition —



Year
1
2
3
4
5
6
7
8
9
V 10
£ 11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
NPVat
Year 1
(A)
Total Environmental
Damages Without
Any Control*
($ x 109)
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

46.1
•Assumes current deposition
(B)
Remaining Damages After
the Chosen Strategy Has
Been Implemented
($ x 109)
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

46.1
levels of 30 kg/ha/yr.

(C)
Benefits
(A-B)
($ x 109)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0



(D)
Control Costs
($ x 109)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0


(E)
Research
Costs.
($ x 109)
.04
.04
.04
.04
.04
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

.173


(F)
Total Costs
(B * D + E)
($ x 109)
3.04
3.04
3.04
3.04
3.04
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

46.27


(G)
Net Benefits
(C-D-E)
($ x 109)
-.04
- .04
-.04
-.04
-.04
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Q

-.173






^
ZJ
(V
— 1
(Q
x:
Ol
0
X
CD
0
c
n
CD
^"*
O
«/>
c
5)
H
•*
p















-------
TABLE 3-7b
Flow of Costs and Benefits for Strategy 1
with a Five Year Research Plan
— Outcome: Moderate Damages from Acid Deposition —



Year
1
2
3
4
5
6
7
8
9
10
w 11
^ 12
fo
M 13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
NPVat
Year 1
(A)
Total Environmental
Damages Without
Any Control*
($ x 109)
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9

138.4
*Assumes current deposition
CB)
Remaining Damages After
the Chosen Strategy Has
Been Implemented
($ x 109)
9
9
9
9
9
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5e

121.8
levels of 30 kg/ha/yr.

(0
Benefits
(A-B)
($ x 109)
0
0
0
0
0
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5

16.56



(D)
Control Costs
($ x 109)
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

11.0


(E)
Research
Costs.
($ x 109)
.04
.04
.04
.04
.04
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

.173


(F)
Total Costs
(B + D + E)
($ x 109)
9.04
9.04
9.04
9.04
9.04
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5
8.5

133


(G)
Net Benefits
(C - D - E)
($ x 109)
-.04
-.04
-.04
-.04
-.04
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5

5.4







rr
3
ft
CC
><
0)
3
Q
X
(I
CO
o
C
~T
o
fO
r
0
3
co
C
-^
cu
3
— »
v>

5"
o














-------
TABLE 3-7c
Flow of Costs and Benefits for Strategy 1
with a Five Year Research Plan
— Outcome: High Damages from Acid Deposition —



Year
1
2
3
4
5
6

7
8
9
u, 10
I 11
E 12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
NPVat
Year 1
CA)
Total Environmental
Damages Without
Any Control*
($ x 109)
15
15
15
15
15
15

15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15

230.6
*Assumes current deposition
(B)
Remaining Damages After
the Chosen Strategy Has
Been Implemented
($ x 109)
15
15
15
15
15
10

10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10

175.4
levels of 30 kg/ha/yr.

(C)
Benefits
(A-B)
($ x 109)
0
0
0
0
0
5

5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5

55.2



(D)
Control Costs
($ x 109)
0
0
0
0
0
3

3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

33.1


(E)
Research
Costs
($ x 109)
.04
.04
.04
.04
.04
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

.173


(F)
Total Costs
(B + D + E)
($ x 109)
15.04
15.04
15.04
15.04
15.04
13.0

13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0

208.7


(G)
Net Benefits
(C - D - E)
($ x 109)
-.04
-.04
-.04
-.04
-.04
2

2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2

21.9







n
a
(C
^
Q
3
Q
X
(I
C
C
n
(D
r
0
3
C
— *
Ql
a
VI
5
p














-------
	Energy and Resource Consultants, Inc.	

 is now the proven best choice.  In year 6, this strategy is implemented incurring control
 costs of $3 billion per year and benefits of $5 billion per year.

 Tables 3-8a through 3-8c show  the stream  of benefits and costs of choosing strategy 2
 with a five year research plan for the  three different outcomes. The other twenty-one
 cases can be calculated similarly. It is  apparent from  these tables that the out of pocket
 research costs play a relatively insignificant role in evaluating the appropriate timing of
 the decision.  They are dominated by the potential environmental damages that may oc-
 cur while the research is being done.

 The results  of the analysis  are  shown in Tables 3-9 and 3-10.  Table 3-9 shows the ex-
 pected total cost for  each strategy and research option.  Total cost is comprised of three
 items:  the environmental damages from the uncontrolled  deposition,  the costs of pollu-
 tion control and the costs of research.  The best policy is then the strategy and research
 plan that minimizes expected total cost. The policy with the lowest expected total cost
 is strategy 2 — a 20 percent reduction in emissions — in conjunction with a five year re-
 search program to determine the actual extent of damages.  The second best policy is to
 implement strategy 2 with  a 10 year research program.  The two  worst strategies are
 strategy 1 with no research to verify its appropriateness and strategy 3 also with no re-
 search. Thus, research  is valuable, but it  does not necessarily mean that  the  optimal
 action is to delay controls until better information is available.

 It is interesting that implementing strategy 2 with no  research to verify its appropriate-
 ness is still a better option than strategy 1 with a 10 year  research program.  This strat-
 egy would delay implementing controls until after a  10 year research program  is com-
 pleted.  Strategy 3 with no  research is preferred  to strategy 1 with research in  spite of
 the large values for additional information calculated in the preceding  section.

 Table 3-10 illustrates the same results, but it is probably clearer since it is presented in
 terms of net benefits.  The best option, strategy  2 with a five  year  research plan, has
 expected net benefits of $9.3 billion per year. This is $1.7 billion higher than the best
 delaying strategy, i.e., taking no action until the completion of a five  year research pro-
 gram.  In  this example, the risks of overcontrolling, i.e., implementing what turn out to
 be overly  stringent controls during the  research period, are outweighed by  the  risks of
 undercontrolling for acid deposition.
                                         3-24

-------
TABLE 3-8a
Flow of Costs and Benefits for Strategy 2
with a Five Year Research Plan
— Outcome: Low Damages from Acid Deposition —



Year
1
2
3
4
5
6
7
8
9
10
11
" 12
S3 ,,
ui 13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
NPVat
Year 1
(A)
Total Environmental
Damages Without
Any Control*
($ x 109)
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

46.1
•Assumes current deposition
(B)
Remaining Damages After
the Chosen Strategy Has
Been Implemented
($ x 109)
2.5
2.5
2.5
2.5
2.5
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

44.0
levels of 30 kg/ha/yr.

(0
Benefits
(A-B)
($ x 109)
.5
.5
.5
.5
.5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

2.15



(D)
Control Costs
($ x 109)
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

4.33


(E)
Research
Costs.
($ x 109)
.04
.04
.04
.04
.04
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

.173


(F)
Total Costs
(B + D + E)
($ x 109)
3.54
3.54
3.54
3.54
3.54
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

48.57


(G)
Net Benefits
(C - D - E)
($ x 109)
-.54
-.54
-.54
-.54
-.54
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

-2.3








n
a
(£
QJ

C
7
0!
I/)
0
C
r
(D
r
0
3
in
C
—
Qj

™<

-------
TABLE 3-8b
Flow of Costs and Benefits for Strategy 1
with a Five Year Research Plan
— Outcome: Moderate Damages from Acid Deposition —



Year
1
2
3
4
5
6
7
8
9
10
w 11
rl> 12
* 13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
NPVat
Year I
(A)
Total Environmental
Damages Without
Any Control*
($ x 109)
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9

138.4
*Assumes current deposition
(B)
Remaining Damages After
the Chosen Strategy Has
Been Implemented
($ x 109)
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5
7.5

115.3
levels of 30 kg/ha/yr.

(C)
Benefits
(A-B)
($ x 109)
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5

23.1



(D)
Control Costs
($ x 109)
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0

15.4


(E)
Research
Costs
($ x 109)
.04
.04
.04
.04
.04
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

.173


OF)
Total Costs
(B + D + E)
($ x 109)
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54
8.54

130.8


(G)
Net Benefits
(C - D - E)
($ x 109)
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46
.46

+ 7.51







IT
3
(1
(fl

cu
3
Q
*
in
O
C
n
O
f-
0
3
in
C
3|
^
~
3
p














-------
TABLE 3-Sc
Flow of Costs and Benefits for Strategy 1
with a Five Year Research Plan
— Outcome: High Damages from Acid Deposition —



Year
1
2
3
4
5
6
7
8
9
10
11
r 12
S 13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
NPVat
Year 1
CA)
Total Environmental
Damages Without
Any Control*
($ x 109)
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15

230.6
•Assumes current deposition


(B)
Remaining Damages After
the Chosen Strategy Has
Been Implemented
($ x 109)
12.5
12.5
12.5
12.5
12.5
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10

165.5
levels of 30 kg/ha/yr.


