United States
Environmental Protection
Agency
Office of Health and Ecological
Effects
Washington DC 20460
EPA-600/5-78-014
June 1978
Research and Development
&EPA
Critical Review
of Estimating
Benefits of
Air and Water
Pollution Control
-------
RESEARCH REPORTIMG SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7 Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL HEALTH EFFECTS RE-
SEARCH series. This series describes projects and studies relating to the toler-
ances of man for unhealthful substances or conditions. This work is generally
assessed from a medical viewpoint, including physiological or psychological
studies. In addition to toxicology and other medical specialities, study areas in-
clude biomedical instrumentation and health research techniques utilizing ani-
mals — but always with intended application to human health measures.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/5-78-014
CRITICAL REVIEW OF ESTIMATING
OF AIR AND WATER POLLUTION CONTROL
By
A. Hershaft (ed.)
A. M. Freeman III
T. D. Crocker
J. B. Stevens
Contract No. 68-01-2821
Project Officer
Thomas E. Waddell
Office of Research and Development
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF HEALTH AND ECOLOGICAL EFFECTS
OFFICE OF RESEARCH AND DEVELOPMENT
WASHINGTON, D.C. 20460
-------
DISCLAIMER
This report has been reviewed by the Office of Research and
Development, U.S. Environmental Protection Agency, and approved
for publication. Approval does not signify that the contents
necessarily reflect the views and policies of the U.S. Environmental
Protection Agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.
11
-------
PREFACE
This report is submitted in fulfillment of U.S. EPA contract #68-01-
2821. It contains a critical review of current methods and data employed
in estimating benefits of air and water pollution control.
In order to provide a balanced perspective in assessing the state-of-
the-art in the "benefits area", the material in this report is derived from
three types of sources:
• Extensive experience by three distinguished consultants in
conceptualizing and analyzing benefit estimates;
t Recent experience in preparation of national benefit esti-
mates by the Enviro Control staff; and
• A two-day conference on benefit estimation involving some
25 participants from a number of government, academic and
private research institutions.
The three consultants on general benefit estimation, air pollution
control, and water pollution control benefits were, respectively:
• Dr. A Myrick Freeman III
Dept. of Economics, Bowdoin College
t Dr. Thomas D. Crocker
Dept. of Economics, University of Wyoming
• Dr. Joe B. Stevens
Dept. of Agricultural and Resource Economics,
Oregon State University
The Enviro Control contributions, including preparation of the over-
view, as well as coordination and editing of the three papers, were made
by Dr. Alex Hershaft, Director of Environmental Studies, with the assis-
tance of Mr. H. Theodore Heintz, Jr., Senior Economic Consultant, and Mr.
Gerald C. Horak, Staff Scientist. Ms. Anita Calcote was responsible for
typing and production of the report. Valuable guidance on content and
format was provided by Mr. Thomas E. Waddell, the U.S. EPA Project Officer.
111
-------
ABSTRACT
This report provides a critical review of the current state of the
art and future prospects of estimating benefits of air and water pollution
control. Benefits of controlling air and water pollution arise from
gains resulting from improvements in air and water quality. Such gains
can be the reduction of damages caused by pollution or the increase in
options now feasible because of improved environmental quality.
This report represents three independent critiques by three experts.
Benefit assessment methodologies were evaluated for the following benefits
categories: human health, recreation and aesthetic properties, productivity,
materials, and natural ecosystem perturbations. Specific aspects discussed
include the nature and role of benefits, damage functions, valuation of
effects, aggregation of results, and representation of uncertainties.
The conceptual foundations of estimating pollution control benefits
were presented and compared with empirical studies. It was concluded that
while available estimates often do not adequately reflect the state of the
art, estimates of pollution control benefits would potentially be very
useful to decision makers. The conceptual basis provided by economic theory
for benefit estimation is adequate in most respects and far ahead of the
corresponding empirical effort. A number of studies are guilty of failing
to list explicitly critical assumptions or to express adequately uncertainty
in the results, while other studies have employed conceptual models that are
inappropriate to the problem at hand or the available data. It was also
concluded that damage functions underlying benefit estimates are frequently
based on insufficient data and/or inadequate~characterization of exposure
and effects. National benefits estimates were found to be based on regional
studies which are frequently inadequate in number and/or quality.
For improving general estimation of benefits it is recommended that
funding for benefits estimation research be reassessed in light of its
potential utility to decision-making process, that a comprehensive national
plan for closely coordinated regional benefit studies be developed and im-
plemented and that ex post evaluation of pollution control benefits gained
through past and current control programs be conducted.
iv
-------
TABLE OF CONTENTS
PREFACE iii
ABSTRACT iv
LIST OF FIGURES ix
LIST OF TABLES ix
I. OVERVIEW 1-1
A. PURPOSE AND SCOPE 1-1
1. Background 1-1
2. Purpose and Scope 1-3
B. CONCEPTUAL FOUNDATIONS 1-5
1. Nature and Role of Benefits 1-5
2. Damage Functions 1-7
3. Valuation of Effects 1-11
4. Aggregation of Results 1-15
5. Representation of Uncertainties 1-17
6. Conclusions and Recommendations 1-19
C. AIR QUALITY BENEFITS 1-21
1. Current Estimates 1-21
2. Health Benefits I_23
3. Aesthetic Benefits 1-24
4. Vegetation and Materials Benefits 1-25
5. Recommendations I_26
D. WATER QUALITY BENEFITS I_28
1. Current Estimates 1-28
2. Recreational Benefits 1-30
3. Aesthetic Benefits 1-31
4. Health and Production Benefits I_32
5. Conclusions and Recommendations 1-33
E. REFERENCES 1-35
II. BENEFITS OF POLLUTION CONTROL - A. Myrick Freeman, III II 1
A. INTRODUCTION II 1
B. POSSIBILITY AND USEFULNESS OF NATIONAL
BENEFIT ESTIMATES II-4
1. On Possibility 11-4
2. On Usefulness II-6
3. Information Content II-S
-------
TABLE OF CONTENTS
(continued)
Page
C. THE THEORY OF BENEFITS 11-12
1. Efficiency and Equity 11-12
2. Measurement of Benefits 11-13
3. The Estimation Process 11-14
D. CONCEPTUAL PROBLEMS IN BENEFIT ESTIMATION 11-17
1. Valuation of Life 11-17
2. Overlaps and Gaps 11-20
3. Temporal Aggregation 11-23
4. Uncertainty of Results 11-24
E. TECHNICAL PROBLEMS IN BENEFIT ESTIMATION 11-27
1. Measures of Pollution 11-27
2. Pattern of Exposure 11-29
3. Multicollinearity 11-30
F. PREPARING NATIONAL ESTIMATES 11-34
1. Criteria 11-34
2. Aggregation from Building Blocks 11-37
3. Specific Examples 11-38
G. CONCLUSIONS 11-46
H. RECOMMENDATIONS II-49
1. Basic Data and Analytical Techniques 11-49
2. Use of Existing Data 11-51
3. Ex Post Evaluation 11-53
I. REFERENCES 11-54
III. BENEFITS OF AIR POLLUTION CONTROL - Thomas D. Crocker III-l
A. INTRODUCTION III-l
1. Purpose and Scope III-l
2. Estimation of Benefits 111-3
3. Organization III-5
B. CONCEPTUAL FOUNDATIONS III-6
1. Motivations for Measuring Benefits III-6
2, Measures of Benefits 111-8
3, Sources of Pollution Control Benefits III-ll
VI
-------
TABLE OF CONTENTS
(continued)
Page
C. EMPIRICAL PROCEDURES 111-16
1. Methods of Estimating Benefits 111-16
2. Availability of Data III-18
3. Expression of Uncertainty in Results II1-22
4. Aggregation Procedures 111-24
D. SHORT-TERM IMPROVEMENTS IN BENEFITS ASSESSMENT 111-29
1. General Equilibrium Models of Prop-
erty Values 111-29
2. Household Production Function 111-31
3. Input Productivity Studies 111-32
4. Environmental Quality Indices 111-34
5. Form of Benefit Functions 111-35
6. Interview Studies 111-35
7. Studies of Agricultural Damages 111-36
E. STATUS AND PROSPECTS OF BENEFITS ASSESSMENT 111-38
1. Status and Role of Benefits Assessment 111-38
2. The U.S. EPA Program 111-40
F. REFERENCES AND BIBLIOGRAPHY 111-43
ATTACHMENT. AN ILLUSTRATION OF SOME ISSUES IN
BENEFITS ASSESSMENT 111-47
1. Introduction 111-47
2. Presentation of the Problem III-48
3. Multi-Dimensional Heterogeneity 111-50
4. Costs of Market Participation 111-53
5. Institutional and Temporal Constraints 111-54
6. Conclusion 111-55
IV. BENEFITS OF WATER POLLUTION CONTROL - Joe B. Stevens jy-l
A. INTRODUCTION IV-1
B. CONCEPTUAL FOUNDATIONS IV-3'
1. Role of Benefit-Cost Analysis IV-3
2. Estimation of Benefits IV-5
3. Special Considerations IV-6
vii
-------
TABLE OF CONTENTS
(continued)
Page
C. GENERAL PROBLEMS IV-10
1. State of Damage Functions IV-10
2. Classification and Aggregation of Benefits IV-13
3. Other Problems IV-14
D. SPECIFIC PROBLEMS IV-16
1. Health Benefits IV-16
2. Production Benefits IV-18
3. Recreation Benefits IV-19
4. Aesthetic and Ecological Benefits IV-24
5. Property Values IV-27
E. CONCLUSIONS AND RECOMMENDATIONS IV-29
F. REFERENCES IV-33
ATTACHMENT A. MODEL I IV-35
ATTACHMENT B.- MODEL 2 IV-37
viii
-------
LIST OF FIGURES
Figures Page
1-1 Hypothetical Damage Function 1-8
II-l Classification Scheme for Capture of Effects 11-22
II-2 True and Observed Relationships Between Pollu-
tion Measure and Mortality Rate 11-28
II-3 Recreational Benefits of Water Pollution Control II-44
IV-1 Representation of Consumer's Surplus III-9
IV-2 Willingness to Pay for Improvement in Air Quality 111-27
LIST OF TABLES
Tables Page
1-1 Estimated National Damages of Air Pollution
for 1973 1-21
1-2 Availability and Reliability of Information
on Air Pollution Damages 1-23
1-3 Estimated National Damages of Water Pollution
for 1973 1-28
1-4 Availability and Reliability of Information on
Water Pollution Damages 1-29
ix
-------
I. OVERVIEW
This first chapter presents a general overview of the state-of-the-
art in estimating benefits of air and water pollution control. The ma-
terial is based on the three papers that follow, as well as other sources
referenced at the end of this chapter (Heintz et al.. 1976 and Waddell,
1974). References to specific material in the three papers are given,
wherever appropriate, to direct the reader to additional information.
This chapter is covered under the headings: purpose and scope, concep-
tual foundations, air quality benefits, and water quality benefits.
A. PURPOSE AND SCOPE
This section sets the scene for subsequent discussion by present-
ing the background for this study, its purpose and scope, and the plan
of work.
1. Background
Nearly everyone is now satisfied that there exists a causal
relationship between environmental pollution and certain damages
suffered by society. These may take the form of increased inci-
dence and prevalence of disease, diminished recreational expe-
rience, decreased property values, reduced crop yields, more fre-
quent maintenance and replacement of exposed materials, and other,
less well-identified losses. This being the case, a reduction in
pollution levels should bring about a corresponding decrease in
these damages and produce a set of benefits equivalent to the dif-
ference in damages before and after the reduction took place.
Legislators, planning officials, and'other environmental de-
cision makers are frequently faced with the decision of how much
to reduce pollutant levels, in the light of the associated direct
costs of pollution control and possible secondary economic impacts.
In the past, the rationale for these decisions was rather obvious,
in that they were often made in response to popular sentiment. How-
ever, as economic conditions have changed for the economy as a whole,
1-1
-------
as well as for certain industries* the costs have become more acutely
felt, especially in the wake of the energy crisis. At the same time,
the beneficial effects of reduced, or stable, pollution levels were
neither obvious, nor easily-measured. Clearly, the decision makers
would need a more sensitive too.l for comparing and trading off the
costs and benefits of various levels and types of pollution control.
It was this need that gave birth to increased interest in en-
vironmental benefit/cost or tradeoff analysis, including'benefit as-
sessment research. Admittedly, this research is an inexact science,
primarily because the underlying biological and physical,relation-
ships need considerable additional research and because social bene-
fits and costs are diffuse and frequently difficult to quantify, much
less to express in monetary terms. Even so, the process of logical
and systematic scrutiny inherent in benefit/cost analysis can contri-
bute substantially to the ability of decision makers to improve the
social welfare through more efficient allocation of the limited re-
sources of the public treasury.
This potential contribution of tradeoff analysis was apparently
recognized by the framers of the National Environmental Policy Act
of 1969 (PL 91-190), primarily in Sections 102 and 204. Section 102
calls for the "identification and-development of methods and proce-
dures which will ensure that presently unquantified environmental
amenities and values may be given appropriate consideration in deci-
sion making, along with economic and technical considerations." Sec-
tion 204 charges the Council on Environmental Quality (CEQ) to gather,
analyze, and interpret timely and authoritative information concern-
ing the conditions and trends in the quality of the environment.
In recent years, there have been a number of estimates of the
benefits of air and water pollution control. Among the most notable
were the reports on air and water pollution-by Waddell (1974) and
Unger et al. (1974), respectively. More recent estimates are report-
ed in CEQ's annual report, Environmental Quality - 1975. However,
there seems to be a widespread skepticism about the validity of avail-
able benefit estimates.
1-2
-------
With the recent reorganization of the U.S. EPA's Office of Re-
search and Development, the benefits assessment program is being re-
evaluated. This report should serve as valuable input into the eval-
uation and re-direction of the benefits research.
2. Purpose and Scope
The purpose of this project is to assit decision makers in U.S.
EPA and other Federal agencies by providing a critical review of the
current state of the art and future prospects of estimating benefits
of air and water pollution control. This should prove valuable
in allocating limited pollution control resources; assessing the pros-
pects for further benefits research; and in scoping out a viable bene-
fits research effort.
The specific objectives of this effort are as follows:
• To assess the current state of the art in estimating
benefits of air and water pollution control, including
validity and accuracy of techniques and data
• To assess the future prospects of benefit estimation
• To identify fruitful areas for additional research.
The scope of this effort can be formulated in terms of the pollu-
tants, benefit categories, affected populations, geographic areas, and
time horizon. The benefit categories for air and water pollution are
listed below:
Air pollution: Water pollution:
Human health • Human health
Vegetation • Production (municipal, in-
Materials dustrial, and agricultural
Aesthetics supplies; commercial fish-
Other (animals, eries and materials)
ecological risks, « Recreation
etc.) « Aesthetics and property
values
1-3
-------
The remainder of this chapter discusses the conceptual founda-
tion of estimating pollution control benefits and summarizes air and
water quality benefit discussions as presented in the subsequent chap-
ters.