(0
Benefits
(A-B)
($ x 109)
2.5
2.5
2.5
2.5
2.5
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.0

66.0




(D)
Control Costs
($ x 109)
1
1
1
1
1
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

37.5



(E)
Research
Costs
($ x 109)
.04
.04
.04
.04
.04
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

.173



(F)
Total Costs
(B + D + E)
($ x 109)
13.54
13.54
13.54
13.54
13.54
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0
13.0

202.2

V

(G)
Net Benefits
(C - D - E)
($ x 109)
1.5
1.5
1.5
1.5
1.5
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0

28.4









IT
3
0)
CO
^<
0)
3
Q
X
0>
in
C
^

r
o
3
C
— H
Q>
3
^

p














-------
         Energy and Resource Consultants, Inc.
                        Table 3-9
Total Costs for the Alternative Strategies and Research Plans
                        ($ x 109)
Research Outcomes for the Level
of Acidic Deposition Damages
Low
Strategy 1 - no control with:
No research 46.1
10 year research plan 46.2
5 year research plan 46.3
Strategy 2 - 20% emissions reduction
No research 53.8
10 year research plan 50.1
5 year research plan 48.5
Strategy 3 - 40% emissions reduction
No research 76.9
10 year research plan 61.7
5 year research plan 55.0
Medium

138.4
134.7
133.0
with:
130.7
130.9
130.8
with:
138.4
134.6
133.0
High

230.6
215.4
208.7

207.6
203.8
202.2

199.8
199.9
200.0
Expected NPV

129.1
123.9
121.6

123.0
120.8
119.8

132.2
125.4
122.4
Rank

(8)
(6)
(3)

(5)
(2)
(1)

(9)
(7)
(4)
                          3-28

-------
	Energy and Resource Consultants, Inc.	

                         Table 3-10
 Net Benefits for the Alternative Strategies and Research Plans
                         ($ x 109)
Research Outcomes for the Level
of Acidic Deposition Damages

Low
Medium
High Expected NPV
Rank
Strategy 1 - no control with:
No research
10 year research plan
5 year research plan
Strategy 2 - 20% emissions
No research
10 year research plan
5 year research plan
Strategy 3 - 40% emissions
No research
10 year research plan
5 year research plan
0
-.12
-.17
reduction
-7.7
-3.97
-2.34
reduction
-30.7
-15.6
-8.7
0
3.71
5.4
with:
7.7
7.6
7.5
with:
-.1
3.7
5.3
0
15.2
21.9

23.1
26.8
28.4

30.8
30.6
30.6
0
5.2
7.6

6.2
8.3
9.28

-3.1
3.7
6.76
(8)
(6)
(3)

(5)
(2)
(1)

(9)
(7)
(4)
                           3-29

-------
                      Energy end Resource Consultants, Inc.
3.2.2.1  Irreversible Capital Investment in Pollution Control Equipment
One argument in favor of delaying decisions to implement environmental controls while
additional research is undertaken is based  on  the irreversible, fixed capital investment
that is required by many control strategies. If the control strategies are capital inten-
sive, implementing a  control strategy only to rescind it five years later may impose high
costs.  The  previous  example assumed perfectly flexible resources, i.e.,  if the research
demonstrated that damages  from acid deposition  were low and the currently required
level of control too stringent; then, the resources used to control pollution could be free-
ly transferred to other productive uses.  This assumption of perfectly flexible capital is
not likely to be the case.

The capital  investment in pollution control  equipment may not be readily transferable to
other uses.   However,  the variable costs (e.g., labor and raw materials) are no longer
incurred if  the  equipment is not used.   This  effect can be  captured in the  analytical
framework presented in the previous section. Flue gas scrubbers for the control of sulfur
emissions are one of the most capital intensive  pollution control technologies. Depending
on the removal levels, about one-third to one-half of the calculated annualized cost of
                                             4t
scrubbers is due to the capital  cost component.  Other pollution control strategies, such
as burning lower sulfur coal,  may have  a  considerably lower proportion of their costs
comprised of fixed capital costs.

To account for this capital inflexibility, it was assumed that 40% of the costs of pollution
control (i.e., the capital cost portion) continued even if the standards are  relaxed and the
control no longer needed.  The annualized control costs used as the basis for the cost of
                                                                     •JHfr
control estimates  used 15 years as the equipment life in the calculation.    Thus, if the
outcome of  the  research program indicated that the current standards were too stringent
and pollution controls were not needed, the capital costs are assumed to continue for the
full fifteen year period, i.e.,  until the capital costs are fully paid off.
 For  example, see:  A Review  of the Literature  Relevant to the Assessment of the
Impacts of  Acid Deposition Mitigation, prepared by Energy and Resource Consultants,
Inc.; and Pechan, E. Reducing Sulfur Oxide Emissions from the Electric Utility Industry,
both  prepared  for  the  Long Range Transport of Air Pollutants Assessment, Office of
Technology  Assessment, October  1981.
**See Pechan (1981).
                                        3-30

-------
	 Energy and Resource Consultants, Inc.	

 This is felt to liberally represent the potential costs from implementing a pollution con-
 trol program prior to conducting additional research and then having the research indi-
 cate that the controls were not necessary. This procedure assumes that there is no other
 use for the equipment, i.e., the equipment has a zero salvage value. This is an extreme
 assumption. Some of the equipment will undoubtedly have value in other uses.

 Although this approach accurately represents the opportunity costs  of the investment to
 society, it overstates the costs incurred by industry. If the pollution control equipment
 were no longer needed, private companies would write off the equipment as a tax loss.
 Thus, private companies would have this loss mitigated by reduced  taxes and would only
 have to absorb roughly 54 percent of  the cost of the capital equipment.  Although this
 tax reduction reduces the impact on private companies, the social costs are unaffected.
 The reduced tax revenue to government would have to be made up from another source.
 As a result, these tax factors were not considered in this analysis.

 Tables  3-11 and 3-12 show the calculations after incorporating the assumption of fixed
 capital investment.  The entries with asterisks are the only entries whose values  differ
 from the previous results shown in tables 3-9 and 3-10.  The only outcomes affected  by
 this additional assumption are outcomes for  scenarios where pollution controls are im-
 plemented prior to completing research and then the having the research show the con-
 trol levels to be too stringent.  As a result, only six outcomes are affected.

 The incorporation  of a fixed capital investment  assumption has little impact on the re-
 sults. In general, the same conclusions hold.  The two best strategies remain the imple-
 mentation of control strategy  2—moderate controls—with either a five-year or ten-year
 research plan.  The third best  strategy is to delay the implementation of pollution con-
 trols until an accelerated five-year research program is completed.  The two worst  strat-
 egies remain the  implementation of strategy 3—tight controls— and strategy 1—no con-
 trol— without conducting any research.  Again, it is interesting that the implementation
 of control strategy 2—moderate control— without conducting any  research is  superior to
 taking no action, i.e., strategy  1 until a ten-year research program is completed.
                                       3-31

-------
                      Energy and Resource Consultants, Inc.
                                    Table 3-11
             Total Costs for the Alternative Strategies and Research Plans
        Assuming Irreversible Capital Investment in Pollution Control Equipment
                                     ($ x 109)
Research Outcomes for the Level
of Acidic Deposition Damages
Low
Strategy 1 - no controls with:
No research 46.1
10 year research plan 46.2
5 year research plan 46.3
Strategy 2 - 20% emissions reduction
No research 53.8
10 year research plan 51.47*
5 year research plan 50.2*
Medium
138.4
134.7
133.0
with:
130.7
130.9
130.8
High
230.6
215.4
208.7
207.6
203.8
202.2
Expected NPV
129.1
123.9
121.6
123.0
121.3*
120.5*
Rank
(8)
(5)
(3)
(4)
(2)
(1)
Strategy 3 - 40% emmlssions reduction with:
No research 76.9
10 year research plan 65.7*
5 year research plan 60.7*
138.4
137.3*
136.4*
199.8
199.9
200.0
132.2
127.9*
125.6*
(9)
(7)
(6)*
* Values changed due to the fixed capital investment assumption.
                                        3-32

-------
                      Energy and Resource Consultants, Inc.
                                    Table 3-12
            Net Benefits for the Alternative Strategies and Research Plans
        Assuming Irreversible Capital Investment in Pollution Control Equipment
                                     ($ x 109)
Research Outcomes for the Level
of Acidic Deposition Damages

Low
Medium
High Expected NPV
Rank
Strategy 1 - no controls with:
No research
10 year research plan
5 year research plan
0
-.12
-.17
Strategy 2 - 20% emmissions reduction
No research
10 year research plan
5 year research plan
-7.7
-5.31*
-4.03*
Strategy 3 - 40% emmissions reduction
No research
10 year research plan
5 year research plan
-30.7
-18.8*
-16.1*
0
3.7
5.4
with:
7.7
7.6
7.5
with:
-.1
.98*
-.6*
0 0
15.2 5.2
21.9 7.6
23.1 6.26
26.8 7.9*
28.4 8.7*
30.8 -3.1
30.6 1.5*
30.6 1.8*
(8)
(5)*
(3)
(4)*
(2)
(1)
(9)
(7)
(6)*
* Outcomes with values changed due to the fixed Capital Investment assumption.
                                        3-33

-------
                    Energy and Resource Consultants, Inc.
3.2.2.2  Sensitivity of the Analysis to Different Probability Estimates

The preceding section presented an example that showed how one can evaluate alterna-
tive options relating to the timing of a decision to control acid deposition.  In particular,
the example examines a decision on whether to delay taking action while additional re-
search is  conducted.   Most of the parameters  used  in the example have some basis in
prior  research.  However, some of the parameters are purely speculative and are chosen
to facilitate the example.  Still, it is hoped that these speculative parameters fall within
a plausible range.  A critical set of speculative parameters are the probabilities  used in
the analysis.  These probabilities represent the likelihood that damages from acid
deposition are large enough  to  justify different control strategies and were depicted in
Figures 3-1  and 3-3.   No reliable estimates of the probability that damages fall within
these ranges were available  for this analysis.  Instead, the probabilities were chosen to
represent  considerable uncertainty regarding  the actual damages resulting from acid
deposition.  This section will examine how the appropriate policy changes as the probabi-
lities of different levels of environmental damage from acid deposition changes.

Although  this example is based on numbers that are  at present speculative, the calcula-
tions  can  help develop insights  on the appropriateness of alternative  strategies for con-
trolling acid deposition.  In particular, this approach can serve to depict the circum-
stances under which different control alternatives would  be  most appropriate.  This sec-
tion,  through sensitivity analysis, will  present  a broader discussion of the conditions
under which each  of  the different  alternatives  would be the most appropriate course of
action.

The preceding section ranked the control strategies according to their desirability given
the following probabilities:

      o     Probability of Outcome A — low damages:
                (O< damages/unit < $200 million) = .35
      o     Probability of Outcome B — medium damages:
                ($200 million ^damages/unit <. $400  million) = .40
      o     Probability of Outcome C — high damages:
                ($400 million^damages/unit = .25

With  these probabilities, costs of control and reductions in deposition used  in this exam-
ple, the best strategy was strategy 2 — a 20% reduction in emissions in conjunction with
                                        3-34.

-------
                      Energy and Resource Consultants, Inc.
an accelerated five-year research plan.   The second  best  option  was to implement

strategy 2 with a ten-year research program. The options of waiting to implement con-

trols until completion of either  a five-year or  ten-year research program were ranked

third and fifth respectively.  It is possible to calculate the set of the estimated probabi-

lities that would change these results.  In particular, it would be of interest to determine

what set  of  probabilities would result in  strategy  1—no control— being  the  correct

choice.


Strategy 1—no control—with a five-year research plan would  be preferred to the pre-

viously preferred  strategy 2—20% reduction in emissions—  with a five-year research plan
under the following circumstances:

     1.)  If the probability of Outcome C—high damages = .25 and;
           -  probability of Outcome A—low damages > .54
           -  probability of Outcome B—medium damages<.21

     2.)  If the probability of Outcome C—high damages = .20 and;
           -  probability of Outcome A—low damages >. 50
           -  probability of Outcome B—medium damages<.30

     3.)  If the probability of Outcome C—high damages —.10 and;
           -  probability of Outcome A—low damages>.43
           -  probability of Outcome B—medium damages  <.47


The probabilities  required to make strategy  1 with a five-year research plan preferable

to strategy 2 with a five-year research plan are  very dependent upon the likelihood of

high  damages occurring.  If there is a reasonable likelihood (i.e., > .25) that damages

from acid deposition are in the high range,  then the probability of damages being in the
low range must be roughly twice as high as the probability of moderate damages.


Strategy 1 with a ten-year research program is preferred to strategy 2 with a five-year
research plan under the following circumstances:


     1.)  If  the  probability of Outcome C—high  damages = .25; then, there is  no
          combination  of probabilities that result in strategy 1 with a ten-year
          research plan being preferred to strategy 2 with a five-year plan.

     2.)  If  the probability of Outcome C—high damages = .20 and;
           -  probability of Outcome A—low damages >.74
           -   probability of Outcome B—medium damages< .06

     3.)  If  the probability of Outcome C—high damage =  .10 and;
           -  probability of Outcome A—low damages >.62
           -   probability of Outcome B—medium damages <.28
                                       3-35

-------
                      Energy and Resource Consultants, Inc.
Strategy 1 with a ten-year research plan will be preferred to strategy 1 with a five-year
research plan under the following circumstances:

      1.)  If the probability of Outcome C—high damages >.0075, then
              strategy 1 with a five-year research plan is always preferred
      2.)  If the probability of Outcome C—high damages = .005 and;
           -  probability of Outcome A—low damages >.935
           -  probability of Outcome B-medium damages <.060
      3.)  If the probability of Outcome C—high damages = .001
           -  probability of Outcome A—low damages > .839
           -  probability of Outcome B—medium damages£.160

These calculations show that for strategy 1 with a ten-year research program to be pre-
ferred to strategy  2 with a five-year  research program, the probability of the actual
damages of acid deposition falling in the low range must be quite high, i.e., greater than
60%.   For strategy  1 with a ten-year research plan to be preferred to strategy 1 with a
five-year research plan, the probability of damages from acid deposition falling in the
low range must  be 84% or higher.  If the probability of damages falling into the high
range is greater than 1%, then strategy 1 with a five-year research plan will always be
preferred to strategy 1 with a ten-year research plan.

This form of  "WHAT  IF" scenario analysis can present problems with the appropriate
interpretation of the results, but it is useful for providing reference points for the deci-
sion parameters where none may otherwise exist.
3.3 Discussion

This section has demonstrated,  through the use of an example, a possible approach for
addressing what are felt to be some of the critical acid deposition policy issues: deter-
mining the dimensions of uncertain parameters, estimating the value of additional infor-
mation, and evaluating the appropriate timing of the decision.

The  estimates of  control costs  and the relationship  between changes  in emissions and
changes in deposition are comparable to other estimates found in the literature.  How-
ever, there are currently no reliable estimates of the probability that damages fall into
any of the ranges considered in this analysis. Instead, the probabilities chosen for this
                                        3-36

-------
                      Energy and Resource Consultants, Inc.
example were selected to represent a situation where there is considerable uncertainty.
The example is intended to provide estimates that may, to a limited extent, generate
insights into plausible policy options for reducing acid deposition. However, the primary
purpose is to demonstrate the usefulness of probabilistic damage estimates by presenting
examples of the method and types of analysis that can be performed.  The problem now is
to estimate the probabilities of the natural occurrence of various levels of damage from
acid deposition.

There were several simplifying assumptions used in the example  that deserve discussion.
First, the uncertainty that  is present in  the  relationship between changes in emissions
and changes in deposition was not considered.  This is a major source of uncertainty. Its
omission in this assessment allows for the use of a two dimensional graph (Figure 3-1) to
depict the implicit assumptions necessary for each  policy option to  be  the  correct
choice.   An assessment that would incorporate the uncertainty in atmospheric transport
as well as in the damages per  unit of deposition could use a three dimensional version of
Figure 3-3. Another approach would be to use a scenario analysis  with each scenario
representing a  different relationship between emissions and deposition. The probabilities
that the emissions to deposition  factor falls within these specific ranges would  then be
estimated.  These  improvements are straightforward, but some effort is required to
develop the necessary probabilistic estimates of the emissions to  deposition relationships.

A second simplifying assumption was the use of piecewise linear approximations to the
cumulative probability distribution depicted in Figure 3-3.  Since the probabilities used in
the example are speculative,  this assumption makes little difference and it greatly sim-
plifies the calculations.  A third assumption is that expected values alone are the proper
measure of benefits and costs.   Risk aversion of any  extent would  necessitate a more
complex analysis.

Some of the advantages of this type of probabilistic approach include the allowances for
the explicit incorporation of uncertainty  into the estimates and the specification of the
assumptions regarding the level  of damages from deposition that are required for  each
policy option to be the correct choice.

These analyses showed that,  given  the  data used in this  example  and the uncertainty
expressed by the probabilities,  the value of obtaining perfect  information that would
eliminate any uncertainty in the decision was estimated at $300 million per year.  This
                                        3-37

-------
                      Energy and Resource Consultants, Inc.
value serves as an upper bound to the value of information.  The value of imperfect in-
formation that would not eliminate all the uncertainty but would revise the probabilities
such that the policy maker's level of certainty regarding the  correct choice is increased
by 50 percent (i.e., increasing the probability of being correct from .4 to .6) was valued
at $100 million dollars per year.