1-4
-------
B. CONCEPTUAL FOUNDATIONS
This section addresses the conceptual foundation of estimating pol-
lution control benefits as presented in the three papers. Detailed re-
ferences to the papers are provided throughout. The topical headings are:
nature and role of benefits, damage functions, valuation of effects, ag-
gregation of results, representation of uncertainties, and conclusions and
recommendations.
1. Nature and Role of Benefits^
Benefits of controlling air and water pollution arise from gains
resulting from improvements in air and water quality. Such gains can
be the reduction of damages caused by pollution or the increase in op-
tions now feasible because of improved environmental quality. The'cor-
responding economic damages result in "out of pocket" losses by increas-
ing the costs of using air and water, by decreasing their use or acti-
vities depending on their use, and by increasing the costs of avoiding
or repairing the effects of pollution. A basic concept in benefit eval-
uation is "willingness to pay," defined as the highest price that indi-
viduals would be willing to pay to obtain the improvement in air or wa-
ter quality resulting from a given pollution control program. Econo-
mists prefer to evaluate benefits in monetary terms, because this pro-
vides a common measure of all types of benefits and costs.
However, many types of damages are not amenable to quantification
in monetary terms, because of their nature and the state-of-the-art of
available measurement methods. These difficult-to-measure benefits
probably account for a substantial portion of the total value of pol-
lution control to society. This is the case with "psychic" damages,
so labeled because they relate to the pleasure or displeasure asso-
ciated with the use of the environment. Psychic damages include de-
creased pleasure and increased pain and anxiety from the use of pollu-
i
ted air or water, as well as the loss of non-user values.
Non-user values arise with people who have no immediate plans of
making direct use of an environmental amenity, but are nevertheless
1-5
-------
willing to pay for its preservation for a variety of reasons. In
the case of option value, they wish to ensure an option of being
able to use it in the future. Vicarious or bequest values describe
the benefits experienced by people who would provide these environ-
mental amenities to their heirs or others. Preservation value is
associated with the desire to preserve a unique natural resource.
Finally, risk aversion refers to the willingness of people to pay
for decreasing or averting the risk of a catastrophic or irreversi-
ble damage, such as the flooding of arable land or extinction of a
biological species. (Heintz, Hershaft, Horak, 1976).
Over time, as the underlying benefit analyses become more reli-
able, estimates of pollution control benefits should provide for es-
tablishment of more efficient environmental policies. However, equity
considerations, as well as various political, institutional, and tech-
nical issues also enter into the decision making process. A thorough
tradeoff analysis would integrate all of these factors.
Other contributions of benefit estimation may be listed as fol-
lows (Chapter II pp. 6-8; IV 3-4):
• Enhanced understanding of the problem, the underlying
factors, and potential solutions
t Development of an analytical framework for entering
inputs from other sources
• Establishment of reasonable environmental quality stan-
dards.
Estimation of pollution control benefits ideally should follow
the sequence of steps listed below (Heintz et al., 1976; II 53-65):
t Project pollutant emissions on the basis of population
levels and economic activity for the area and time pe-
riod under consideration
• Estimate reduction of pollutant emissions attributable
to implementation of given control policy
1-6
-------
• Estimate improvements in environmental quality asso-
ciated with stipulated reduction of emissions
t Estimate reduction fn adverse effects associated with
improvement in environmental quality
• Estimate regional benefits (in monetary terms if pos-
sible) of reductions in adverse effects and other con-
siderations
• Extrapolate and aggregate regional benefit'estimates
across all relevant regions and time periods of in-
terest to obtain national estimates.
The first two steps involve the projection of a suitable eco-
nomic scenario and evaluation of the cost-effectiveness of various
administrative and technological pollution control fixes. The third
step requires the use of complex models of the diffusion and assimi-
lation of specific pollutants within their respective media. The re-
maining steps rely on (1) the development of damage functions and (2)
economic benefit analysis. These two topics form the basis of this
report and receive closer scrutiny in the pages that follow.
2. Damage Functions
Damage functions provide a quantitative expression of the rela-
tionship between exposure to specific pollutants. The type and ex-
tent of the associated effect can then be estimated based on a tar-
get population, or "population at risk". Despite their crucial role
in the formulation of benefits of pollution control, damage functions
have been treated lightly in the three papers that follow (Chapters
II, III and IV), because the essence of damage functions is more phy-
sical and biological than economic. Thus, most of the material pre-
sented in this section is abstracted from Hershaft et al. (1976).
A typical S-shaped damage function, showing the damage corres-
ponding to a given exposure to a specific pollutant, is presented in
Figure 1-1. The ordinate may represent either the number of individ-
uals affected or severity of effect. The abscissa indicates the dos-
age in terms of time at a given ambient concentration, or in terms of
1-7
-------
ambient concentration for a fixed period of time. The lower portion
of the curve suggests that, up to a certain exposure value, known as
a "threshold level", no damage is observed, while the upper portion
indicates that there exists a damage saturatipn level (e.g., total
destruction of the target population) beyond which increased expo-
sure levels do not produce additional damage. The middle, quasi-
linear portion is very useful in that any data points here can be
readily interpolated, and the frequent assumption about linearity
of a damage function is most valid in this sector.
O)
en
I
to
Q
With
Controls
Without
Controls
Saturation
Level
Exposure
Figure 1-1. Hypothetical Damage Function
Exposure is typically measured in terms of ambient concentra-
tion levels and their duration and it may be expressed as "dosage"
or "dose". Dosage is the integral of the time and ambient concen-
tration to which the subject has been exposed, whereas dose repre-
1-8
-------
sents that portion of the dosage that has been actually instrumental
in producing the observed damage. The damage can become manifest in
a number of ways and can be expressed in either physical and biologi-
cal, behavioral response, or economic terms. If the effect is physi-
cal or biological, the resultant relationship is known as a physical
or biological damage function, or a dose-effect function. In an eco-
nomic damage function, on the other hand, the effect is expressed in
monetary terms. Economic damage functions can be developed by assign-
ing dollar values to the effects of a physical or biological damage
function, or by direct correlation of economic damages with ambient
pollutant levels.
In reporting a damage function, one needs to specify the pollu-
tant (or surrogate), the dose rate, the effect, and the "population
at risk". Dose rate, or the rate at which ambient concentration var-
ies with time, has a major influence on the nature and severity of
the resultant effect. For example, long-term exposure to relatively
low concentrations of air pollutants'may result in manifestations of
chronic disease9 characterized by extended duration of development,
delayed detection, and long prevalence. Short-term exposure to high
concentration levels, on the other hand, may produce acute symptoms,
characterized by quick response and ready detection, as well as chron-
ic cumulative, or delayed effects.
Specification of the population at risk involves the characteri-
zation of the nature and magnitude of the exposed population. Damage
functions, when extrapolated on the basis of population at risk, serve
to define the total damages produced by a given level of exposure by
multiplying the corresponding unit damage (e.g., increased mortality)
for the specified population at risk (e.g., white males over 65) by
the total number of units within this population. Detailed popula-
tion characteristics also permit investigators to adjust their results
to reflect the influence of various intrinsic (e.g., age, race, sex)
and extrinsic (e.g., general health, occupation, income, and education)
variables in assessing the specific effects of air pollutants (e.g.,
increased incidence of lung cancer). Finally, population-at-risk in-
1-9
-------
formation can provide useful guidance for allocating pollution con-
trol resources by identifying areas with particularly susceptible
populations exposed to relatively hazardous levels of pollutants.
(Takacs and Shea, 1975).
The data required to develop physical or biological damage
functions are obtained primarily through epidemiological, field,
clinical, toxicological, or other laboratory investigations. The
first approach involves the comparative examination of the effects
of pollutants on selected segments of population exposed to differ-
ent levels of pollution, in order to deduce the nature and magni-
tude of the likely effect. Field observations represent a similar
approach to assessment of effects on animals, vegetation, and ma-
terials, and they are characterized by similar analytical tech-
niques and concerns. Clinical studies are concerned with the ef-
fects of exposure on human subjects. Toxicological investigations
involve deliberate administration of controlled doses of pollutants
to animal, and occasionally, human subjects, and observation of the
resulting effects. Laboratory studies represent essentially the
same approach for determining effects of pollutants on plants and
materials.
The three principal techniques for analyzing the relationship
between exposure and effect indices in epidemiological studies are
known as cross tabulation, multivariate regression, and non-paramet-
ric or distribution-free analysis. Cross tabulation is the simplest
of the three. Multivariate regression provides a rapid indication
of the degree of association between a large number of independent
and dependent variables and is readily programmable for computer op-
eration. However, its validity is heavily contingent on a fairly
precise a priori definition of the relationship between each indepen-
dent and the dependent variables and on precise measurement of the
independent variables. Thus, this technique is especially vulnerable
to the poor precision in measurement and reporting of air pollution
levels, for example, for a given segment of population. Non-para-
metric analysis, on the other hand, is free of these assumptions,
I-IO
-------
but requires laborious data reduction for each of the many pairs of
independent and dependent variables and interposition of expert judg-
ment at each step of the process.
In the development of damage functions on the basis of epidemi-
ological or field studies, it is important, albeit complex, to iso-
late or control the influence of cofactors and covariates. The for-
mer may be defined as factors that act in concert with the indepen-
dent variables (e.g., relative humidity or cigarette smoking and pol-
lutant level). Covariates, on the other hand, may be thought of as
factors that vary jointly with the principal dependent variable. The
interference of cofactors and covariates with results of the analysis
can be reduced by holding them constant or with the aid of statistical
techniques. When this cannot be done, it is frequently assumed that
the distribution of these factors in the target population is suffi-
ciently uniform to avoid vitiating the basic conclusions.
Finally, it should be noted that epidemiological and field stud-
ies and observations can only indicate an association between expo-
sure to pollution and the observed effect, suggesting the existence
of a causal relationship. Such a relationship can then be tested by
clinical, toxicological and laboratory studies.
3. Valuation of Effects
Estimation of economic benefits of a given improvement in en-
vironmental quality involves either conversion of the associated
reduction of adverse physical and biological effects into monetary
terms, or direct determination of the user's willingness to pay for
the improvement. The three common methods to estimating benefits
of pollution control are:
• Alternative cost
« Opportunity cost
® Willingness to pay.
1-11
-------
The alternative cost method is employed most frequently, be-
cause of its rapid applicability and avoidance of complex economic
analysis. This involves development of a damage function, leading
to estimation of the total damages corresponding to exposure of the..
target population over a specified period. Because there is no pro-
vision for substitution or any other mitigative adjustment by the
target population, the damage estimate may be excessive. (Ill 16).
Opportunity cost, on the other hand, is estimated on the basis
of the cost of substitution and other adjustment opportunities .open
to the target population. This formulation presumes that ownership
to the environmental good is held by the target population, which is
then entitled to trade it away for a substitute good.
Finally, the willingness-to-pay method is based on the deter-
mination of how much the affected population is willing to pay to
avoid the particular environmental degradation. Here, the title is
presumed to be vested in the perpetrators of environmental degrada-
tion, rather than the target population. (Ill 17).
An alternative formulation of approaches to estimation of pol-
lution control benefits is as follows (IV 5-6):
© Market prices, if a market for these goods and ser-
vices exists and the additional output does not af-
fect price significantly
® Simulated market prices (especially useful for esti-
mation of recreation and aesthetic benefits)
c Changes in net income of producers for consumers, if
outcome represents an intermediate good (useful for
production benefits)
® Cost of most likely alternatives for obtaining same
objective.
In spite of general reliance on simulated markets, some market
price data are also useful in estimation of benefits. These data
need to be germane, relatively free of price imperfections, and ad-
1-12
-------
justed to long-term expected levels. Technological externalities
should be included and monetary externalities should be excluded
in estimating these benefits. Finally, the annual changes in bene-
fit streams should be identified for the various benefit categories.
(IV 5-9).
Estimation of benefits on the basis of willingness to pay en-
tails the intermediate steps of estimating changes in user behavior
associated with anticipated reductions in the adverse effect and the
marginal willingness to pay associated with these changes. Changes
in user behavior reflect the user's perception of environmental qual
ity. Moreover, they are affected by local socioeconomic conditions.
(IV 10-11).
The benefits associated with a change in environmental quality
are defined by the area under the demand curve over the applicable
range, but, since environmental amenities are generally not bought
and sold in the marketplace, there may not be any direct way of es-
timating the shape of the demand curve by conventional econometric
techniques. Instead, discreet market values may need to be inferred
from individual responses to changes in environmental quality. The
techniques for drawing these inferences are known as market demand
analysis, net productivity, net factor income, travel cost studies,
land value studies, and personal interviews. (II 15-16).
Another representation of the relationship between changes in
environmental quality and the user's utility or welfare is known as
"consumer's surplus". This is defined as the difference between
what an individual is willing to pay and the market price. Determi-
nation of consumer's surplus can be affected by an alternative for-
mulation of the problem. (Ill 9-10).
An important element in estimation of pollution control bene-
fits is the aversion of grave environmental risks brought about by
pollution. The value of risk aversion can be measured through ob-
servation of individual behavior, or through lottery games. The
1-13
-------
f-irst method is preferred, because it is more realistic, permits
construction and testing of behavior models, and provides for ap-
plication of opportunity cost or willingness-to-pay formulations.
(Ill 19-21).
In closing this discussion, it is important to bear in mind
that the data required to implement these valuation steps are fre-
quently difficult to obtain and that some of the economic concepts,
such as the value of human life, or non-user benefits, have not been
adequately defined. Thus, judgmental estimates need to be frequently
substituted for hard data and a balance must be struck between the de-
sirable and available data and techniques. In such cases, it is very
important to>state explicitly all assumptions, the reasons for their
•
selection, and the implications of choosing different assumptions.
Otherwise, the estimate can easily be misleading and discredit the
entire benefit estimation process. (II 5-8).
Usefulness of benefit estimates is governed by their adherence
to the following provisions (II 9-11):
• Consistent terminology
• Clearly defined pollutant levels
• Clearly identified pollutants and sources
• Appropriate degree of detail
• Clearly defined relationships between pol-
lutant levels and human values.
The term "costs" should apply to pollution control resources;
"damages" should refer to the total loss to society due to a pollu-
ted environment; "benefits" should represent the gains realized by
society from a given improvement in environmental quality. Similar-
ly, the quality levels being compared should be already defined as
those "with" and "without" controls, rather than those "before" and
"after" applications of controls, to avoid time-related complications,
such as inflation. (II 9-10).
1-14
-------
The pertinent pollutants and their sources should be clearly
identified with the environmental improvement sought, to avoid gaps
and overlaps in the estimation procedure. Moreover, policy deci-
sions should be based primarily on marginal benefits associated with
efects of the particular decision, rather than on average or total
benefits that would accrue in the absence of pollution. (II 9-10).