Although large values for additional information were calculated, this only indicates that
research on the issue could prove valuable.  It says nothing about the  appropriate timing
of the decision to implement a control strategy.  For example, a large calculated value
for additional information does not necessarily imply that the correct policy choice is to
delay until  additional research is performed.  To assess this, three research plans with
different time frames and  costs were considered in conjunction with  the three policy
options.  The outcomes of the hypothesized research plans were very favorable In that it
was  assumed  that each  research  plan  would yield  perfect information, i.e., each plan
would reveal with certainty which strategy  is the correct choice.  Thus, in the examples
used, the policy maker faced a choice of delaying the implementation  of controls until
the completion of either a five year research program at a cost of 200 million dollars, or
a 10 year research program at a cost  of $150 million.  Even with the assumption that
each research plan  would yield perfect information at its conclusion, it was not the op-
timal strategy to delay  the Implementation of controls. The best choice over a wide
range of assumptions was to implement strategy 2 — a 20 percent reduction in emissions
— in conjunction with the five year research plan.  This option remained the best choice
even when the probabilities of each strategy being  correct were changed from .35 for
strategy I, M for strategy  2, and .25 for strategy 3  to equal probabilities for each (i.e.,
.33,  .33, .33); and even to a set of probabilities where strategy 1 was most likely to be
correct (i.e., .40, .35, .25).

Thus, this set of data, although speculative,  indicates that even in the presence of a large
amount of uncertainty and a research plan  guaranteed to yield perfect  information, the
best policy  option was not  to  delay  implementing controls.  Instead, the best policy in
this  example, was the implementation of a moderate level of control accompanied by a
research plan designed to determine, as quickly as possible, whether  the correct choice
has been made and  If the emissions reductions should be increased or relaxed.  These
results were relatively insensitive to small changes in the key parameters.
                                        3-38

-------
                       Energy and Resource Consultants, Inc.
There are two additional scenarios that could have important policy implications but that
were not addressed in this example.  The first is the possibility of irreversible environ-
mental damages and the second is the possibility of a higher upper bound on the magni-
tude of environmental damages. The consideration of either of these effects would make
it even  more desirable  to implement strategy 2 (moderate control)  or strategy 3  (tight
control) while research is being undertaken.

What is  now needed is research to provide better estimates of these parameters, particu-
larly the  probability that  damages  fall into different levels.  The next chapter will
address the sources of uncertainties in the damage estimates and methods for estimating
probabilistic damage functions.
                                       3-39

-------
                      Energy and Resource Consultants, Inc.
                 PROBABILISTIC DAMAGE FUNCTIONS AND ESTIMATES
The applicability of the approach outlined in Chapter 3.0 is dependent upon the ability to
estimate a probability distribution of damages  due to  acid deposition.  The probability
distribution used in  Chapter 3.0 is for aggregate  damages from acid deposition.  This
aggregate  probability  distribution  would be constructed using  a  bottom-up approach.
This chapter discusses how the probability distribution for aggregate damages from acid
deposition would be generated.
4.1 Elicitation of Probabilistic Damage Functions

The approach that would be used to develop the probability distribution for damages from
acid deposition is characterized as a bottom-up approach.  Damages from acid deposition
would be divided into many separate categories and probabilistic damage functions would
be estimated for each category.  These probability distributions would be combined to
generate an aggregate damage function. This approach would require the cooperation of
scientists conducting research on damages due to acid deposition.

During the course of the project, a number of probabilistic damage functions for differ-
ent species of trees and fish were developed. The principal  reason for developing these
damage distributions was to provide a basis for discussion of  the estimation methodology
with scientists currently conducting  research on damages.  This was undertaken to de-
termine the usefulness of the approach  in representing the uncertainty in the estimation
of damages from acid  deposition.  In general, this  procedure received support from the
scientists, and, as a result, a formal  elicitation of  probabilistic damage  functions seems
feasible and should yield  valuable information.

A probability elicitation exercise is a  session in which an interviewer queries an expert to
develop a consistent estimate of the frequency distribution of an uncertain quantity.  The
procedures for eliciting judgmental probability estimates have received substantial study
yet there is no  one accepted  method for  generating these probability distributions.  A
number of probability encoding techniques for use in policy analysis have been reviewed
In  Morgan et al. (1979).
                                      4-1

-------
                      Energy and Resource Consultants,  Inc.
Regardless of the particular elicitation technique, certain basic principles apply:

      1.    The person interviewed should have specific knowledge of the aspect of
           the environmental effect being assessed.

      2.    There should be an explicit  understanding between the elicitor and the
           respondent as to what  is  being assessed. For example, specifying that a
           discussion is about the impact of acid rain is not sufficient.  The elicita-
           tion  should be specific  regarding the parameters felt to  induce the
           effect  (e.g.,  H+  ion  concentration  or toxic  metals),  the  time frame
           under consideration, the  species of plant or fish, and the dimensions of
           the impact (e.g., reduction in biomass or population of fish).  The appro-
           priate specification of the  dimensions of  the damage is an important
           task.

      3.    Some respondents may be uncomfortable thinking in terms of frequency
           distributions.  They may tend to offer a single estimate and not devi-
           ate.  Their willingness to consider variations about the single  estimate
           may  be enhanced by  postulating  higher or  lower impacts  and asking
           them what circumstances could  result in these different impacts.  This
           broadens the discussion to more than their 'best guess' scenario.

      *.    The elicitor should not begin by asking for a best guess. This will focus
           the  discussion  on a single  value and  limit consideration  of  the ex-
           tremes.  Instead, the elicitor should first attempt to define the range of
           potential impacts.

      3.    The elicitor should explore the extreme values suggested by the respon-
           dent  by  postulating scenarios that would extend the limits or by having
           the respondent do this. The ultimate limits achieved in this fashion are
           more realistic bounds  than the ones  that will be  suggested initially  as
           they are more nearly free of the central tendency bias.

      6.    Graphical  tools to illustrate the concepts of probability may be useful in
           helping the respondent make tradeoffs between ranges of outcomes.

-------
                      Energy and Resource Consultants, Inc.
     7.    Care must be taken to assure that  the  opinions of the respondent are
           unbiased.  The possibility of bias should be investigated in a discussion
           at the beginning of the interview.
During the course of the project, probabilistic damage functions were estimated for five
species  of trees and eight species of fish. These damage functions were elicited from
only one or two scientists and should only be interpreted as examples of the procedure.
The purpose of the solicitation was to discuss the process with scientists and to ascertain
whether  they felt  it could yield useful results.  Figure 4-1  shows the outcome of  one
elicitation for the estimated effect of changes in pH on the population of rainbow trout.

The damage function measures the change in population of rainbow trout as the pH of a
lake declines. The range of possible  impacts is shown by the  dotted lines and the dashed
line shows the most likely estimate.  The damage function shows that at a pH of 7.0,
there is  no effect  from pH on the population of fish and that  this is known with cer-
tainty.   As the pH declines to 6.5, there is the possibility of an impact on the  rainbow
trout population.  The impacts range  from no effect (i.e., a zero change in population) to
a 10% decline in fish population, with the most likely impact being a 3% decline in fish
population. This slight impact would be due  to only the most susceptible fish being af-
fected.   As the  pH declines to 5.0, the  range  of possible  impacts  increases and the ex-
pected,  or most likely, impact also increases.  With further declines in pH, the range of
possible impacts tends to narrow since at an extremely low pH, there is a high degree of
certainty that the fish population will experience significant population decreases.  Then,
at a pH of 4.0, it is determined that rainbow trout cannot survive. There is a 100 percent
decline in population and this effect is known with certainty.

This probabilistic  damage function was constructed in the following manner.  The first
step was  to construct a table such as that shown in Table 4-1.  The  elicitor begins by
asking the expert if there is any possibility of a pH of 7.0 resulting in a reduction in the
rainbow trout fish population.  After  some discussion of the possibilities, the expert con-
cludes that there really is no chance  of a pH of 7 having an adverse impact on  the fish
population. As  a result, zeros are entered in the low, midpoint, and high columns of the
                                        4-3

-------
Energy and Resource Consultants, Inc.
           FIGURE «<-l

  Estimated Damage Function for
 Rainbow Trout  (Salmo gai rdneri)
                                            upper  Iimi t




c
O
4~f
H>
3
0.
O
a.
c
c
O
4J
u
3
X)

-------
                  Energy and Resource Consultants, Inc.
                               TABLE
                            Elicitation Table

                      Species of Fish: Rainbow Trout
                                     Range of Impacts
pH Level           Low         (Weight)      Midpoint     (Weight)     High

   7.0
   6.5
   6.0
   5.5
   5.0
   4.5
   4.0
                                  4-5

-------
	Energy and Resource Consultants, Inc.	

 table.  The next set of queries concerns the possibility of adverse impacts occurring at a
 pH of 6.5  At this pH level, the expert is uncertain as to whether adverse impacts may
 occur.  After some discussion of the scenarios and assumptions that could result in there
 being no impact or there being some adverse  impact on the fish population, the  expert
 concludes that the possible impacts range from  zero on the low end to 10% on the high
 end.  As a result, a 0 is entered in the low column and 10% is entered in the high column
 (See  Table ^-la).  These lower and upper bounds were chosen to represent roughly a 95
 percent confidence interval.  Thus, the lower  and upper bounds entered in the table do
 not determine the absolute limits to the range of possible damages but, instead, serve to
 define effective limits such that there is only  a  small probability that damages fall out-
 side  the range.  In this case, the  assumed 95 percent confidence interval implies that
 there is a 2.5 percent chance that the damages will be below the lower bound  and an
 equivalent probability that damages will be above the upper bound.