Benefit estimates corresponding to a given environmental im-
provement should be expressed in monetary terms and should be based
on individual behavior and preference. Empirical measures and tech-
niques should be appropriate to the specified theoretical model and
available data. Finally, these estimates should reflect the regional
and temporal variations in both economic and environmental variables.
(II 9-11, 34-35).
The type and degree of geographic, temporal, and economic de-
tail should be tailored to the specific needs of the user. For ex-
ample, only a rough estimate may be needed to make an initial deci-
sion, with more detailed estimates being provided for more refined
future decisions. (II 11).
4. Aggregation of Results
Most benefit estimation studies address a specific geographic
location, group of pollutants, population at risk, and time period.
Extension of these results to the national level and some future
time frame requires the aggregation of the regional estimates and
projection of a number of variables^ including ambient levels, popu-
lations at risk, personal incomes, and costs of damages. In aggre-
gating over several variables, it is important to specify the order
in which the variables are being aggregated. (Ill 24-28).
The benefit categories for which the data are collected are
often dictated by availability of information sources and analyti-
cal expediency, rather than the needs for a uniform and self-con-
sistent framework. Consequently, different studies evaluate dam-
I 15
-------
ages that are not necessarily additive, or even comparable, and care-
ful interpretive techniques must be applied to the results of such
studies to prevent gross overlaps or omissions of damage estimates.
Moreover, in aggregating such fractional results, it has not been
possible to reflect the potential impacts of changes ;'n individual
components on one another, nor the impact of the general adjustments
of the economy and the resulting reduction in damages. (Heintz et al.,
1976).
Aggregation of benefit estimates entails a tradeoff of detailed
information about form and structure in return for treatability and
ease of comprehension. Attempts to apply aggregated national esti-
mates to local pollution control decisions can introduce substantial
errors, because the information lost in the aggregation process can-
not be recovered. In principle, national estimates may be developed
directly, rather than by aggregation of local studies. (Ill 24-28).
However, additional research in this area is necessary.
Overlaps and gaps between categories of benefits may arise when
two types of effects (e.g., health effects and property values, in
the case of air pollution, and recreation benefits and property val-
ues, in the case of water pollution) are estimated by different meth-
ods which may count the same benefit component twice or fail to cap-
ture certain other components. It is important, therefore, to at-
tempt removal of the excess count and to inpute a value for the mis-
sing component. Another problem is the inconsistency in quality of
estimates for different benefits categories. (II 20-23; IV 13-14).
If the benefits of interest accrue over a number of years, then
it may be useful to compute the present value of the total stream of
benefits with the aid of an appropriate discount rate and time hori-
zon. This approach becomes less effective when the projected effects
extend over a very long time period that spans several generations
(intergenerational effects), because of the large reduction factor.
For example, using a discount rate of 5 percent, the current benefit
of an effect occurring 200 years hence is reduced by a factor of
1-16
-------
6 x 10 . In such cases, the consideration of intergenerational
equity should outweigh that of intertemporal efficiency. Another
problem in computing the present value of the stream of benefits
is the differential growth rates. (II 23-24; IV 13-14).
5. Representation of Uncertainties
Uncertainties about benefit estimates arise from errors in
the four intermediate steps of the estimation process:
t Specification
t Measurement
• Valuation
• Aggregation.
Errors of aggregation are associated with attempts to extra-
polate national values from regional estimates or future values
from current or past estimates. They arise from geographic and
temporal variability of user behavior and market conditions. Er-
rors of valuation are due to the difficulty of assigning monetary
values to certain physical, biological, aesthetic, or psychic ef-
fects. (Ill 22-23).
Errors of specification include any type of error in specify-
ing the functional form of the relationship under study or in ac-
counting for important variables. A particularly common and grave
error of specification is committed in attempting to extrapolate a
complete functional relationship from a few data points that are
barely adequate to characterize a small portion of the curve. Even
if one were willing to make an assumption about the overall shape
of the function, there is frequently no way of knowing which portion
is represented by these data points. (Hershaft et al., 1976).
Errors of measurement may be incurred in the course of the fol-
lowing steps of the benefit estimation process:
• Location of monitoring station and subjects
• Sampling and analysis
1-17
-------
• Averaging and aggregation of ambient pollu-
tant levels
• Determination of effect
• Impact of covariates
• Characterization of population at risk.
If the errors of measurement of the independent variables are
relatively small, occur at random, and follow a normal, or Gaussian
distribution about the mean value of each variable, then the total
error of all the independent variables can be computed by standard
statistical techniques. However, this is seldom the case, because
measurement of such independent variables as pollutant level, meteo-
rological conditions, and socioeconomic characteristics is subject
to errors that are both large and biased. The advantages of this
statistical approach include an opportunity to incorporate more in-
formation in the reported results and the assignment of a probabil-
ity to the various outcomes. (Hershaft et al., 1976; II 24-26; III
23-24).
Envelopes characterizing errors and uncertainties of benefit
estimates can be also obtained by more practical means, including:
• Replicating a specific study using new data or
methods
• Providing ranges of values of the more important
variables
• Combining results of several studies
• Applying "best" and "worst" case assumptions.
Replication and data manipulation are essentially empirical
techniques of determining the errors and corresponding confidence
bands. Combining the results of several studies is a rare and un-
certain opportunity, in light of the great variety of conditions
and populations that often characterize the different efforts. Ap-
plication of "best" and "worst" case assumptions is more an argu-
mentative than a statistical technique. The lower boundary, or best
1-18
-------
case, is established by attributing all reasonable portions of the
effect to any plausible cofactors and covariates and associating on-
ly the residual effects with pollutant exposure. The upper boundary,
or worst case, is determined by Inverting this procedure and assuming
a minimal impact of other variables. (Hershaft et al.. 1976).
6. Conclusions and Recommendations
A number of major conclusions emerge from the discussion of con-
ceptual foundations of estimating pollution control benefits:
• Estimates of pollution control benefits would poten-
tially be very useful to decision makers (II 46)
• However, available estimates do not adequately re-
flect the state of the art (II 48)
• The conceptual basis provided by economic theory for
benefit estimation is adequate in most respects and
far ahead of the corresponding empirical effort (II
46)
• Nevertheless, a number of studies have employed con-
ceptual models that are inappropriate to the problem
at hand or the available data
• A number of studies have failed to list explicitly
critical assumptions or to express adequately un-
certainty in the results
t National benefit estimates are based on regional
studies which are frequently inadequate in number
and/or quality
• Damage functions underlying benefit estimates are
frequently based on insufficient data and/or inade-
quate characterization of exposure and effects (II
27-30)
• Benefits assessment research has been accorded a
relatively low priority by the U.S. Environmental
Protection Agency, perhaps because of skepticism
about its value, or the public health and ecosys-
tems focus of environmental legislation (III 40-
41)
• Past U.S. EPA research strategy has been charac-
terized by short-term, piecemeal studies which
are not conducive to the accumulation of a con-
sistent and comprehensive data base useful in
decision making (III 41-42)
1-19
-------
Some aspects of benefit estimation are subject to
value and equity judgments, rather than applica-
tion of economic theory (II 47-48; IV 13-15)
relative merits of different distributions
of benefits among population groups
adverse consequences of present actions on
future generations
Recommendations for improving general estimation of benefits
are as follows:
• Reassess funding for benefits estimation research
in light of its potential utility to decision-mak-
ing process and the uncertainties permeating other,
currently funded national assessment programs (III
41-42)
• Develop and implement a comprehensive national plan
for closely coordinated regional benefit studies
that is well grounded in economic theory and pro-
vides for consistent aggregation and expression of
the uncertainty of all estimates (III 42)
• Develop and implement a comprehensive program of
ex post evaluation of pollution control benefits
gained through past and current control resources
(II 53)
• Convene a group of experts every few years to de-
velop comprehensive estimates of national benefits
of pollution control (II 52)
t Employ scientists from other disciplines in plan-
ning and implementing benefits estimation studies
(IV 25).
1-20
-------
C. AIR QUALITY BENEFITS
This section presents an overview of the discussion on benefits
of air pollution control contained primarily in Chapter III, Speci-
fic references to this chapter are provided wherever appropriate. The
topics covered are: current estimates, health benefits, aesthetic ben-
efits, vegetation and materials benefits, and recommendations.
1. Current Estimates
The latest estimates of national air pollution damages have
been compiled by Heintz et al. (1976). These are equivalent to
benefits that would accrue annually from reduction of air pollu-
tion to threshold levels. Table 1-1 summarizes the benefit es-
timates for the four major classes of benefits, in terms of best
estimates and corresponding ranges. The latter were derived
largely through a proportional representation of the ranges quo-
ted by Waddell (1974).
Table 1-1. Estimated National Damages of Air Pollution for 1973
($ billion)
Damage Category
Health
Aesthetic
Vegetation
Materials
Total
Best Estimate
5.
9.
2.
1.
20.
7
7
9
9
2
Range
Low
2.
5.
1.
0.
9.
0
7
0
8
5'
High
9.4
13.7
9.6
2.7
35.4
Source: Heintz et al., 1976
To gain a proper perspective of these estimates, it is im-
portant to recognize that they do not reflect all of the potential
damages from air pollution. There are a number of categories of
1-21
-------
potential damages for which estimates are^not available. The most
important of these is the threat to preservation of the natural
environment, including unique ecosystems and species. But, even
within the categories for which estimates are available, the
existing monetary measures tend to understate the total damages.
For example, the estimated health damages reflect the direct and
indirect costs of illness, such as health care costs and lost earn-
ings, but do not reflect the value of lost leisure time or the
psychic costs of illness and death.
In combining estimates from different classes of damages,
care has been taken to minimize double counting. For example,
studies of the differences in residental property values asso-
ciated with differences in air pollution reflect primarily the
aesthetic and soiling effects, rather than health effects. This
is based on the argument that the aesthetic effects are experi-
enced directly in everyday life, whereas health effects are
mostly long-term, and are not distinguishable by the general
population from other causes of illness. Although improved edu-
cation may be altering people's awareness of the health effects
of air pollution, it is not likely that this has been signifi-
cantly reflected in past property values, on which these benefit
estimates have been based.
The damage estimates presented here are based on the inter-
pretation of the results of numerous studies of varying scope,
methodology, and data quality. The availability and reliability
of information from these studies is indicated in Table 1-2. It
will be noted that effects data are most available for effects of
SO , oxidarits, and particulates and for the damage categories of
/\
human health and vegetation.
1-22
-------
Table 1-2. Availability and Reliability of In-
formation on Air Pollution Damages
Damage
Categories
Human Health
Aesthetics
Vegetation
Materials
NOX
SF
U
IF
SF
Oxi-
dants
SF
SF
IG
IF
S0x
IG
IG
IF
IG
CO
SG
U
SG
SP
HC's
SF
U
SF
SP
Parti -
culates
IG
IF
SF
SG
Other
IF
U
SF
U
Source: Heintz et al., 1976
Availability
A - ample
I - insufficient
S - scarce
U - unavailable
Reliability
E - excellent
G - good
F - fair
P - poor
2. Health Benefits
The key consideration in estimation of health benefits of air
pollution control is the valuation of human life. This can be ac-
complished by the Thaler-Rosen or the productivity method. The for-
mer estimates the average value of human life as somewhat in excess
of $200,000 on the basis of individual preference for higher wages
versus more risky occupations. The problem with this method lies
in the fact that willingness to pay for preservation of human life
rises steeply with proximity of death. (II 17-18).
The productivity, or human capital method values each life at
the present value of the expected stream of the individual's future
earnings. The difficulties here are the conversion of a probabilis-
tic estimate into a specific stream of earnings and the failure to
value lives of people who receive no compensation for their labor
(e.g., housewives, children, retirees, volunteers). It should be
1-23
-------
noted also that there is no logical connection between value of fu-
ture earnings and willingness to pay. (II 18-19). • '
f.
1 •
The household production function is an analytical framework
that can be applied in estimating health benefits of pollution con-
trol. This theory views the consumer as combining market purchased
goods and his own resources to produce a given state of health. The
major advantage of this method is that it provides for dealing with
the inadequacies of previous methods, such as failure to take into
account the value of leisure and uncompensated time as well as indi-
vidual adjustment to pollution. (Ill 31-32).
3. Aesthetic Benefits
Estimation of aesthetic benefits of air pollution control en-
compasses a number of the concepts discussed in Section B. This is
illustrated in Chapter III Attachment, which discusses a case study
involving the impact of air pollution on the outdoor recreational
market in San Bernardino, California. These concepts may be de-
scribed as (III 46-54):
• Multidimensional heterogeneity
• Cost of market participation
• Institutional and temporal constraints.
The first term refers to the fact that outdoor recreational
sites exhibit substantial differences in a number of dimensions
(e.g., size, topography, vegetation, facilities, congestion, ac-
cessibility). The second concept deals with the relatively high
cost of participating in the outdoor recreation market due to tra-
vel requirements, search for a suitable site, and purchase of spe-
cial implements (e.g., hiking boots and backpacks). Finally, the
market is subject to a number of controls exercised by public in-
stitutions, such as health and safety regulations and restrictions
on attendance, as well as constraints on the time schedule avail-
able for recreation. (Ill 49-53).
1-24
-------
Another aspect of aesthetic benefits is reflected in property
values. An improved method of obtaining an aggregated national es-
timate of property value changes has been developed. This method
does not assume that the same set of weights applied to each locale,
time period, and household, nor does it ignore the manner of distri-
bution. Moreover, it is soundly grounded in economic theory and is
capable of capturing several elusive features of the market. (Ill
29-30).
4. Vegetation and Materials Benefits
Estimates of air pollution damages to agricultural crops have
focused on the physical measures of damage and have ignored the ef-
fect of the resulting changes in output on the unit price. Yet,
models of agricultural markets that are capable of capturing these
price effects are generally available and should be applied to the
estimation of crop damages (III 36).
Past studies of production damages of air pollution have fo-
cused on materials damages and have neglected the involvement of
these inputs in the production process. It appears desirable, in
this connection, to assess the effects of air pollution on labor
supply and productivity. The findings should be transferable to
other cases of occupational exposure. (Ill 33).
Past attempts to construct indices of air pollution control
benefits failed, because they viewed benefit changes as nearly syn-
onymous with changes in ambient concentrations, and social valua-
tions were based almost exclusively on air quality standards. This
representation could be improved by viewing ambient quality as a
factor of production, e.g., changes in morbidity characteristics
associated with certain pollutant levels. (Ill 34).
Results of interviews to ascertain individual preferences are
frequently biased by the interviewees' anticipation of how their
responses will be used. Nevertheless, differences in willingness
to pay for various pollution levels may provide a valid representa-
1-25
-------
tion of the marginal willingness to pay, which is of primary impor-
tance in policy decisions. Moreover, techniques have been developed
for eliciting truthful responses about preference orderings. (Ill
35-36).