 Once the range of impacts for a given pH is specified in the above manner, the experts
 were queried regarding  the probabilities of the different outcomes within this range. It
 was quite difficult to elicit responses regarding the most likely damage outcome within
 the estimated range.  Typically, the experts were either unwilling to give an estimate of
 where the  most likely value would fall or  felt  that they did not have the information
 necessary to  develop an estimate of the most  likely impact.  In such circumstances, the
 usual procedure was to enter the midpoint in the table and then ask the expert a question
 similar to:  "If you were a betting man, would you bet that the actual adverse impact of a
 pH of 6.5 would fall above or below the midpoint?" Often, when  questioned in this way,
 the experts would express a strong conviction regarding where they would place a bet,
 even  when  they previously were  not able to provide estimates of where the most likely
 impact would fall.  Once the  expert decided  whether the adverse impacts on the fish
 population were likely to  be above or below the midpoint of the range, an "X" is  placed
 either above  or below the midpoint to express this weighting of the probabilities.  In
 Table 4-lb, the "X" placed between the low value of zero and the midpoint impact of 5%
 indicates that the expert felt that the most likely outcome for  the impact on fish popula-
 tion from a pH of 6.5 is between 0% and 5% rather than between 5% and 10%. The elici-
 tation procedure  is continued until estimates  for all the pH levels were developed (see
 Table 'f-lb). When an "X" was placed in the interval, it  was assumed that it was one third
 more likely that the actual adverse impact would  fall in the interval weighted  by  the
 "X".  Figure 
-------
                 Energy and Resource Consultants, Inc.
                              TABLE
                            Elicitation Table

                      Species of Fish: Rainbow Trout
                                     Range of Impacts
pH Level          Low         (Weight)      Midpoint     (Weight)     High

   7.0               0                          0                     0
   6.5               0                                               10%
   6.0
   5.5
   5.0
                                  4-7

-------
r
                                  Energy and Resource Consultants, Inc.

                                              TABLE «Mb
                                            Elicitation Table

                                      Species of Fish: Rainbow Trout
                                                     Range of Impacts
pH Level

7.0
6.5
6.0
5.5
5.0
4.5
4.0
Low

0
0
0
5 %
15 %
70 %
100 %
(Weight) Midpoint

0
"X" 5 %
10 %
27.5 %
52.5 %
85 %
100 %
(Weight)




"X"
"X"
"X"

High

0
10 %
20 %
50 %
90 %
100 %
100 %
                                                  4-8

-------
	 Energy and Resource Consultants,  Inc.	

 *.2  Discussion of the Elicitation Procedure

 The ellcitatlon procedure described above Is admittedly crude, but it is adequate to serve
 as a starting point for discussion of the viability of the general approach. In addition, if
 damage functions for many species of fish and vegetation, as well as materials damage
 and human health effects are to be estimated, increased sophistication in the estimation
 of each individual damage  relationship may not be warranted.  An adequate  representa-
 tion of the uncertainty surrounding estimates of aggregate  damages from acid deposition
 may not require a procedure any more sophisticated than that presented in the preceding
 section.

 There are several problems in the design of an elicitation process that require careful
 thought.  One important problem is the appropriate specification of the dimensions of the
 damage function.  The  preceding example used  pH as  the  independent  variable  (or
 causative agent) and percent  reduction in fish population  as  the dependent variable,
 however, there are other options.  For example, changes in the biomass of the fish popu-
 lation could have been used as a measure of the damage instead of  the  reduction in
 numbers of fish. In addition, pH is not the only index of water quality  that can be used to
 dimension the impact of acid  deposition on fish populations.  There are other important
 factors, such as concentrations of toxic metals — aluminum and mercury concentrations
 are of particular concern.  Often, low pH levels and high concentrations of toxic metals
 are closely correlated since rainfall with a  low pH tends to leach toxic metals from the
 soil.  Another problem  is the specific measure of pH used,  i.e., peak measurements or
 average pH.   The pH  of  a lake is not constant  and much  of  the scientific literature
 indicates that considerable damage results  from episodic events.  The spring snow melt
 or particularly heavy rains can cause short-term, but severe, depressions in pH.  These
 episodes  often occur in the spring and can be very  damaging to fish reproduction.  In
 some  cases, they can be so severe as to cause the elimination  of a fish  species from a
 lake with otherwise  high  pH levels.  Considerations  of this type can be folded into the
 elicitation of  the probabilistic  damage function presented  earlier by incorporating
 different scenarios into the range of possible outcomes.

 The incorporation of these additional factors into the damage function complicates the
 interpretation of relationship.  For example, a damage function expressed in terms of pH
 and percent reduction in fish  population may, in fact, be representing a  wider range of
 variables and a  more complex  relationship.  Also, this makes estimation of the range of
 outcomes and probabilities even more difficult for the experts.

-------
                      Energy and Resource Consultants, Inc.
The preceding discussion indicates that the process used to generate subjective probabili-
ties can be simply stated but, in actual applications, a number of complications invari-
ably arise. In spite of these problems, the approach has many advantages.  These include:

     o     Environmental policy decisions, as well as most other regulatory deci-
           sions, will  always  be subject to uncertainty.   A dimensioning of this
           uncertainty by technical experts will provide additional information that
           has not been available to policy makers in the past.  Since any decision
           requires, at least, an implicit assessment of the uncertainty, the alter-
           native is to require the  often less-knowledgeable policy makers to form
           their own opinion of the probabilities without assistance from the scien-
           tific community.

The following advantages are taken from Morgan et al. (1979):

     o     The approach may  contribute to "better"  decisions.  Since there is con-
           siderable evidence  that people are poor statistical processors, a formal
           analytical technique may  help to avoid  some of the pitfalls of a "seat of
           the pants" approach.

     o     Results are  obtained in a quantitative  form which can be easily incor-
           porated into subsequent analyses.

     o     Results explicitly incorporate a statement of the uncertainty associated
           with the knowledge. .  .  something which has been  all too frequently
           ignored in more of the previous regulatory decision-making.

     o     Results can be obtained at a fairly low cost.

The use of subjective expert judgments has substantial benefits  but the use of  this ap-
proach is  not without pitfalls.  In their review of  the use of subjective probability esti-
mates, Morgan et al. (1979)  cite the following concerns:

     o     People may have an incorrect or incomplete perception of how "good" a
           job  expert  can do  at making subjective  quantitative assessments, and
           thus may be misled by the results.
                                        f-10

-------
                       Energy and Resource Consultants, Inc.
     o     People who obtain such  assessments from experts may not adequately
           use current understanding of the problems in this field to obtain the best
           possible results.

     o     Use of quantitative subjective judgements in regulatory decision-making
           without a proper understanding of their limitations could lead to subse-
           quent litigation and legal decisions which might significantly limit a
           regulator's ability to employ such techniques in the future.

     o     Because they are relatively cheap, produce quantitative "technical look-
           ing"  results,  and because in some fields (e.g., toxic substances) regu-
           lators are  faced  with overwhelming data needs, there is a real  danger
           that the elicitation of expert subjective  judgements may begin to be-
           come a substitute for doing needed research.

After consideration of the advantages and disadvantages, Morgan  et al. (1979) conclude
by sayings  "To our mind, these potential problems do not outweigh the substantial bene-
fits that can result from using subjective expert probabilistic judgement in policy anal-
ysis."  The arguments and analysis  presented in this report support this conclusion, but
they also reinforce the importance of careful application of decision analysis techniques
and an understanding of the limitations of the approaches.
4.3 A Sample Application; Damages to Forests from Acid Depositions

An example of how the elicitation process outlined in Section 4.1 can be used to develop
probabilistic estimates of damages is presented in this section.  The steps that must be
performed to develop probabilistic  damage estimates are presented in Table 4-2.  An
elicitation for five species of trees was conducted.  The results from this elicitation are
subject  to several important limitations.  The primary limitation is that only two indivi-
duals were used in the elicitation. As a result, the full range of views and research out-
comes may not be represented.  It is important that not too much significance be placed
on these early elicitations.   An elicitation encompassing a larger number of experts and
the inclusion of a broader range of  tree species  and scenarios in the elicitation process
could result in different estimates of the  range and probabilities of damages.  The pur-
pose of  this section is to serve as an  example of the method.

-------
                 Energy and Resource Consultants, Inc.
                               TABLE

                Damages to Forests from Acid Deposition:

                          Steps in the Analysis
Step 1.    Estimate Probabilistic Damage Functions.