5. Recommendations
The following research projects are recommended to improve es-
timation of air pollution control benefits:
• Apply the household production function framework to
obtain health benefits of pollution control (III 31-
32)
t Conduct research on value of human life along two ap-
proaches (II 50):
determination of willingness to pay for small
changes in probability of death
determination of the value implicit in public
decisions involving public safety
• Assess effects of air pollution on labor supply and
productivity (III 32-33)
• Prepare improved estimates of benefits derived from
available property value studies and conduct new
studies (II 52; III 29-30)
• Apply available models of agricultural markets to es-
timation of crop damages (III 36)
• Investigate improved methods of assessing wi'llingness
to pay through interviews (III 35-36).
Benefit estimation would also gain from the following recom-
mendations designed to improve development of air pollution damage
functions (Hershaft et al., 1976).
• Determine how well air quality measurements at the
monitoring station represent the ambient quality af-
fecting the population at risk
• Determine the impact of mobility and shielding of
the subjects on their exposure
1-26
-------
Develop a composite index of exposure representing
both prevailing and extreme values
Define current air quality monitoring requirements
for potential future damage studies
Assess and improve the accuracy and reliability of
various methods for determining morbidity
Develop a method for determining changes in life
expectancy due to exposure to air pollutants
Provide uniform guidance for designing studies of
the effects of air pollutants on vegetation
Conduct studies on the following effects of air pol
lutants:
specific effects of NOp and nitrates
specific effects of different types of
sulfates
chronic effects of low-level exposures to
specific pollutants
carcinogenic, teratogenic, and mutagenic
effects of specific pollutants
effect of oxidants on yields of major
crops
effects on paints, cement, rubber, plas-
tics, and wood
effects on climate and local weather con-
di tions.
1-27
-------
D. .WATER QUALITY BENEFITS
This section presents an overview of the discussion on benefits of
water pollution control contained primarily in Chapter IV. Again, spe-
cific references to this chapter are provided wherever appropriate. The
topics covered are: current estimates, recreation benefits, aesthetic
and ecological benefits, health benefits, production benefits, and con-
clusions and recommendations.
1. Current Estimates
The latest estimates of national water pollution damages have
been compiled by Heintz et al. (1976). These are equivalent to ben-
efits that would accrue annually from reduction of water pollution
to "threshold" levels. Table 1-3 summarizes these estimates for
four major classes of damages, in terms of best estimates and the
corresponding ranges. The latter reflect the substantial uncer-
tainty associated with £hese estimates.
Table 1-3. Estimated National Damages of Water Pollution for 1973
($ billions)
Damage Category
Outdoor Recreation
Aesthetic and Ecological
Health
Production
(including municipal ,
industrial , agricultur-
al supplies, commercial
fisheries, and materials
damage)
Total
Best Estimate
6.3
1.5
0.6
1.7
10.1
Low
2.5
0.6
0.3
1.1
4.5
Range
High
12.6
2.8
1.0
2.3
18.7
Source: Heintz et al., 1976
1-28
-------
Since the largest damages are in the recreational, aesthetic,
and ecological categories, the total estimates strongly reflect the
uncertainties inherent in the currently available data and techniques
for estimating the monetary value of these uses of water. The esti-
mation of recreation damages depends'upon understanding the behavior
of recreationists when confronted with a complex set of choices con-
cerning types of recreation, travel, sites, and the quality of the
recreational experience. In determining aesthetic and ecological
damages, it is necessary to consider the value which individuals
place on viewing or preserving waterways of high quality.
A broad spectrum of studies have been reviewed to obtain infor-
mation on water pollution control damages. Table 1-4 reports on the
availability and reliability of the information contained in these
studies.
Table 1-4. Availability and Reliability of Infor-
mation on Water Pollution Damages
Pollutant
Acidity
BOD
Col iform Bacteria
Floating Solids
Hardness
Nutrients
Odor
Oil
Pesticides
Sediment
Temperature
TDS and Salinity
TSS and Turbidity
Toxic Metals
General Pollution
Outdoor
Recreation
SF
IF
SF
IF
U
SP
U
SP
U
IF
IF
U
SF
SP
AF
Aesthetic and
Ecological
U
SP
U
U
U
U
U
SP
U
IF
SP
U
U
U
SP
Health
U
U
SP
U
U
U
U
U
U
U
U
U
U
U
SP
Production
Losses
SP
IF
SP
SP
IP
SP
U
SP
U
IF
SP
IF
SP
SP
IF
Source: Heintz et al., 1976 Availability:
A - ample
I insufficient
S scarce
U - unavailable
Reliabi1ity:
E excellent
G good
F - fair
P - poor
1-29
-------
2. Recreational Benefits
Estimates of recreational benefits are based largely on
studies of the behavior of recreationists faced with changing wa-
ter quality. The analysis attempts to predict how recreational
patterns and the resulting values gained by recreationists would
be affected by changes in water quality parameters. Unfortunately,
damage functions relating these changes are very sketchy. (IV
11, 20).
Most national benefit estimates are based on a few partial
studies of local recreational patterns. This presents two sour-
ces of uncertainty. First, there is considerable variation in
the relationship between water quality and consumer behavior from
one geographic location to another. Second, local studies cannot
account for the large number of options open to the recreationist,
i.e., how deterioration of water quality at one site affects his
behavior at other sites. (IV 19-20).
Two other points are worth noting. Most studies consider only
two pollution levels: the current and the one meeting water qua-
lity standards. Instead, a range of pollution levels should be
examined to determine that level where marginal costs become equal
to marginal benefits. In addition to the usual efficiency grounds,
improvement of water quality for recreation could be justified on
other grounds, such as compensation to those who can not afford to
recreate elsewhere, preservation of future options, and national
pride. (IV 20-21).
Water pollution affects recreation by reducing the quality
of the recreational experience, by causing the consumer to re-
create less often, and by causing him to incur higher travel costs
to enjoy substitute sites. The last of these components is mea-
sured most readily. However, the use of travel costs for this
purpose is susceptible to the following complications (IV 21-22);
1-30
-------
• Unusual income elasticities
• Impact on travel of highway construction,
crowding, and gasoline prices
• Inertia in recreational preference
t Transfer of recreation from a more dis-
tant, but more desirable site to a
closer site, as a result of pollution of
the distant site.
Two conceptual models estimating national recreational bene-
fits by aggregation of site and situation-specific results have
been developed. The first examines the conceptual links between
quality, quantity, and price of the recreational experience. The
second model demonstrates the usefulness of a "reduction in travel
cost" construct and specifies data requirements. (IV 23-24, Attach-
ment A and B).
3. Aesthetic Benefits
The state of the art in estimating aesthetic and ecological
benefits is even more limited than that for estimating the recrea-
tional benefits. Effect of water quality on aesthetic and ecologi-
cal concerns and the resultant consumer behavior modification are
difficult to measure, because they are not well reflected in market
transactions. Consequently, such estimates are based on personal
interviews which seek to determine the value placed by consumers
on these concerns. It is widely suspected that the interviewees
are attempting to affect the results of the interview by their re-
sponses, i.e., that there is a substantial Disparity between what
they say they would do and what they would actually do. Although
psychologists and other behavioral scientists have been able to
cope with this problem, their expertise has not been brought to
bear on these studies. (IV 25).
It is difficult to draw general conclusions from current empi-
rical work, because it focuses on high-quality, unique environmen-
tal services, rather than medium or low-quality services. Moreover,
1-31
-------
there is some question whether estimates of option values should
be incorporated in benefit estimates, in light of their softness
and consequent weakening of the validity of total benefit esti-
mates. (IV 26-27).
Changes in property values have been .employed to estimate
aesthetic benefits of water pollution control. These changes
have been regarded with some suspicion, because it was felt that
they reflect a redistribution of wealth, rather than a net addi-
tion to national productivity or efficiency benefits. On the
other hand, property value changes may reflect an actualized op-
tion demand which should be included as an efficiency benefit, be>
cause it indicates that buyers seek to insure a continued future
supply of environmental services for private use. (IV 27-28).
The use of property value changes to estimate the aesthe-
tic benefits suffers from several limitations. One is the diffi-
culty of avoiding double counting of benefits with those in the
recreational category. Moreover, there is a general scarcity of
property value data which can be adequately linked to water qua-
lity levels. (IV 28).
4. Health and Production Benefits
Improvements in estimating health benefits hinge on a more
precise definition of the relationship between water quality
and the incidence of waterborne diseases, such as hepatitis, gas-
troenteritis, salmonellosis, and typhoid. The key problem in
quantifying health benefits is how best to deal with uncertain,
long-term relationships, such as the' presence and effects of car-
cinogens. Also, determination of the value of human life is
complicated by the fact that society may be willing to expand the
effective willingness to pay of an individual beyond his future
earnings. (IV 16-17).
1-32
-------
Production benefits 'aYise from water pollution control, be-
cause water is a direct input to many production activities. These
activities may be divided to include municipal, industrial, and ag-
ricultural water supplies, commercial fishing, and materials damage
categories. Available estimates of these benefits have been made
with the aid of the alternative cost, or technical coefficient ap-
proach. This approach is based on estimating the additional cost
incurred in these production activities, because of water pollution
and assumes that the reduction in such costs would be proportional
to the reduction in pollution levels. The costs addressed include
treatment of water before-use, maintenance and repair of equipment,
and decreased production. (Heintz et a!., 1976).
Production benefit estimates obtained under this approach suf-
fer from the failure to consider substitutions and other market ad-
justments made by the producer to mitigate the effects of pollution
on production activities. Another problem lies in the assumption
that marginal costs of treating man-made pollutants are equivalent
to the -prorated portion of the average cost of treating both man-
made and natural pollutants. This may lead to over estimation of
the benefits due to control of man-made pollution. A problem in
the estimation of agricultural damages on the basis of crop prices
is due to distortions brought about by price support for farm pro-
ducts and other institutional manipulations. (IV 18-19).
5. Conclusions and Recommendations
The following conclusions are drawn from the preceding dis-
cussion:
t The theoretical framework for estimation of water
pollution control benefits is far more advanced
than the corresponding empirical work (IV 29)
• Estimates of national benefits of consumption
losses (e.g., recreation, aesthetics) are much
less firm than those for production losses and
prospects for improvement of this situation are
not good (IV 29)
1-33
-------
• Reduction of travel costs offers a- potential for as-
sessing recreational benefits at a higher level of
aggregation (IV 30-31)
The following actions aire recommended to improve estimation
of water pollution control benefits:
t Prepare more complete and valid behavioral response
functions, including demand curves for specific re-
creation sites (II 50; IV 30)
• Establish a long-term "behavioral monitoring system"
that would measure simultaneously basic economic and^
ecological factors, objective and perceived water qua-
lity parameters, recreational activity, and property
values (IV 31-32)
• App],y the new models to the estimation of recrea-
tional benefits (IV 23-24, Attachments A and B)
• Improve interview techniques for estimating aesthetic
benefits by seeking help from behavioral scientists
(III 35-36; IV 25)
t Prepare improved damage functions defining the
relationship between water quality and the inci-
dence of waterborne diseases (IV 16)
• Prepare improved estimates of the value of human
life (II 50; IV 17-18).
1-34
-------
REFERENCES
Heintz, H. T., Jr., A. Hershaft, and G. C. Horak, National Damages
of Air and Water Pollution Control, U.S. Environmental Protec-
tion Agency (Contract No. 68-01-2821), September 1976.
Hershaft, A., et al.. Critical Review of Air Pollution Dose-Effect
Functions, Council on Environmental Quality (Contract No. 5-
AC012). March 1976.
Takacs, I., and G. B. Shea, Human Population at Risk to Various^
Levels of Air Pollutants. U.S. Environmental Protection Agency
(Contract No. 68-01-2820), February 1975.
Unger, S. G., et a!., National Estimates of Water Quality Benefits,
U.S. Environmental Protection Agency (Contract No. 68-01-2997]",
November 1974.
Waddell, T. E., The Economic Damages of Air Pollution, U.S. Environ-
mental Protection Agency, EPA-600/5-74-012, May 1974.
1-35
-------
II. BENEFITS OF POLLUTION CONTROL
A. Myrick Freeman, III
Bowdoin College
A. INTRODUCTION
The issues to be discussed in this paper can be stated most con-
cisely in the form of four questions. First, "would national estimates
of benefits from controlling air and water pollution be useful to public
officials and others concerned with evaluating pollution control policy
alternatives?" Second, "is it possible to use the available economic the-
ory, empirical techniques, and data base to make national estimates which
can be supported on logical and scientific grounds?" The third and fourth
questions presuppose generally affirmative answers to the first two ques-
tions. The third is, "do available estimates of national pollution control
benefits adequately reflect the present state of the art?" And finally,
"what can be done to produce better estimates of national benefits?"
It is perhaps misleading to present the "possible" and "useful" ques-
tions as if they were entirely separable. The way the questions are asked
implies the possibility of a simple yes or no answer. But what is "possi-
ble" depends upon the resources and effort devoted to the exercise of mea-
surement. And this in turn depends upon the expected usefulness or utility
of the information being sought. The importance of the question and the
likelihood that more information will improve the decision determine the
value of additional information such as estimates of national benefits.
Policy makers should seek more information on national benefits as long as
the marginal value of the improved benefit measure exceeds the marginal re-
source cost of obtaining it.
Section B continues the discussion of the possibilities and usefulness
of national estimates of benefits in somewhat more detail. While the ques-
tions posed at the outset focus primarily on national estimates of benefits,
it is not possible to provide complete answers without some reference to
the problems and possibilities for estimating benefits for individual pollu-
tion control projects, river basins, and air sheds. Generally, national es-
timates of benefits involve some form of extrapolation or aggregation based
II-l
-------
on parameters and relationships derived from more narrowly focused studies
of a particular region or pollution control project. Of necessity then,
the evaluation of the possibility of national estimates must consider the
M» \ .
theory and practiced benefit estimation in general. This is the top.ic
of Section C.
Secion D discusses three conceptual problems in benefit estimation.
The representation of uncertainty through the use of probabilistic state-
ments and confidence intervals is recommended. The willingness to pay
versus forgone earning approaches for valuing human life are discussed.
The latter seems to be unsatisfactory on both conceptual and empirical
grounds. In the absence of definitive estimates of willingness to pay,
it is suggested that decision makers experiment with explicit assumptions
or value judgments about the value of life. The problem of temporal ag-
gregation of benefit streams over time is discussed in terms of the prin-
ciple of discounting and present value. The problem of intergenerational
effects seems to raise questions which are probably fundamentally ethical
in nature.
Section E reviews a number of problems in the empirical techniques
and data for benefit estimation. Topics covered include overlap and gaps
among categories, the effect of multi-collinearity on measurement and pre-
diction, -the biases due to misspecification of variables, and the problem
of adequately summarizing air pollution data collected over a substantial
interval of time. The section concludes with a review of the adequacy of
available data and technique for dealing wHh air pollution data, property
value effects, health effects of air pollution, recreation benefits, and
benefits from increased industrial productivity and reduced vegetation and
materials damages.
Section F discusses the process of developing national estimates of
benefits from the building blocks of region-specific, pollutant-specific,
and effect-specific studies. Specific suggestions are made for estimating
national mortality changes from cross-section regressions of mortality ver-
sus air pollution and aggregating the information contained in property value!
studies for individual cities. Also the problem of national recreation benel
fit estimation is discussed.
II-2
-------
Section G makes a number of specific recommendations for improving
estimation techniques and for additional research. The final section is
devoted to some concluding observations.
II-3
-------
B. POSSIBILITY AND USEFULNESS OF NATIONAL BENEFIT ESTIMATES
The feasibility, or possibility, and usefulness of national benefit
estimates are the cardinal issues in this critical assessment. They are
examined here along with the corollary issue of information content.
1. On Possibility
Since the nation is simply the aggregation of its regions, how-
ever! defined (as air sheds, river basins, or political subdivisions),
the possibility of developing national benefit estimates hinges on
our ability to make estimates of particular classes of benefits in a
specific region. Data from particular regions constitute the building
blocks of national estimates. If nothing meaningful can be said about
the benefits of a specific pollution control plan in a particular re-
gion, then no meaningful aggregation to the national level is possible.
Estimating benefits is essentially an ecpnomic process. The dis-
cipline of economics is the source of the concept of benefits; and the
definition of benefits and the conceptual basis of their measurement
are derived from a type of economic model. Yet benefit estimates must
be built upon a non-economic base. For example, health benefits due
to air pollution control must be built upon scientific knowledge of
the relationship between pollutant concentrations and human health.
In recreation and fisheries, benefits stemming from water pollution
control may require knowledge of the relationship between pollutant
levels and biological productivity. It is clear that lack of informa-
tion in some non-economic areas severely limits our ability to make
estimates of economic benefits. But aside from acknowledging the exis-
tence of this problem, no more will be said here. Rather our atten-
tion will be devoted to the economic aspects of benefit estimation.
Suppose the Administrator of EPA wished to know the magnitude
of the benefits to be expected from a particular pollution control
program. If asked, could the economist specify what economic the-
II-4
-------
ory and models he would use, what data he would like to have, and
what empirical techniques he would apply to the data to obtain ben-
efit estimates? The answer is a qualified "yes". After completing
a review of the available techniques for measuring the benefits of
water pollution control for the U.S. Environmental Protection Agency,
I concluded: "The economic theory and analytical techniques for
evaluating benefits for most types of uses are relatively well devel-
oped. At least, this is the case for the most significant classes
of user benefits (Freeman, 1975)."
The qualifications are threefold. First, some of the data the
economist would like to have is non-economic, e.g. dose-response
functions, physical damage functions, and these data may not be
available. Second, in the economic realm there are still major un-
resolved theoretical and empirical issues with respect to valuing
human life and health effects and various forms of non-user benefits
such as aesthetics and option value. And third, the kind of economic
data that the economist would 1 ike to have are seldom already avail-
able or easily obtainable at modest cost. Major data gathering ef-
forts would be required to obtain the primary economic data called
for by the correct theoretical models. Lacking data in the ideal
form, the economist must resort to second best theoretical constructs
and utilize sophisticated econometric techniques in an attempt to
wrest some modicum of useful information from the limited and imper-
fect body of data.
Because of the several qualifications, successful benefit esti-
mation, either at the local or national level, requires striking a
balance between rigorous theory and correct empirical technique on
the one hand and pragmatism and informed judgment on the other.
Where there are gaps in knowledge concerning key parameters and rela-
tionships, and where it is not possible to devote resources to clos-
ing these gaps within the scope of the existing research effort, it
may be appropriate to make an informed assumption about the unknown
relationship in order to push the analysis through to its conclusion.
II-5
-------
This procedure is valid provided that: the assumption is made ex-
plicitly and openly; the basis for choosing the assumed value ra-
ther than some alternative is fully discussed; and the implications
of alternative assumptions are made clear through sensitivity analy-
sis on the outcome or carrying confidence limits through to the fi-
nal outcome. The necessity for making assumptions may stem from ig-
norance concerning empirical relationships, e.g., dose-response re-
lationships for air pollutants, or inherent problems of evaluation,
e.g., the valuation of human life. These issues are discussed in
more detail in a recent National Academy of Sciences Report to U.S.
EPA (National Academy of Sciences, July 1975, esp. Chapter 6, and
Appendix 8).
If the practice of benefit estimation were highly developed,
the estimation of benefits at the national level would proceed sim-
ply by estimating benefits separately for each region or basin, and
then summing them over the nation as a whole. But this is clearly
not the case. The real question for policy analysts is how far can
one depart from the ideal of rigorous and theoretically precise es-
timation' for all sites, moving toward the bastardized but pragmatic
process of first approximation, reasonable assumption, and judgment,
and still obtain estimates which are considered reasonable and use-
ful for the questions at hand. It is hard to provide an answer to
this question in the abstract. As was indicated in the introduction,
this depends in part on the resources available to support adequate
investigation, and in part on the kinds of policy questions being
posed and the value of additional information.
2. On Usefulness
Useful to whom? To officials responsible for environmental
policy planning and decision making. Benefit estimates are useful
if they provide an improved information base for environmental de-
cision making. Achieving air and water pollution control object-
ives established by Congress will involve massive expenditures both
through the public sector and private sectors. There is a poten-
tial for substantial gains through more effective utilization of
II-6
-------
the resources devoted to pollution control via the judicious use
of benefit-cost analysis in evaluating policy alternatives. It
must be emphasized that benefit-co'st analysis as we use the term
is not a rule for decision making. Environmental decision makers
may have other objectives besides the simple economic efficiency
objective which underlies the benefit-cos-t rule. For example de-
cision makers may be concerned with equity considerations, inter-
generational effects, or risk aversion, none of which can be incor-
porated in a simple equation for decision making purposes. Also,
it is seldom feasible to provide the full range of complete and
accurate benefit and cost information to the decision maker which
would be required for the unqualified acceptance of the benefit
cost rule. Rather, questions of uncertainty and information gaps
are 1ikely to exist.
Benefit-cost analysis can be very valuable as a framework and
a set of procedures to help organize available information. In this
way, benefit cost analysis does not dictate choices, nor does it re-
place the ultimate authority and responsibility of decision makers.
Rather, it is a tool for organizing and expressing certain kinds of
information on the range of alternative courses, of action. It is in
the context of this framework for arraying information that the use-
fulness of national .benefit estimates must be assessed.
Although the major thrust of Federal pollution control policies
has already been established by the U.S. Congress, there are still
many important decisions to be made. Also, looming problems may
force reconsideration of major elements of existing policies. Most
of these issues involve either an implicit or explicit weighing of
benefits versus costs. For example, U.S. EPA and U.S. Congress are
weighing proposals to change the emissions standards and/or the time-
table for compliance for automobiles. National Secondary Air Quality
Standards require at least an implicit balancing of benefits versus
costs. And to defend effluents standards based upon the best practi-
cal treatment and best available treatment criteria against legal
challenge, U.S. EPA will have to have information on the benefits to
II-7
-------
be expected from the standards they are setting. There are two
things to note from this sampling of future U.S. EPA and Congress-
ional decision problems. First, while decision makers may have
other objectives besides economic efficiency, it is clear that eco-
nomic considerations, broadly defined, will play a significant role
in the decision making process. And second, each of these decision
problems is national in scope and requires national estimates of
benefits and costs.
This discussion has established the potential value of national
estimates of benefits. But it must be recognized that, under some
circumstances, benefit estimates can be damaging rather than useful.
If estimates succumb to the fault of misplaced concreteness, if they
contain conceptual or analytical errors, if they do not make expli-
cit the key assumptions and value judgments involved, and of they do
not express their inherent lack of accuracy through some device such
as confidence limits, the estimates may do more harm than good. They
may mislead policy makers in key decisions. And as their faults and
weaknesses become known, they may discredit the whole process of bene-
fit estimation and policy analysis.
Finally, even if benefit information does not become an input
into a decision process, the exercise in quantification and measure-
ment that benefit estimates require may be itself of educational
value to researchers and policy makers. The exercise requires that
investigators gain a better understanding of the nature of different
pollutants, the paths that pollutants follow in reaching people, and
the variety of ways in which pollutants affect the uses that people
make of their air and water environments. Even if the exercise does
not bear fruit in the form of monetary estimates of benefits, the in-
sights gained and the pieces of information developed may be of value
themselves.
3. Information Content
A major factor influencing the usefulness of estimates of na-
tional benefits is the nature and quantity of specific information
II-8
-------
contained in the estimates. A statement that "air pollution costs
$20 billion per year" contains almost no useful information. To be
useful for policy purposes, any statement about national benefits
must abide by the following minimum provisions:
• Consistent terminology
• Well-defined pollutant levels
• Identity of pollutants
« Sources of pollutants.
,'
The statement must be consistent in its use of the terms "bene-
fit," "cost," and "damage". The benefits of pollution control are
measured by comparing the present situation characterized by a known
degree of pollution with some specified hypothetical alternative for
which pollution has been reduced or eliminated. Benefits are the
gain associated with reduced pollution. The statement of benefits
should make clear whether the measure refers to total benefits for
some non-marginal change in pollution levels or marginal benefits
for some small change in pollutant levels. The statement should
also indicate whether benefits are defined per capita or in the ag-
gregate.
Damages represent the mirror image of benefits measuring what
is lost in moving from some hypothetical clean state to the existing
level of pollution. Pollution control benefits are the same as re-
duced damages. The term "costs" should be used only to refer to re-
sources absorbed in the process of controlling or reducing pollution.
Its use to denote damages should be avoided.
The statement concerning benefits must clearly specify the al-
ternative pollution levels being compared, both the present dirty
level and the hypothetical cleaner alternative. Is the polluted
state defined as conditions in 1975, or for example, 1973 (when
pollutant levels and adverse effects may have been either higher
or lower)? Specifying the reference year is also important for
II-9
-------
determining the price level for expressing dollar measures of bene-
fits (e.g., in 1973 dollars), and for specifying values for other
socioeconomic variables which might influence the analysis of bene-
fits.
Four choices for the hypothetical alternative seem to present
themselves:
• Zero ambient concentrations of pollutants (this may be
unrealistic where there are natural sources of the pol-
lutant in question)
t Zero manmade emissions -- that is, ambient levels equal
to background levels from natural sources
t Ambient levels above background levels but below all
known thresholds (this alternative would be difficult
to implement, because of the considerable controversy
over threshold levels)
0 Ambient levels at promulgated air or water quality
standards (this is the most useful alternative for
evaluating present pollution control policies).
The estimates should make clear which pollutant or group of
pollutants is being referred to. Where the benefit estimates are
provided for a group of pollutants (e.g., all air pollutants), the
statement should include a disaggregation or breakdown by individual
pollutants, unless the pollutants within the group act synergisti-
cally, so that such a separation of effects is not analytically pos-
sible. Benefit estimates for a group of pollutants may also be use-
ful where the group is controlled jointly.
For each benefit estimate,the statement should indicate whether
all sources are included, or, for example, only mobile sources.
Where the purpose of the benefit estimate is to evaluate a pollution
control policy, benefit estimates should exclude potential benefits
associated with non-controlled sources. For example, pollution con-
trol benefits associated with the control of non-point water pollu-
tion sources should not be included in an estimate of benefits asso-
ciated with municipal and industrial point-source control program.
11-10
-------
The kind of information which should be contained in a state-
ment of benefits also depends upon the nature of the policy ques-
tion for which the estimate was made. For example, in the early
stages of the development of a pollution control policy, the most
important question might be one of priorities, i.e., which classes
of pollutants should be controlled most quickly. For this purpose,
very rough order-of-magnitude estimates of total benefits by pollu-
tant would be very helpful. Later, as control plans are developed,
marginal benefits, and benefits by source would be required.
The main thrust of Federal policy since 1970 has been away from
regionally differentiated air and water quality standards and toward
uniform national ambient standards and national standards for emis-
sion levels and control technologies. This evolution of a nationally
standardized set of policies has increased the importance of national
estimates of benefits. The process of setting national primary air
quality standards would be aided by national estimates of marginal
benefits by pollutant. And the development of pollution control
strategies requires national estimates of benefits by source.
In conclusion, the type of policy question being considered
governs the nature and detail of the benefit information required.
Planning on a regional basis requires estimates of regional benefits
and costs. National policy choices require national estimates. And
since the building blocks of valid national estimates of benefits are
regional studies of specific pollutants, once_these studies are done,
their aggregation to the national level is relatively simple. And
the effort to get better national estimates will stimulate research
on obtaining better building blocks.
11-11
-------
C. THE THEORY OF BENEFITS
This section reviews the conceptual b'asis of benefits and some em-
pirical techniques for estimating benefits. These topics have been trea-
ted extensively in other studies for the U.S. EPA (Haveman and Weisbrod,
1975; Freeman, 1975).
1. Efficiency and Equity
By definition, benefits are desirable consequences. But we
require some criterion for defining and measuring desirability.
Welfare economics has conventionally distinguished between effi-
ciency and equity as criteria for judging alternative economic
states. Efficiency refers to the total availability of those
goods and services valued by individuals without reference to whom
they accrue. Equity refers to the pattern of the distribution of
available goods and services among individuals. What constitutes
equity or fairness in the distribution of goods and services can-
not be defined objectively. Equity involves value judgments about
the relative "deservingness" of individuals. For this reason, eco-
nomic analysis focuses on the efficiency criterion on the argument
that it is objective or value-free.
The efficiency criterion requires a rule for assigning values
or prices to goods and services so that they can be added up to de-
termine a total value. The price system'of a market economy pro-
vides the necessary set of relative values, provided markets are
competitive, there are not externalities, etc. These prices would
appear to constitute the set of values necessary for the implemen-
tation of an objective efficiency criterion, and to validate the
dichotomy between efficiency and equity. But this is not the case.
And this is one of the insoluble problems in the analysis of benefits.
The price set is not independent of the initial distribution of wealth
among individuals in the economy. Thus, if the price set is taken to
be valid as a measure of values, this implies that the underlying
distribution of wealth which produced those values is acceptable in
equity terms. A redistribution of wealth will lead to a different
11-12
-------
set of relative prices. And an allocation of resources judged effi-
cient under the initial distribution would likely be inefficient un-
der the new distribution. Which measure of efficiency is to be pre-
ferred depends upon one's value judgment about the equity of the two
wealth distributions. There is no way around this problem, so it is
typically ignored in the practical business of applied welfare analy-
sis. In other words, the existing distribution of wealth and its
resulting set of relative prices are assumed to be acceptable in
equity terms.
2. Measurement of Benefits
Benefits are defined in terms of the increase in welfare or
utility of consumers stemming from some action to control pollution.
Benefits can accrue to consumers either directly in the form of in-
creases in the availability of utility conveying goods and services
(e.g., improved health, better visibility, improved recreation oppor-
tunities), or indirectly through increases in production efficiency
(for example, due to reduced defensive expenditures, reduced mater-
ials damages, etc.). Pollution control involves the use of limited
resources which have opportunity costs measured in dollars. This
means that the benefits must also be defined and measured in com-
mensurate dollar terms. Furthermore these dollar measures must be
firmly rooted in and based upon individual preferences and values.