           1.1    Obtain the subjective judgements required to fill in the elicita-
                 tion table (i.e., Table 4-1).

           1.2    Translate the elicitation table into probabilistic damage func-
                 tions.
Step 2.    Develop Baseline Estimates of Damages.

           2.1    Obtain estimates of the current levels of acid deposition that
                 tree species are exposed to.

           2.2    Calculate the level of damages that are occurring at current de-
                 position levels using the estimated damage functions.


Step 3.    Estimate how the forest damages will change under the different policy
           options being considered.

           3.1    Develop estimates  of how the exposure of each  tree species to
                 acid deposition will change.

           3.2    Use the probabilistic damage functions to estimate the new level
                 of forest damages that will result under each policy option.

           3.3    Subtract the levels of damages occurring after implementation
                 of the policy from the baseline level of damages to calculate the
                 estimated improvement  or benefit

-------
                      Energy and Resource Consultants, Inc.
Probabilistic damage functions were developed for five species of trees — two softwood
species  (spruce  and balsam fir) and three  hardwood  species (yellow  birch, maple  and
beech).  Table 4-3 shows the outcome of the elicitation of the effects of acid deposition
on red spruce.  Table 4-3 shows the range of damages to the red spruce population that
                                                                      .£
could be expected to occur, given an exposure to an annual average rain pH.

Table 4-3  is interpreted in the same manner as the previous elicitation example present-
ed in Table 4-1.  The "X" entered in the "weight" columns on Table 4-3 are used to indi-
cate  that portion of the range of damages the experts felt included the most  likely
damage outcome.  The elicitation table is used to develop a probability distribution of
damages for each pH exposure level.

Figure 4-2 shows the cumulative probability distribution of damages to red spruce asso-
ciated with a rainfall pH of  4.5  This was constructed from Table 4-3 using the following
assumptions:

      o     The range of damages, i.e., a 15% to 80% reduction in population due to
           a rainfall pH  of 4.5, represents a  95% confidence interval.  That  is,
           there is a 2.5% probability that  the actual reduction in population may
           be below 15% and the probability that damages exceed 80% is also 2.5%

      o     Actual damages are one third more likely to fall on the side of the mid-
           point deemed most likely, i.e., the side marked with the "X".
The cumulative  probability distribution indicates the following probabilities for the re-
duction in red spruce biomass due to exposure to rainfall with a pH of 4.5:

            Prob. (Of % population reduction <. 15%) = .025
            Prob. (15%£% population reductions47.5%) = .57
            Prob. (47.5%f % population reduction < 97.5%) = .38
            Prob. (97.5%£% population reduction <. 100%) = .025
  In order for the pH to have an adverse effect on red spruce, presented in Table 4-3, the
reduction in pH would have to last for several years.
                                       4-13

-------
                           Energy and Resource Consultants, Inc.
                                      Figure 4-2


             Cumulative Probability Distribution for Damage to Red Spruce

                 Exposed  to Rainfall with an Annual Average pH of 4.5
 Cumulative
 Probability
,975  -  -  -  -  -



          .90 •




          .80 .




          .70 .




          .60 .
,595
          .50 -
          .40 -
           30 -
          .20 _
           10 -
.025	
            0
                                          47.5
                                   % Reduction in Biomass

-------
                     Energy and Resource Consultants, Inc.
                                  TABLE 4-3
             Percent Decline in Biomass for Red Spruce Due to Acid Rain
Indicator Variable        Lower Bound                                        Upper Bound
     Rain pH              of Damage     (Weight)      Midpoint     (Weight)     of Damage
     5.5                    0                           00
     5.                     0                           00
     4.5                   15            X*             47.5                     80
     4.                    30             X             60                       90
     3.5                   60             X             80                      100
     3.                    60                           80          X           100
*The "X" indicates which side of the midpoint the expert felt actual damages were most likely
to occur.
                                     4-15

-------
                      Energy and Resource Consultants,  Inc.
Figure 4-3 shows the marginal probability distribution corresponding to the cumulative distribu-
tion shown in Figure 4-2.

The elicitation  table  shown in Table  4-2 is  used to generate a  probability  distribution  of
damages for trees exposed to rainfall of a given pH.  The next step in constructing the proba-
bility distribution for baseline (i.e., current) damages from acid rain is to determine the rain pH
red spruce trees are currently exposed to.  This was done by overlapping maps of forest popula-
tions  with maps showing average rain pH.  Only the New  England region and the states of New
York  and Pennsylvania were considered in this analysis.  Table 4-4 shows the estimates of the
population of the five treo species exposed to different levels of rain pH.

This exposure information can be combined with the estimates of damages at the different pH
exposure levels to generate  an estimate of current damages from acid rain. This is done separ-
ately for each pH range using the midpoint of the pH range to generate the probability distribu-
tion of damages for each pH level.  The aggregate distribution of damages for all five species of
trees is calculated in three steps:

      (1)   The distribution of damages for each species for each pH exposure cate-
           gory are calculated.

      (2)   The distribution  of total damages for each  species of tree is calculated
           by combining the distributions across the  pH categories  for the same
           tree species.

      (3)   The distribution  of total damages for all species of trees is calculated
           by combining the distributions for each species.

The frequency distributions were combined using a Monte-Carlo simulation.  A critical
assumption in this calculation is that the  damage functions are independent.   That is,
that the damage done to one species at a given pH is not  influenced by the damage done
to another species at the same pH.  Uniform biases in  the damage function estimates or
biological  factors such as the damage to one species being offset by reduced competition
                                       4-16

-------
                            Energy and Resource Consultants, Inc.
                                         Figure 4-3

           Marginal Probability Distribution for Damages to Red Spruce  Exposed  to
                         Rainfall with an Annual Average of pH of  4.5
Probability
in Percent
            2.0 -
            1.8 -
    1.75	
             1.6 -
             1.2 -
    1.17	
             1.0-
              .8-
    .16 -
    .125 -
              .2-
                                  30
                                             i
60
    40   , 50
       47.5
% Reduction in Biomass
70    80    90
100

-------
                      Energy and Resource Consultants, Inc.
                                   TABLE
              Net Volume of Growing Stock on Commercial Timberland
                    in New England, New York and Pennsylvania
                               (million cubic feet)
Current Rain
pH Levels
(Exposures)
pH 4.5-4.4
pH 4.4-4.3
pH 4.4-4.2
pH 4.2

Spruce
1959
5035
266
neg.

Balsam
Fir
1959
5035
266
neg.
Tree Species
Yellow
Birch
246
1398
409
195

Maple
852
6965
5466
3510

Beech
221
1114
1053
712
Sources;
     1.   Timber materials from:  An Analysis of the Timber Situation in the United
          States 1952-2030, United States Department of Agriculture, Forest Service.
     2.   pH levels  from  maps  in  Memorandum of Intent on  Transboundary Air
          Pollution, Phase 2, Interim Report, Working Group 2, July 1981.
                                      4-18

-------
	 Energy and Resource Consultants, Inc.	

 from other damaged  species could invalidate the assumption of  independence.    This
 would widen the  distribution of combined damages  in  the case of estimation bias, or
 narrow it in the case of offsetting biological factors.

 In surveying the biological literature, it has become clear that damages may not be inde-
 pendent. The extent of the interdependence is not clear at this time as more research is
 needed. It would  certainly be possible  to include elicited views on issues such as species
 replacement in the decision analytic framework.

 The estimates of the baseline damages from acid deposition are presented for softwoods
 (spruce and balsam fir) and hardwoods (yellow birch,  maple and beech). A value of $1.60
 per cubic foot was used for the stumpage price for softwoods and $.60 per cubic foot was
                                       M M
 used as the stumpage price for hardwood.