Economic theory shows that there are two ways in which mone-
tary measures of welfare changes can be derived from individuals'
preferences. The two approaches differ only in their definition of
the status quo ante^ from which the welfare change is measured. One
approach, termed the willingness-to-pay approach, asks how much an
individual would be willing to pay to receive an extra quantity of
the utility-conveying good or service, rather than do without. It
assumes that the individual does not have a legal property right in
additional units of the good, and therefore must give up control over
other goods and services (pay) in order to receive more of this good.
The other approach, termed the compensation approach, asks how much
the individual would have to be compensated in money terms in order
11-13
-------
to induce him to give up or forego some quantity of the good. This
approach assumes that the individual has a legal property right in
the good, and could only be induced to give up the good if the
amount of the compensation conveyed an equivalent amount of utility.
(National Academy of Sciences, 1974, pp. 400-404).
The two approaches lead to quantitatively different answers
only if the marginal utility of income to the individual changes as
the individual moves from one position to another. If the marginal
utility of income is diminishing, the compensation approach will
lead to a higher measure of benefits than the willingness-to-pay
approach. Where the marginal utility of income is constant, or
can be assumed to be approximately constant (for example, where
the changes are ve>y small relative to total income), the two ap-
proaches can be taken to be equivalent.
The benefit associated with a particular pollution control
action is defined as the sum of the willingness-to-pay of all in-
dividuals affected directly or indirectly by that action. Analy-
tically, willingness-to-pay implies the existence of a demand curve
for the consequences of that action. Benefits are equal to the area
under that demand curve, properly defined. Benefit estimation in-
volves an attempt, either indirectly or directly, to determine the
shape of that demand curve. This definition of benefits meets two
1 important criteria for an adequate framework for benefit estimation:
'the definition is firmly rooted in the economic theory of individual
preferences and values and it leads to objectively determined mone-
tary measures of welfare or benefit.
3. The Estimation Process
Although the definition of benefits focuses on the existence
of a demand curve, its application in the realm of pollution control
must also deal with two other problems associated with the "produc-
tion" of benefits. The process of producing pollution control bene-
fits has three distinct stages:
11-14
-------
t Improvement in environmental quality through re-
duction of pollutant discharges
• Increase in uses of environmental amenities due
to their improved quality
§ Increase in willingness to pay due to changes in
use of environmental amenities.
A reduction in the quantities of polluting substances being
discharged into the air or water leads to an improvement in various
measures of air or water quality. Modeling this transformation pro-
cess between changes in residual flows and changes in ambient pol-
lutant levels is outside of the realm of economics. Yet it is a
task which must be done before the benefits of pollution control
projects can be estimated.
Changes in ambient pollutant levels lead to changes in the
ways in which individuals make use of the air or water, that is,
changes in the level and composition of the stream of environmental
services yielded by the air and water bodies. For example, changes
in ambient air pollutant levels may lead to changes in morbidity and
mortality rates. Understanding these relationships is an essential
part of the estimation process. This might be termed the measurement
process to distinguish it from the process of valuation, which consti-
tutes the next stage.
Changes in uses or service flows have their counterpart in
changes in aggregate willingness to pay for uses of the environment.
Determining the relationship between changes in quality and changes
in use is only a partly economic problem. Determining the relation-
ship between use and value is strictly the economist's domain, since
here the economic concepts of demand and value are central to the
analysis.
The benefit associated with a change in air or water quality
has been defined as equal to the area under the demand curve for
the environmental service over the relevant range. But, since en-
vironmental services are generally not bought and sold in organized
11-15
-------
markets, there may not be any direct way of estimating the shape of
this demand curve by applying normal econometric techniques to ob-
served market data. The basic methodology used for estimating bene-
fits involves techniques for inferring values of environmental ser-
vices where market processes have failed to reveal these values di-
rectly. Thus, benefit estimation often involves a kind of detective
work for piecing together the clues about the values individuals place
on these goods as they respond to other economic signals.
There are several different situations where the way in which
environmental services enter economic processes permits estimation of
willingness to pay for improvements in environmental quality by indi-
rect means. In an earlier study for the U.S. EPA (Freeman, 1975), I
analyzed the economic theory underlying several types of cases, where
environmental services were inputs in production, where environmental
services were complementary to other goods and consumption, where en-
vironmental services were perfect substitutes for other goods in con-
sumption, and where environmental quality levels were characteristics
of fixed assets such as land. In that study, I also reviewed a number
of techniques, e.g., market demand analysis, net productivity, net fac-
tor income, Clawson-Knetsch travel costs studies, land value studies,
and interview techniques, in light of the earlier discussion of eco-
nomic theory. Each of these techniques (except the interview techni-
que) relies on some model of individual behavior and some set of as-
sumptions about the nature of the benefits being analyzed to deduce
a relationship between the unobserved values individuals assign to
environmental benefits and related observable market s.ignals.
11-16
-------
D. CONCEPTUAL PROBLEMS IN BENEFIT ESTIMATION
There are a number of conceptual issues that need to be addressed
in the process of benefit estimation. Among the most pressing of these
are valuation of human life, handling of overlaps and gaps, aggregation
of the stream of benefits over time, and expressing uncertainty of the
results.
1. Valuation of Life
A major category of potential air pollution control benefits
is improved health and reduced mortality. Assuming that these
benefits can be measured or quantified, how can monetary values
be attached to them? Some might say that putting a price on hu-
man life is insensitive, crass, and even inhuman. But this point
of view must be rejected on practical grounds because both indi-
viduals in their day to day actions and governments in their deci-
sions about social policy do in fact make tradeoffs between changes
in probability of death and other goods with money values; and these
tradeoffs imply money price tags being attached to life.
Individuals in a variety of actions act as if their preference
functions included probability of death or life expectancy. They
make decisions which involve reductions in life expectancy in re-
turn for increases in income or other goods and services, reveal-
ing, thereby, that they perceive themselves to be better off having
made these choices. Individuals who continue to smoke despite the
Surgeon General's warnings could be viewed as trading off the con-
sumer surplus attached to smoking against the now widely publicized
reduction in life expectancy. Individuals also make implicit trade-
offs between life expectancy and reduced travel time when they chose
a mode of travel, for example private autos vs. bus. The individual
willingness to pay framework has been developed by Schelling (1965),
Mishan (1971), Zeckhauser (1974), and Thaler and Rosen (1975). Thaler
and Rosen in particular have developed a framework for estimating the
individual tradeoff between probability of death and higher wages in
11-17
-------
more risky occupations; and they have tentatively estimated the
implied value of human life to be between $176,000 and $260,000
per life. Although this estimate cannot be taken as definitive,
it does provide a point of reference for further work in the area
and for assumptions and value judgments as described below. (Na-
tional Academy of Sciences, 1975, Chapter 6).
What should be the basis for determining these values? The
conceptually correct basis for valuation is individual willingness
to pay. But this willingness to pay must be defined and evaluated
in the correct context. The question is not how much would one be
willing to pay to avoid certain death tomorrow. Rather, for the
relevant policy questions, e.g., health effects of pollution con-
trol, safety and accident prevention, the precise identity of the
people affected is not known. Safety regulations reduce the prob-
ability that any one individual in a population at risk will be
killed during a given period of time. It is this kind of question
that Thaler and Rosen investigated. Reductions in air pollution,
for example, are likely to affect the whole probability distribu-
tion of the date of death for affected individuals. One measure
of this change in the probability distribution would be to define
the change in life expectancy. The changes in higher moments of
the probability distribution might also be important, especially
for things such as accident prevention. (Freeman, 1975).
The other major approach to valuation in the literature is
the so-called productivity or human capital approach. It values
each life lost at the present value of the expected stream of fu-
ture earnings for that individual had that individual's death been
avoided. There are several serious weaknesses and limitations to
this approach. First, it assumes that it is possible to identify
ex ante those whose lives will be saved. Second, it does not take
into account the probabilistic nature of death and death avoidance
in the health and safety problems. Third, there is no logical con-
nection between earnings and willingness to pay, especially in the
11-18
-------
probabilistic framework. And finally, this measure places no value
on the lives of those who are not working for reasons of age, sex,
or other factors.
While lost earnings can be readily calculated, that approach
is conceptually unsatisfactory. In contrast, the conceptually va-
lid willingness to pay for changes in life expectancy approach is
appealing,, but it is only now being elaborated and subjected to
empirical testing. In the absence of generally accepted empirical
estimates of individual willingness to pay, it is still useful for
the researcher to proceed beyond the measurement stage in benefit
estimation by applying assumed unit values to predicted changes in
mortality and morbidity. In this way, the analyst can examine the
implications of alternative unit values for aggregate benefit esti-
mates, and for benefit cost comparisons. (National Academy of Sci-
ences, 1974).
However, if such assumptions are made, several points must be
kept in mind. First, the assumptions should be made openly and
explicitly. This is not the case with estimates of air pollution
health benefits provided by Lave and Seskin (1970) who fail to ex-
plain the implications of their assumptions about health values.
They used earlier data from Dorothy Rice on direct and indirect
costs of mortality and morbidity. The indirect cost of mortality
was defined as forgone earnings due to early death.
Second, researchers should provide an examination of the sen-
sitivity of final benefit estimates to changes in assumed unit
values. For example, final figures could be presented in the for-
mat of high and low range and "best estimate". A third consider-
ation is that, whatever assumptions are made about unit values, it
would be helpful for readers of these studies to have the assump-
tions related to other work. For example, if a researcher chose to
use a value substantially below that used in other work which might
be familiar to the reader, the researcher should offer some reason
for making that assumption. Finally, where assumed values are used
11-19
-------
in making resource allocation decisions, such as allocating medi-
cal services, or setting standards for air pollution or toxic
chemicals, there is much to be said for using a consistent set of
values across the whole range of public decisions.
While this discussion has been carried out in the context of
valuing human life, most of the same points can be made concerning
assumed unit values for other difficult to price environmental ser-
vices, for example recreation user days. The Water Resources Coun-
cil guidelines call for agencies to apply unit values to predicted
recreation days in calculating benefits of water resource develop-
ment. While there are problems in the application of particular
unit values (Freeman, 1975), there can be no objection to the prac-
tice in principle, provided that the considerations outlined above
are recognized. Assumed unit values can be utilized provided they
are based upon a review of other analytical and empirical studies
designed to measure unit values of recreation under similar circum-
stances.
2. Overlaps and Gaps
One major problem area in estimating benefits is the possi-
bility of either overlap among or gaps between categories of bene-
fits as estimated by presently used techniques. This kind of prob-
lem can arise where two kinds of effects associated with the same
pollutant are estimated separately. For example, do estimates of
health effects and property value changes associated with particu-
lates involve some double counting of particulate damages, or are
there additional effects not captured by either measure? Although
definitive answers to this question cannot be given here, the dis-
cussion will show that it should be possible to determine when such
gaps and overlaps exist. It should also be possible to make some
judgment as to the significance and likely direction of any biases
in the estimated total benefit figures. And this should be useful
in deriving confidence limits or high and low bounds surrounding
best estimates of total benefits.
11-20
-------
Consider the case of air pollution where health benefits are
measured by mortality rate studies and aesthetic benefits are mea-
sured through property value differentials. The justification for
measuring two classes of benefits separately and adding them to ob-
tain aggregate national benefits is that only those effects of air
pollution which are perceived by individuals can influence property
values, and that people have for the most part been ignorant of the
effects of air pollution on their health and life expectancy.
But one can feel somewhat uneasy about this simple resolution
of the problem. Some kinds of short term health problems (eye ir-
ritation, shortness of breath) may be directly perceived as being
caused by air pollution. And there has been a substantial increase
in the information (and misinformation) available to the general
public about long term health effects and relative air pollution
levels around urban areas. Thus, perceptions of health effects
may be influencing property value differentials and leading to the
possibility of double counting. To gain a better understanding of
the problem, it is necessary to consider in more detail exactly
what is captured by each of the two approaches to measurement.
It will be helpful to develop a system of classification for
different types of effects. Aesthetic effects are defined to in-
clude odor and taste, reduced visibility, soiling, damage to exter-
nal paints, etc. By definition, aesthetic effects are perceived
by individuals. Impacts on health constitute the other broad class
of effects. For purposes of this discussion, we need not distin-
guish between morbidity and mortality effects. But we can, at the
conceptual level, distinguish between those health effects which
are perceived by individuals and those of which they are ignorant.
The former are likely to be primarily clinical manifestations of
short term exposures at relatively high levels. Finally, it is
necessary to distinguish between those effects which are caused by
pollution exposure at home and those which are caused by the per-
son'^ exposure as he travels around the urban air shed, to work,
for shopping, for recreation, etc. These "away from home" effects
are independent of the individual's place of residence.
11-21
-------
This classification scheme is displayed schematically in
Figure II-l. There are six subsets of effects classified by
aesthetics, health—perceived, health-^unperceived, and in all
cases further divided between home and away. Property value
studies can only capture those effects associated with the
home which are perceived. Health benefits derived from mor-
tality and morbidity studies capture both home and away effects
and both perceived and unperceived health effects.
\
Aesthetics
Vs
sN -- Home
S
\S
Health
Perceived
Home
Aesthetics
-- Away
-- Perceived
-- Away
,
yy
y
y
/ Health
Unperceived
Home
Health
-- Unperceived
-- Away
Vvvv
captured by property value differential
captured by morbidity and mortality studies
not captured by either measure
Figure II-l. Classification Scheme for Capture of Effects
11-22
-------
The figure illustrates both an overlap and a gap. Perceived
health effects at home are captured by both the property value and
health effects approaches. But aesthetics associated with away
from home exposures are not captured by either measure. Whether
the addition of property value and health effect benefits results
in an overestimate or an underestimate depends upoh the relative
size of the doublecounted and omitted categories. Nothing can be
said about this question on a priori grounds.
There is a similar kind of problem in the case of water pol-
lution where recreation benefits and property value benefits are
both being estimated. Property value measures capture both the
value of easy access to the improved water based recreation oppor-
tunities and other non-recreation aesthetic benefits (e.g., wild-
life appreciation, elimination of odor and unsightly flotsam). In
estimating recreation benefits the double counting problem can be
avoided by identifying recreation participants as either from abut-
ting property or from away, and counting for benefit purposes only
anticipation.by those from away. This can easily be accomplished
if benefits are being estimated for a particular site, for example
by the Clawson-Knetsch method. However, if benefits are being es-
timated at the national level from a national recreation partici-
pation survey, it may not be possible to identify that part of par-
ticipation attributable to property owners using abutting water
bodies. Thus, adding property value benefits to those estimated
from a national participation model will involve some degree of
double counting.
3. Temporal Aggregation
For many pollution control programs, benefits accrue over
many years. We have discussed benefits under the assumption
that an estimate is for a given year. But for some circumstances,
we may wish to examine the whole time stream of benefits. This
would be the case, for example, if the pollution control program
involved capital costs now to produce a stream of benefits over
11-23
-------
the future. Each year's benefit is temporally distinct and not
comparable or additive to benefits accruing in other years, or
with costs without some system of weights. The appropriate
weighting system is derived from the discount rate.