 The cumulative probability distribution for damages  to  both hardwoods and softwoods Is
 presented in Figure 4-4. The estimated range of damages from acid deposition at current
 deposition levels range from 520 million to 1,480 million.  The estimated  median damage
 level is $1,280 million.  Given this cumulative probability distribution, the probabilities
 for damages from current levels of acid rain falling in selected ranges is:

      o    Prob (0 
-------
                                                    FIGURE  4-4
.c-
 I
ro
              ati ve

          Probability
1.00




 .90




 .80




 .70




 .60




 .50




 .40




 .30




 .20




 .10




   0
                                   Baseline Distribution of Damages to Both Hardwoods and

                                    Softwoods in New England, New York, and Pennsylvania
                                                                                                                           m
                                                                                                                           (O
                                                                                                                           Q.
                                                                                                  o
                                                                                                  c

                                                                                                  n
                                                                                                  n

                                                                                                  n
                                                                                                  o
                                                                                                  3
                                                                                                  CO
                                                                                                  C_

                                                                                                  oT
                                                                                                                           y>


                                                                                                                           5"
                  50
60
                      70
80
90
100
110
120
130
—i—

 140
150
                                                                                           160
                                                    1 x 10' $

-------
                      Energy and Resource Consultants, Inc.
This particular elicitation showed relatively high probabilities of there being substantial
damage to forest from acid precipitation.  In particular, the median damage estimate for
the area comprised  of New York State, Pennsylvania and the New England Region was
calculated to be $1.28 billion.  The probability that damages fall  between $1.28 billion
and $1.41 billion is estimated at 40%, and the probability the damages are less than .7
billion is  10 percent.    There  are  several  possible  reasons for these high  damage
estimates.  First, the area under consideration has  the most severe acid deposition and
the vast  majority of damages to forests can be expected to occur in this region. Other
possible reasons for these high damage estimates stem from potential biases in the elici-
tation and application of the damage functions.  Only two experts were used in the elici-
tation and it is possible that the full range of research results on  damages are not ade-
quately represented.  Another possible bias, and potentially the most serious is that if a
specific species of tree is shown to have a growth reduction due to acid precipitation no
other commercially valuable trees are assumed to replace it. For example, if acid depo-
sition  reduces the stock of red spruce, the stock of balsam fir trees (a more  resistant
species) may increase due to reduced competition among the species. The commercial
value of  the replacement trees  Is not considered.  This would tend to overestimate  the
damages  due to acid deposition.

The  next step in the analysis is evaluating how damages will change as anthropogenic
emissions change. This poses a difficult problem.   It is not as clear how the pH of  the
rain  will change as  emissions  change.   Naturally occurring  rain is commonly cited as
having a  pH of 5.6, although this can vary  widely at different locations. In this analysis,
the assumption is made that the naturally occurring pH of rain is 5.6. If the pH of  the
rain  is less than 5.6, then this reduced pH  is due entirely  to anthropogenic sources.  The
H+ ion concentration that results in a pH of 5.6 is 2.51 micro-equivalents  per liter; if  the
observed pH of the  rain is 4.4  the H* ion concentration is 39.8 micro-equivalents  per
liter.  The difference between the two, i.e., 37.3 micro-equivalents per liter, is assumed
to be the contribution from anthropogenic sources.

Three  additional  scenarios are evaluated to determine the effect of different policy  op-
tions on the damages from  acid deposition.  The scenarios are:

      o    a 15 percent increase in H* from anthropogenic sources

      o    a 25 percent decrease in H+ from anthropogenic sources
                                       4-21

-------
                      Energy and Resource Consultants, Inc.
     o     a 50 percent decrease in H* from anthropogenic sources

By considering only changes in the H* ion concentration of the rain, it was possible to
calculate the new pH levels the forests would be exposed to.  The new exposures for the
different species of trees for each of the three scenarios are shown in Tables 4-5 to 4-7.
For example,  using  Table 4-4 as the baseline and examining the first row, Table 4-5
shows that a 15 percent increase in H+ ions from anthropogenic sources would decrease
the pH of the rain  from 4.5 to 4.44. Similarly, if the pH of the current rainfall is 4.5,
then a 25 percent reduction in H+ ions from anthropogenic sources would increase the pH
to 4.61.  A 50 percent reduction in  H+  ions from anthropogenic  sources would further
increase  the pH to  4.77.  These results are used to generate tree species exposure esti-
mates for each scenario (see Tables 4-5, 4-6, and 4-7).

The cumulative probability distributions for damages from acid rain for each of the three
scenarios are shown in Table 4-8  and in Figures 4-5, 4-6 and 4-7.  A 15  percent increase
in anthropogenic  Induced H* ions increased the estimated median value  of damages from
$1,279 million to $1,378 million.  A 25  percent  reduction  in H*  ions reduced the esti-
mated  median value of damages from $1,279 million to $1,079 million. Therefore, the
benefits of achieving a 25 percent reduction in H+ ions from anthropogenic sources would
have an estimated median benefit, in terms of reduced forest damages, of $200 million.
The range of benefits from  a 25 percent  reduction is from $96 million to $220 million.  A
50 percent  reduction in H* ions  reduced the  estimated median value of damages  from
$1,279 million to $780 million.  The estimated median benefits from a 50 percent reduc-
tion in H*  ions are $499 million.  The range of benefits from a 50 percent reduction  is
from $216 million to $573 million.

Given the cumulative probability  distributions shown in Table 4-8 and Figures 4-5, 4-6
and 4-7,  it  is possible to calculate the probability that benefits from a strategy  to reduce
H+  ion  concentrations  in rainfall are equal to  or  exceed  a certain  value  for  each
scenario.  For example, if the costs of achieving a 25 percent reduction in H+  ions  from
anthropogenic sources were estimated at  $150 million,  the probability of benefits, in
terms of reduced forest damages, exceeding $150 million is approximately 70 percent.
This is found by subtracting the total damages distribution in Table 4-8 with a 25 percent
reduction in anthropogenic emissions from the baseline distribution of  damages.
                                       4-22

-------
                    Energy and Resource Consultants, Inc.
                                   TABLE
              Net Volume of Growing Stock on Commercial Timberland
                    in New England, New York and Pennsylvania:
          Exposures to Rainfall pH given a 15% Increase in Anthropogenic H+
                                (million cubic feet)
15% Increase in
Anthropogenic H+
pH Levels
(Exposures)
pH 4.44-4 .34
pH 4.34-4.24
pH 4.24-4. 14
pH 4.14


Spruce
1959
5035
266
neg.

Balsam
Fir
1959
5035
266
neg.
Tree Species
Yellow
Birch
246
1398
409
195


Maple
852
6965
5466
3510


Beech
221
1114
1053
712
Sources:
     1.   Timber materials from:  An Analysis of the Timber Situation in the United
          States 1952-2030, United States Department of Agriculture, Forest Service.
     2.   pH levels  from  maps  in Memorandum of  Intent on  Transboundary Air
          Pollution, Phase 2, Interim Report, Working Group 2, July 1981.
                                      4-23

-------
                     Energy and Resource Consultants, Inc.
                                   TABLE
              Net Volume of Growing Stock on Commercial Timberland
                    in New England, New York and Pennsylvania:
         Exposures to Rainfall pH given a 25% Reduction in Anthropogenic H+
                                (million cubic feet)
25% Reduction in
Anthropogenic H+
pH Levels
(Exposures)
pH 4.61-4.51
pH 4.51-4.42
pH 4.42-4 .32
pH 4.32


Spruce
1959
5035
266
neg.


Balsam
Fir
1959
5035
266
neg.

Tree Species
Yellow
Birch
246
1398
409
195


Maple
852
6965
5466
3510


Beech
221
1114
1053
712
Sources:
     1.   Timber  materials from:  An Analysis of the Timber Situation in the United
          States 1952-2030, United States Department of Agriculture, Forest Service.
     2.   pH levels  from  maps  in  Memorandum of Intent on  Transboundary Air
          Pollution, Phase 2, Interim Report, Working Group 2, July 1981.
                                    4-24

-------
                    Energy and Resource Consultants, Inc.
                                   TABLE
              Net Volume of Growing Stock on Commercial Timberland
                    in New England, New York and Pennsylvania:
         Exposures to Rainfall pH given a 50% Reduction in Anthropogenic H+
                                (million cubic feet)
50% Reduction in
Anthropogenic H*
pH Levels
(Exposures)
pH 4 .77-4.67
pH 4.67-4.58
pH 4.58-4.48
pH 4.48


Spruce
1959
5035
266
neg.


Balsam
Fir
1959
5035
266
neg.

Tree Species
Yellow
Birch
246
1398
409
195


Maple
852
6965
5466
3510


Beech
221
1114
1053
712
Sources:
     1.   Timber  materials from:  An Analysis of the Timber Situation in the United
          States 1952-2030, United States Department of Agriculture, Forest Service.
     2.   pH levels  from  maps  in  Memorandum of Intent on  Transboundary Air
          Pollution, Phase 2, Interim Report, Working Group 2, July 1981.
                                    4-25

-------
                          Energy and Resource Consultants, Inc.

                                        TABLE 4-8

                           Cumulative Probability Distributions of
                           Forest Damages for the Four Scenarios
Cumulative
Probability
    Baseline
Current Damages

    $ x 107)
  15% Increase
in Anthropogenic
  H+ ions from
    ($ x 107)
 25% Decrease
in Anthropogenic
    H+ ions
    ($ x 107)
 50% Decrease
in Anthropogenic
    H+ ions
    ($ x 107)
   .025
   .10
   .20
   .30
   .40
   .50
   .60
   .70
   .80
   .90
   .975
      51.9
      69.6
      83.9
      98.3
      113.2
      127.9
      129.9
      136.4
      138.4
      140.4
      148.8
      57.2
      76.2
      91.0
     106.3
     122.2
     137.7
     139.7
     145.9
     147.9
     149.8
     158.1
      42.2
      58.3
      70.4
      82.6
      95.4
     107.9
     109.7
     116.2
     118.0
     119.8
     126.6
     30.2
     42.7
     51.3
     60.1
     69.1
     77.9
     79.1
     84.3
     85.4
     86.6
     91.3
                                           4-26

-------
.e-
 t
ISJ
-J
                       ve
             Profcabi1i ty
1.00



 .90



 .80



 .70



 .60



 .50



 .40



 .30



 .20



 .10





            50
                                                                                               m

                                                                                                 o
                                                                                               m -••
                                                                                               3 «/>
                                                                                              <£3 ft-

                                                                                               O) -i.
                                                                                               3 O"
                                                                                            O Q- C
                                                                                            O •  r*
                                                                                            3   -••
                                                                                            O Z O
                                                                                            **3

                                                                                            rf   O
                                                                                            -I -< -h
                                                                                            Q> O
                                                                                            c^ n o
                                                                                            -"• ?T 0»
                                                                                            o«3
                                                                                            3   &>
                                                                                               O> U3
                                                                                            o 3 n>
                                                                                            -+i Q. l/>

                                                                                            S» -D r+    -n
                                                                                            3 ro O    >—
                                                                                            rf 3       O
                                                                                            3" 3 03    C
                                                                                            -J l/> O    ?O
                                                                                            O «< r*    m
                                                                                            o <
                                                                                            10 0»  Z
                                                                                            (0 3  a*
                                                                                            3 -^ T
                                                                                            _.. QJ  Q.
                                                                                                                                                 <_n
                                 60
70
80
90      100     110      120     130      140      150     160
 i -•• o
«< rl- Q.
 Q. 3- «/>
 -J
 O O* O>
«£>    3
 fD —• Q.
 3 tn
   fr« (/)
 «-«    O
 O >—' -h
 3 3 rf
 1/1 n X
   i o
   fD O
   01 O.

   n

   -•• 3
                                                                                                                                                                (V

                                                                                                                                                               (O
                                                                                                                                                                01

                                                                                                                                                                Q.

                                                                                                                                                                X)
                                                                                                                                                                It
                                                                                                                                                                to
                                                                                                                                                                O
                                                                                                             n
                                                                                                             o
                                                                                                                                                                3
                                                                    1  x 10'  $
                                                                                               n

-------
-t-
I
f-0
co
         Cumulative
         Probabi1ity
1.00



 .90



 .80



 .70



 .60



 .50



 .40



 .30



 .20



 .10




   0
                                            FIGURE 4-6

                         Distribution of Damages to  Both Hardwoods  and Softwoods
                    in New England,  New York, and Pennsylvania with  a 25% Reduction
                            the Concentration of Anthropogenic Hydrogen Ions
                                                                                         in
                                                                                re

                                                                                co


                                                                                O>

                                                                                Q.

                                                                                JO

                                                                                «/»
                                                                                O
                                                                                c

                                                                                o


                                                                                n
                                                                                o

                                                                                to


                                                                                tt

                                                                                
                                                                                                                  o
                   40
                  50
—i—

 60
70
80
90
100
110
120
—i—

 130
                                                  1  x  10' $

-------
             Cumulative
             Probability
IS)

U3
                                                                                                                                          ro
                                                                                                                                             10
                                                                                                                                             r*-
                                                                                                                                        3- —' -1
                                                                                                                                        ro a> ->•
                                                                                                                                          3 o-
                                                                                                                                        c~> a. c
                                                                                                                                        o -  c+
                                                                                                                                        3    -"•
                                                                                                                                        o z o
                                                                                                                                        CD ro 3
                                                                                                                                        3 s:
                                                                                                                                        rf   O
                                                                                                                                        -5 •< -*i
                                                                                                                                        JU O
                                                                                                                                        <-*•-!<=)
                                                                                                                                        ->• 7T JD
                                                                                                                                        O ••  3
                                                                                                                                        3    CJ
                                                                                                                                          Qi 03
                                                                                                                                        o 3 ro
                                                                                                                                        -h Q. 
 3 ro  o

 3- 3  ro
 -j   o
 O <<  c-t-
T3 —• 3"
 O <
CO D*  "T"
 ro 3  oj
 3 -•• -J
 _.. Q»  Q.
 O    «:
   s  o
 31 -.. O
<< rh  CL
 0. 3- (/>

 O CU  O)
CO    3
 ro tn  Q.
 3 o

 •-.    o
 O X)  -h
 3 ro  c+
 to Q.  S
   c  o
   0  O
   a-  a.
                                                                                                                                                  m
n>

(O
><
a*

Q.
 o
 c
 r>
 (D

 O
 O
 3
                                                                                                                                                            3
                                                                                                                                                            o
                                                                    1   x  TO7  $

-------
                      Energy and Resource Consultants, Inc.
 *.* Summary
This section  demonstrated how the probabilistic damage functions can  be derived and
how they can be used to generate estimates of damages. It remains to be  seen how sensi-
tive policy decisions are to the distributions and hence, the degree of resolution needed
for dependable decision  making.  ERC has not pushed the limits on elicitation in this
study,  but there  is every indication that  despite the  substantial  uncertainty about
damages, the basic understanding is there to proceed with a decision analytic formulation
of acid deposition policy decisions.

Probabilistic  damage functions for a number of fish species were also estimated but the
lack of information on fish populations as well as a lack of information on the number of
lakes at different  pH levels prevented the presentation of a similar example for aquatic
damages.
                                       4-30

-------
                     Energy and Resource Consultants, Inc.
                               5,0 BIBLIOGRAPHY

Battelle  Laboratories, 1980.  Research  Report on  Acid Precipitation,  for  American
     Electric Power Service Corporation, September, 1980.

Conrad,  J. M., 1980.  "Quasi-Option Value and the Expected Value of Information."
     Quantity Journal of Economics, June, 1980, p. 813-820.

Crocker, T. D., J. T. Tschirhart, R.M. Adams, and B. Forster, 1981.  Methods Develop-
     ment for Assessing Acid  Deposition  Control  Benefits, prepared  for  Office  of
     Research and Development, U.S.E.P.A., Grant //R806972010, January, 1981.

Edison Electric Institute, 1981.  Clean Air Act Issue Paper on Acid Deposition, March 3,
     1981.

Hirshleifer, J. and J. Riley, 1979.  "Analytics of Uncertainty and Information",  Journal of
     Economic Literature, December, 1979, pp. 1375-1421.

Keeney,  R. L. and H. Raiffa, 1976.  Decisions with Multiple Objectives, John Wiley and
     Sons, Publishing, 1976.

Mclntyre, Shelby  H.,  1982.  "An Experimental  Study of the Impact of Judgment-Based
     Marketing Models, Management Science, Vol. 28, No. 1, January 1982.

Morgan,  M. G.,  S. C., Morris, A. K. Meier, S. L. A.  Schenk, 1978.  "A Probabilistic Meth-
     odology for Estimating Air Pollution Health Effects form Coal-Fired Power Plant",
     Energy Systems Policy, 1978, J33, pp. 287-309.

Morgan,  M. G., et al., 1979.  Expert  Judgments for Policy Analysis, National Center for
     Analysis of  Energy Systems Department of Energy and Environment, Brookhaven
     National Laboratory. Upton, New York.

Morgan,  M. G., et al., 1981. Implications of Diverse Expert Opinion for Estimates of
     Sulfur Mass  Balance in Power Plumes,  Department  of Engineering and  Public
     Policy, Carnegie-Mellon University,  Pittsburgh, Pennsylvania.
                                       5-1

-------
                     Energy and Resource Consultants, Inc.
National Academy of Sciences, 1975a. Decision Making for Regulatory Chemicals in the
     Environment,. Environmental Studies Board, Washington, D.D., 1975a.

National Academy of Sciences, 1975b. Air Quality and Stationary Source Emission Con-
     trol, report  by the Commission on Natural Resources for the Committee on Public
     Works, U.S. Senate; Serial No. 94-4,  March, 1975.

National Academy of Sciences,  1977.  Decision Making in the Environmental Protection
     Agency, Report to the U.S.E.P.A. from the Committee on Environmental Decision
     Making, 1977.

New England River  Basins  Commission,  Regional Acid Deposition Economic/Social Im-
     pacts Work Group: Technical Report, May, 1981.

National Academy of  Sciences, 1981. Atmosphere - Biosphere Interactions;  Toward a
     Better Understanding of the Ecological Consequences of Fossil Fuel Combustion.
     National Academy Press.

Raiffa, H., 1970. Decision Analysis, Addison-Wesley Publishing, second printing,  July,
     1970.

Schailfer,  R., 1969.  Analysis of Decisions Under Uncertainty, McGraw-Hill, Inc., 1969.

Spetzler,  C.  S.  and Von  Holstein,  1975.   "Probability Encoding Decision Analysis",
     Management Science, Vol. 22, November, 1975.

Utility Air Regulatory  Group, 1981. Briefing Document for the Council on Environmen-
     tal Quantity, March,  1981.
                                       5-2

-------