There are two issues: the choice of a discount rate and
the time horizon. I see no reason to depart from the accepted
practices of project analysis on either of these issues, at least
for benefits and costs accruing over the next, say, fifty years.
(Kneese and Herfindahl, 1974).
The issues are more complex when we consider effects ex-
tending over very long periods of time., for example between gen-
erations. Discounting with the normal discount rate can reduce
even catastrophic distant future effects to nominal present values.
For example, with a relatively low discount rate of 5%, the benefit
of avoiding some number of deaths 200 years in the future is re-
duced by a factor of 0.0006. On the basis of conventional discount-
ing, the benefits of preventing some future catastrophe, for example
the effects of possible destruction of the ozone layer, may seem
less than the present costs of preventing the effect. It appears
that in cases of this sort, the major consideration is not inter-
temporal efficiency in resource allocation, but rather intergenera-
tional equity in the distribution of welfare. Hence, it should not
be surprising that decision rules based on the efficiency criterion
should appear to give unreasonable results. This is a problem which
is beginning to receive attention in the literature. But as yet,
there are no generally accepted answers.
4. Uncertainty of Results
A major problem area is the development of a framework and a
language for conveying the uncertainty about the accuracy of bene-
fit estimates. Even the most careful estimate of benefits contains
inaccuracies because of errors in the measurement of variables and
errors in the statistical estimation of relationships. In addition,
it may be necessary for the analyst to make assumptions regarding
11-24
-------
unknown parameters, relationships, and values. The uncertainty
inherent in these errors must be expressed somehow in the final
benefit figure. The most appealing way of dealing with these
problems is through the use of probabilities to quantify uncer-
tainties. This is straightforward in theory because probability
is a natural language for describing uncertain situations. (Na-
tional Academy of Sciences, 1974, Appendix 8).
Most people are accustomed to using probability informally
as a language for describing uncertainty. Weather forecasts are
stated in terms of probability of rain. And expectations concern-
ing elections and sporting events are often expressed in terms of
odds or probabilities. The same concepts may be used to describe
uncertainties related to benefit estimates. Probability theory
provides an unambiguous and logically consistent language to rea-
son about uncertainty. In fact, a forceful argument can be made
that any logical'process for reasoning about uncertainty is equiva-
lent to probability theory.
The same people who use probability naturally in informal
situations may be extremely reluctant to use the same ideas in im-
portant decision situations. To some extent the problem is one of
measurement. Many people are accustomed to viewing probabilities
as exact numbers based on objective evidence, for example, as de-
duced from physical symmetry or observed frequencies from large
number experiments. They are uncomfortable with the idea of using
subjective probabilities as a language for expressing their judg-
ment on a matter where the available information is limited.
Consider the air pollution/human health relationship. If we
had a large body of statistical data on the incidence of mortality
over a range of known air pollution exposures, it would be rela-
tively straightforward to assign a probability to a given indivi-
dual's dying as the result of exposure to that air pollutant. But,
if estimates of benefits are desired before this type of data is
available, or if those data are very difficult, to obtain because
11-25
-------
of cost, measurement problems, etc., then the analyst must resort
to informed judgment and logical reasoning. Judgment must be ex-
trapolated from other information, for example, animal tests, or
a review of several less than perfectly successful attempts to
measure the desired relationship.
Once the basic assumption has been accepted that probability
assignments are not "objective" but represent judgments, that they
can summarize information or a state of mind, rather than being
physically measurable, then the probability theory for expressing
these judgments in a consistent manner is relatively straightfor-
ward. For example, what is frequently called the "best guess esti-
mate" can be interpreted as an expected value or mean of the sub-
jective probability distribution. High and low values should be
expressed in terms of confidence limits. Where the final estimate
is derived from several variables, each with their own estimated
degrees of error or uncertainty, the uncertainty surrounding the
final estimate is a compound of the component uncertainties. And
probability theory provides a logical framework for determining
the overall degree of uncertainty.
To those who might argue that the assignment of probability
should be rejected because it is subjective, the response is that
the present approach to determining best guesses and high and low
bounds is also subjective. And often, the method of deriving best
guesses and high and low bounds from the underlying data are not
consistent from one part of the study to the next. Probability is
a language for dealing with subjective estimates In a consistent
and logical manner; and, it is one which can be interpreted consis-
tently by others.
11-26
-------
E. TECHNICAL PROBLEMS IN BENEFIT ESTIMATION
In addition to the conceptual issues discussed in the preceding
section, there are a number of more technical problems that need to•
be resolved. These include pollution measures, exposure measures,
and multicol linearity.
1. Measures of Pollution
A major problem in estimating benefits due to air pollution
arises because of systematic errors in the use of aggregate or
average measures of the air pollution variable. Consider, for
example, the cross section mortality studies of Lave and Seskin
which use SMSA's as the sample unit. Assume that there is some
single valued measure of the air pollution exposure of an indi-
vidual. Further, assume that the "true" relationship between
this pollution measure and the mortality rate is known. This is
portrayed by the solid line in Figure II-2.
Air pollution measures are typically taken at one centrally
located monitoring station in the smaller SMSAs, whereas the
larger SMSAs may have networks of monitoring stations. In the
latter case, pollution measures are some average of readings at
the several stations. Assume that for a "dirty" city the air
pollution reading at the downtown station is 0-P^. Because some
portion of the population lives in suburban areas quite far from
the downtown monitoring station, and some portion of the suburban
population does not regularly travel to the downtown area, O-P^ is
an overestimate of the true average pollution exposure of this ur-
ban population. Assume that the true measure is 0-P^. Reference
to the true pollution-mortality relationship shows that this dirty
city would have mortality rate of 0-M.,. The observed pollution-
mortality data would be plotted as point A in Figure II-2.
11-27
-------
Mortality
Rate
True
iiiiniiiiiiiiiiiiiiiiiiiiiiiiiiitii
Observed
Pollution Measure
for Population
Figure I1-2.
True, and Observed Relationships Between
Pollution Measure and Mortality Rate
In a relatively "clean" city, there will also be a divergence
between recorded and actual pollution exposures. However, the di-
vergence will be much smaller, since the difference between the
"clean" suburbs and "almost clean" downtown area will be much small-
er. If 0-P2 is the recorded pollution measure for the clean city,
the clean city observation would be point B in Figure II-2. It can
be seen that a regression line fitted to the recorded data would
lead to an underestimate of the regression coefficient relating pol-
lution to mortality.
However, while the estimate of the regression coefficient is
biased by the error in the pollution measure, if the regression equa-
tion is used to predict changes in mortality conditional upon changes
in pollution, the regression equation will give "true" predictions.
Consider a pollution control program which is expected to reduce re-
corded pollution in the dirty cities from 0-P, to 0-P?. The regres-
sion equation predicts a reduction in mortality of 0-M3 to 0-M,.
11-28
-------
The pollution control program will actually reduce true exposure
from 0-P- to 0-P,. And the true reduction in mortality will be
just that predicted by the regression equation.
Similar problems might arise for mortality or morbidity
studies based upon a sample of individuals, if the pollution
variable is measured by some proxy. For example, suppose each
individual's exposure was taken to be a function of pollution le-
vels at his residence. Then, the recorded pollution measure would
understate true exposure for those individuals living in the "clean"
part of the city and spending some part of the working day in the
"dirty" part of the city. Similarly, for those living in the dirt-
iest part of the city, the recorded pollution measure is likely to
overstate'true exposure. By a line of reasoning similar to that
developed through Figure II-2, it can be shown that this divergence
between recorded and true exposure measures will lead to an under-
estimate of the slope of the regression line relating exposure to
morbidity or mortality. However, unlike the' preceding case, the
biased regression equation will not provide correct predictions.
Rather, the regression equation will underpredict reductions in
mortality or morbidity associated with reductions in pollution ex-
posure.
2. Pattern of_Exposure
One of the problems in doing research on dose-response rela-
tionships or epidemiological studies of the air pollution-mortality
relationship is a choice of a time dimension for the air pollution'
variable. The air pollution level at any particular point in an
urban area is an instantaneous variable which fluctuates over time.
The true exposure of an individual located at that point is mea-
sured by a trace of the time path of that instantaneous variable
over the relevant time period. The published data on air pollu-
tion levels which are used to generate exposure variables for em-
pirical research involve various approaches to summarizing this in-
stantaneous time path. Inevitably these summaries, such as the
annual mean, lose information.
11-29
-------
Also, at least in part because of measurement methods, aver-
ages are struck over shorter time periods, for example the 24-hour
average for particulates, and 8-hour and 1-hour averages for other
pollutants. These shorter averages can also provide a basis for
summary measures of exposure. For example, Lave and Seskin used
the lowest 24-hour average recorded during a year as one measure
of long term exposure. No one or two of these measures completely
represent the true exposure of any individual. Empirical research
is hampered by this inability to accurately characterize exposure
over long periods of time.
An alternative approach to summarizing recorded air pollution
data may lead to more accurate characterizations of exposure. The
suggested approach is based upon the assumption that the instanta-
neous air pollution reading can be treated as a stochastic variable.
For a city with given emissions patterns, meteorological conditions,
and location of monitoring stations, the instantaneous air pollution
reading can be viewed as being drawn from a probability distribution.
That probability distribution has a mean, variance, and other higher
order moments. If the probability distribution is stable, its para-
meters can be estimated on the basis of readings taken over a long
period of time. The air pollution of a city could be reported in
the published data in terms of these parameters. These parameters
in turn could be used to characterize exposure in empirical research.
3. Multicol linearity
When two pollutants are both thought to be related to an "ef-
fect" variable such as mortality rate, and when multiple regression
techniques are used to estimate the hypothesized relationship, the
process of estimation may be complicated by the problems of multi-
col linearity. Thus, differences in mortality rates and property
value differentials have been associated with differences in the
levels of both suspended particulates and sulfates.
11-30
-------
For example, suppose the hypothesized relationship is as
follows:
M = aQ + a.P + a2S + a3Z + e (1)
where M = mortality rate
P = sulfate concentration
Z = an index of other variables
e = disturbance term
Suppose further that P and S are positively correlated, producing
multicol linearity.
It is important to distinguish between problems in estimating
the coefficients for equation (1), and problems in using equation
(1) for prediction of mortality rates. It can be shown that when
equation (1) is estimated with multicollinearity present, the esti-
mates of a-, and a^ will be unbiased. However, they are likely to
be imprecise in the sense that the estimates will have relatively
large standard errors. In fact, it is possible that the equation
2
may have an R well over 0.95, with none of the regression coeffi-
cient estimates being significantly different from zero. It is also
possible that estimates may have the wrong sign. But it is impor-
tant to note that, while the individual coefficient estimates may
be imprecise, the equation as a whole may still provide good esti-
mates of mortality rates when the independent variables are known.
Now suppose that equation (1) has been estimated and that the
estimate of a? was insignificant by the ordinary statistical criter-
ion. One might be tempted to drop the sulfate variable and esti-
mate a new equation (2):
M = aQ + a'.P + 837 + 6 (2)
However, this would be incorrect. The estimate of a^ will be bi-
ased. Given a positive correlation between P and the omitted sul-
fate variable, it can be shown that a, > a-j This is because the
11-31
-------
estimate reflects not only the partial relationship between P and
M, but also part of the relationship between the omitted variable
and M. Similarly, if P were dropped and the equation were estima-
ted with only the S, the estimated coefficient for the S variable
would be biased upward.
In summary, when the regression equation is to be used to pre-
dict changes in the dependent variables, the correct procedure is
to use all variables and all coefficients, including those which
are insignificant, to predict changes in the dependent variable.
In the case of air pollution health benefits, multicol linearity
does not pose insuperable barriers to benefit estimation. How-
ever, in the case of the property value approach to air pollution
control benefits, the problems are more severe.
This is because benefits are not measured by changes in prop-
erty values stemming from changes in pollution levels (Freeman,
1974). Rather, the partial relationship between the pollution vari-
able and property values is used to determine a marginal willingness
to pay function. Where there is multicollinearity between a pair
of independent variables (e.g., two pollution variables, or pollu-
tion and distance from central business district), this partial re-
lationship can be estimated only imperfectly or imprecisely. In
other words, the standard error of the pollution coefficient will
be high. This uncertainty should be reflected in the final benefit
figure calculated from the regression equation.
Most of the published property value and mortality studies use
only one or, at most, two air pollution variables, for example some
measure of particulates and sulfates. But if there are other air
pollutants which also affect the property value or mortality (for
example, oxidants or nitrogen oxides) and these other pollutants
are correlated with the pollution variables included in the estima-
ting equation, then the estimates will be biased upward, as was
pointed out above. I am not aware of any studies that have at-
tempted to determine the significance of this problem.
II 32
-------
Published studies of Lave and Seskin have focused primarily
on long-run or chronic effects of exposures to relative low levels
of pollutants. They have found that long-term measures, such as
annual average and minimum bi-weekly exposures work best- in their
regression equations. This raises the question of whether benefit
estimates based on the Lave-Seskin equations might be missing the
contribution to mortality made by'health effects associated with
short-term exposure to high levels of pollutants. A consideration
of the multicol linearity problem suggests that mortality equations
should be estimated with, say, two sulfate variables, one to reflect
the presence of short-term high level exposures and other to reflect
long-term chronic exposures. These two variables are likely to be
correlated and when their coefficients are estimated in the same
equation, the standard errors may be high. However, to omit one of
the variables on those grounds will lead to biased estimates of the
coefficient for the other variable. Since the objective of obtain-
ing estimates of air pollution benefits is best served by obtaining
regressions with unbiased estimates and high predictive power, in-
significant variables should not be omitted where multicollinearity
is present.
11-33
-------
F. PREPARING NATIONAL ESTIMATES
The conceptually correct approach to aggregation of benefits at
the national level is to examine each region separately, within that
region to analyze each pollutant and each effect separately, then to
sum benefit measures across all effects, all pollutants, and all re-
gions to get the national figure. This section reviews the criteria
for selection of regional studies and the process of aggregation from
these building blocks.
1. Criteria
National estimates of benefits can be judged in terms of
two general considerations. The first looks at the underlying
benefit study used as a building block. The second general cri-
terion is the logic underlying the extension or aggregation from
the building block benefit study to the national level. The
following criteria are intended to provide a benchmark against
which to evaluate specific benefit studies which are to be used
as building blocks.
Since the objective of the exercise is determination of
values, a technique should lead to monetary measures of values or
benefits. This means, for example, that a measurement technique
which predicts the increase in recreation user-days associated
with a given reduction in pollution does not meet this criterion.
For, although the information is useful, it does not fully meet
the needs of water resource planners for benefit measures which
are commensurable with their monetary estimates of pollution con-
trol costs. Similarly, while studies of the relationship between
mortality and air pollution are essential building blocks in the
estimation of benefits, they do not determine value. In some
cases, the analyst may not have credible estimates of individual
values. In these circumstances, he may be justified in making
some explicit assumption about unit values, e.g., value per death
avoided, and determining the implications of alternative value
judgments.
11-34
-------
The technique and estimating procedures should be based an-
alytically and empirically on individual behavior and preference.
Some measures of the value of reduced mortality have been based
upon the earnings of affected individuals (Lave and Seskin, 1970).
Such measures do not meet this criterion, since there is no known
relationship between willingness to pay, on the one hand,and earn-
ings, on the other. The most obvious limitation of the lost earn-
ings measure is that it places no value on the lives or health of
those who are not working for reasons of age, sex, or other factors.
The measures of use and value should be related to changes
in pollution levels. For example, the analysis of the demand for
water-based recreation is not adequate, unless the analysis shows
how that demand is affected by changes in pollution and the quality
of the water bodies being analyzed. Particularly in the realm of
water quality, the difficulty of establishing the relationship be-
tween quality'and use is a major barrier to better estimates of
benefits. We do not yet have a clear idea of what indexes of water
quality are relevant to recreational uses of the water, and how peo-
ple will respond to changes in quality parameters. And furthermore,
the links between many of the more subtle water quality parameters
and discharge rates are not understood.
Benefit estimates should be based upon a correctly specified
theoretical model of individual behavior and the relationships
among economic units. The logic of the model must be internally
consistent and also be consistent with established theory. When
observed relationships are measured empirically, without benefit
of an underlying theoretical model, researchers may be led to make
faulty or erroneous interpretations of the data. An important ex-
ample is the discovery by early researchers that land value and air
pollution levels were inversely related in urban areas, ceteris
paribus. They then assumed that changes in welfare associated with
reduced pollution would be adequately measured by the associated
increases in land values. However, subsequent research based upon
11-35
-------
theoretical models of urban land markets, has shown that this as-
sumption is not true in general. (Freeman, 1974; Lind, 1973; Na-
tional Academy of Sciences, 1974, Chapter 4).
The actual measures used in the empirical work should cor-
respond to the variables of the theoretical model. For example,
i priori reasoning about the relationship between air pollution
ixposure and mortality suggests that the time profile of cumula-
ive exposure of an individual is the appropriate independent or
ausal variable. But most empirical research on the air pollution-
lortality relationship has been based upon air pollution measure-
lents taken at a point in time or over a short period, and at a
jpecific geographic location within an urban area. Thus, the em-
jirical measure of air pollution is not perfectly correlated with
the measure called for by the correctly specified theoretical
model. The correct theoretical measure may not be available, but
the analyst must be sensitive to the problems that might be caused
by the lack of correspondence between the theoretically correct
variable and the one actually used.
Benefit studies should use the empirical techniques appro-
priate to the theoretical model and the data at hand. Empirical
techniques should be able to cope with such problems as multicol-
linearity, simultaneity, and identification problems. Empirical
techniques which do not adequately control for other variables
should be used with great caution. Researchers should be sensi-
tive to the problems caused by misspecification of the variables
used and of the functional forms estimated.
No benefit study yet done is completely satisfactory in terms
of all six of these criteria. Yet, the researcher and analyst
need to push on and do the best they can with available data and
analytical techniques using these criteria as benchmarks. Part of
the art of benefit analysis involves a sensitivity to the gap be-
tween the ideal and the available and knowing how much confidence
11-36
-------
to place in the estimates being generated. At some point, the
problems become so great and likely margins of error so large
that the effort is not producing usable numbers.
2. Aggregation from Building Blocks
A correct approach to estimating national benefits from
valid building blocks involves several distinct steps. These
are now described in more detail.
The first step is to specify the present levels of pollution
in each region or unit of analysis, and to postulate the change
in pollution or environmental quality which is to be the subject
of analysis. The postulated changes in pollution will be deter-
mined by the nature of the policy question to which the national
benefit estimate is to be addressed. For example, in the case of
air pollution, it might be most useful to postulate the changes
in air pollution required to achieve national primary standards
throughout the country. In the case of water pollution, one could
postulate changes necessary to achieve the water quality standards
established under the 1965 law. But an analysis which postulates
a 50 percent across the board reduction in air pollution does not
shed much light on important policy issues.
The next step is to determine the functional relationship be-
tween use of the environmental amenity and environmental quality
Use is likely to be a function not only of quality, but also of
socioeconomic and other variables which will differ across regions
For example, differences among regions in industrial composition,
or importance of agriculture, may mean that the same changes in
environmental quality will have different physical effects in dif-
ferent regions.
The third step is to use the functional relationships de-
scribed above to predict the changes in use for each region con-
ditional upon its postulated change in environmental quality and
its other socioeconomic variables.
II 37
-------
The next step is to estimate the functional relationship be-
tween uses of the environment and marginal value or willingness to
pay. This relationship will also be a function of socioeconomic
variables and other factors (such as the availability of substi-
tutes as in the case of recreation).
The fifth step is to calculate the value or monetary benefit
measure for each region on the basis of predicted change in use
and other factors influencing demand or marginal willingness to
pay-
The final step consists of summing the predicted benefits
across all regions to obtain the national estimate.
These six steps are offered as a benchmark against which to
evaluate the procedures used to generate existing national bene-
fit estimates. I know of no study which has followed faithfully
all six steps as outlined here. Most of the studies take a short-
cut by generalizing from the results of one or two specific studies
to the nation as a whole. Sometimes this is reasonable, where it
can be reasonably assumed that the variables involved are constant
across regions. In other cases, the generalization may involve
significant distortions and inaccuracies. Because of the great
variety of cases and types of benefits, it is not possible to
state general rules as to when such assumptions are valid.
3. Specific Examples
I will now turn to some specific suggestions for generating
national estimates for pollution control benefits in the three
quantitatively most significant areas: health benefits from air
pollution control, aesthetic benefits from air pollution control
(property values), and recreation benefits from water pollution
control. I believe that it is possible to develop useful estimates
of pollution control benefits in these three areas on the basis of
present theory and analytical techniques.
11-38
-------
In the areas of health and property value benefits from air
pollution control, data bases exist which would permit useful a!T
though not definitive benefit estimates. Existing published esti-
mates in these two areas (Waddell, 1974) have used faulty pro-
cedures to move from the underlying functional relationships to
estimates of benefit and aggregation to the national level. In
the case of water-based recreation, the necessary data base does
not presently exist. However, work in progress by the National
Planning Association under contract to the National Commission on
Water Quality may help to close this gap.
A major component of total air pollution control benefits is
accounted for by the property value approach to aesthetics. Most
of the property value studies have used data from the decennial
>
censuses. The census collects data from each homeowner respondent.
on the owner's estimate of market value. It would be highly desir-
able to improve the data base for property value studies by obtain-
ing value estimates more directly related to market transactions,
and by obtaining estimates more frequently than .during census years
Crocker (1970) has shown that this is possible and more investiga-
tors should follow his lead.
Consider the benefits accruing to households due to improve-
ments in air quality at their residence site. The first step is
to analyze the property value-air pollution relationship for a
given city using the appropriate regression techniques. The first
derivative of the property value function with respect to air qua-
lity can be interpreted as a locus of household equilibrium margi-
nal willingnesses to pay. It is necessary to make some assumption
about the shape of each household's marginal willingness to pay
function through that known equilibrium point. Several alternative
assumptions were described by Freeman (1974). Next, for each unit
of analysis (household, block, census tract), a change in air qua-
lity is postulated and, for each unit of analysis, the area under
the assumed marginal wil1ingness-to-pay curve between the original
air quality level and the postulated new clean air point is deter-
11-39
-------
mined by integration. Finally, these are summed across all units
to determine the estimate of air pollution control benefits for
that city. (Freeman, 1974; National Academy of Sciences, 1974,
Chapter 4).
To obtain conceptually correct national estimates, it is ne-
cessary to repeat the estimation of the property value-air pollu-
tion relationship for each city, then repeat the calculations de-
scribed above. This is because different cities have different
levels of average air pollution, different spatial patterns of
air pollution, and different vectors of socioeconomic characteris-
tics. (National Academy of Sciences, 1974, Tables 4-7 and 4-8).
I have also described a technique based upon the estimation of
hedonic price indices for identifying the demand curve for clean
air directly. It requires a pooling of air pollution, property
value, and socioeconomic data from several cities. (Rosen, 1974;
Nelson, 1975). If the demand curve has been properly drawn, it
is then possible to integrate areas under these demand curves for
all individuals in all urban areas to obtain a conceptually valid
estimate of national benefits.
A shortcut approach would be to review the existing air pol-
lution property value relationships and to construct a synthetic
(average) relationship which is assumed to apply to all cities.
But this would have a higher margin of error than the method de-
scribed above. Alternatively, one could carry out this procedure
for subsets of cities grouped according to some set of characteris-
tics. Then the postulated changes in air quality could be combined
with the appropriate assumption about functional form to compute
benefits for each unit of analysis within each city. Finally, these
could be summed within each city and across all cities to obtain the
national benefit figure.
The present state of the art in health effects research can
be characterized as follows. Lave and Seskin have produced a sub-
stantial body of research showing an association between mortality
and air pollution and a number of less comprehensive studies have
11-40
-------
corroborated this finding. The Lave-Seskin research is based up-
on cross sectional studies using U.S. SMSA's as the sample unit.
At the other end of the spectrum, a substantial number of micro
dose-response relationships have been established for specific
clinical manifestations. These are primarily related to high
acute dosages, at levels not often experienced in urban areas.
The substantial gap between these two bodies of studies must
be closed. At the clinical level, more work should be done to
establish the specific mechanisms and cause-effect relationships
involved with lower-level, longer-term exposures, such as those
found associated with mortality by Lave and Seskin. On the other
hand, epidemiological work should begin to focus on more disag-
gregated research based upon smaller units of analysis. But this
work must await the establishment of a better data base on both
mortality and morbidity associated with exposure to pollutants.
I would recommend the following procedure for estimating
health benefits from air pollution control. First, existing mor-
tality-air pollution studies are reviewed. Either the single best
relationship is selected, or a synthetic relationship is construc-
ted which reflects the range of experience and confidence in dif-
ferent estimates revealed in existing studies. For each unit of
analysis defined for the mortality relationships selected (e.g.,
SMSA, city), a change in air pollution levels is postulated. Then
for each study the change in mortality based_upon the change in air
pollution and other socioeconomic variables is computed. It should
be noted that, if the selected relationship is linear, the predicted
change in mortality is independent of other variables. However,
if the function is semi-log or log, the vector of other variables
must be incorporated in the calculation. Summing th£ predicting
changes in mortality over all SMSA's will give an aggregate measure
of deaths avoided for the nation as a whole due to the postulated
air pollution control program. To obtain a monetary measure of bene-
fit, it is presently necessary to make some explicit assumption about
the value of deaths avoided or postponed. Morbidity effects can be
measured and valued in the same manner.
11-41
-------
Unfortunately, the most widely quoted estimate of national
health benefits associated with air pollution control did not use
a correct method for predicting national changes in mortality.
After choosing their "best" regression associating total mortality
with measures of suspended particulates, sulfate particles, and
several socioeconomic variables, Lave and Seskin (1970) calcula-
ted elasticity coefficients of the dependent variable with respect
to each of the independent variables at the mean values for all
variables. Strictly speaking, the elasticity enables one to cal-
culate the percentage change in the dependent variable for a given
percentage change in any independent variable.
If the elasticities are constant throughout (i.e., if a log-
log regression form had been estimated), then the elasticity co-
efficients can be used for any magnitude percentage change in an
independent variable for any sample point. However, Lave and
Seskin estimated a linear form and it is well known that linear
functions have different elasticities throughout the range of the
function. Therefore, the calculated elasticity coefficient can
strictly speaking only be used to estimate changes in mortality
for the "average" city for small changes in its population level.
A sample calculation suggests that with relatively low elastici-
ties, as in the case cited, the magnitude of the error in using
the same elasticity figure throughout for all of the sample points
can be substantial. Therefore, the estimated changes in mortality
consequent on a reduction in air pollution must be called into ques
tion.
Recreation appears to be the largest category of benefits to
be realized from water pollution control. There are two major
problem areas in the estimation of national recreation benefits.
The first is our ignorance of the relationship between water qua-
lity and recreation use or participation. Second, it is very dif-
ficult to generalize to the nation as a whole from the small num-
ber of studies that have been done on the unit value or shadow
price to be attached to a unit of recreation (e.g., a recreation
day). A priori reasoning suggests that the appropriate shadow
11-42
-------
price depends upon local socioeconomic characteristics, the avail-
ability of substitutes, and taste formation mechanisms, such as
the "learning by doing" phenomena. Also, many of the recreation
sites which have been studied have unique attributes which make it
difficult to separate the value of the general recreation experi-
snce from the value attached to those site-specific characteristics
In estimating the national recreation benefits from water pol-
lution control, there are several points to bear in mind. First,
;he major barrier to developing adequate estimates of water pollu-
tion control benefits is our lack of understanding of the relation-
ship between various measures of water quality, on the one hand,
and various recreational uses of water bodies, on the other. The
nost commonly measured water quality parameters are at best only
crude proxies for the more complex vectors of characteristics which
influence recreational uses. Second, it is very dangerous to ex-
trapolate predictions of changes in recreation participation or re-
creation demand from narrowly focused studies of limited geographic
areas.
Third, the analysis of recreation benefits must recognize the
variety of likely individual responses to changes in the availabi-
lity of recreation sites. Existing recreationists will increase
participation levels; non-participants will become participants;
and there will be shifts in the geographic pattern of recreation
activity, as recreationists shift from one site to another. Fur-
thermore, there will be both changes in the level and pattern of
economic/market transactions associated with recreation, e.g.,
travel expenditures to recreation sites, and changes in utility
levels which have no counterpart in market transactions.
It would be useful to outline briefly the theory of recrea-
tion benefits as developed for a single recreation site holding
all other relevant economic magnitudes constant. There exists a
demand for this recreation site relating quantity demanded of re-
creation services (e.g., in recreation days) to price. This can
11-43
-------
also be interpreted as a marginal willingness to pay curve, rela-
ting marginal value to quantity. This demand curve is plotted
holding income, prices and availability of substitutes and alter-
natives, and the quality of this recreation site constant. When
price is known, as indicated by the line P in Figure II-3, actual
recreation use for quantity demanded can be predicted. Of course,
price could be zero. When the facility is polluted and the demand
curve with pollution is as shown, the quantity of recreation days
would be 0-Q-,.
A
Recreation Days
Figure II-3. Recreational Benefits of Water Pollution Control
The net value or rent attributable to this recreation site
is the area A-B-C. This is a consumer surplus measure of value.
Now, let us assume that a pollution control program increases wa-
ter quality at this site. In economic terms, the effect is to
shift the demand curve out and to the right - the new demand curve
as shown in the figure. The net economic benefit of this cleanup
is the increase in consumer surplus, i.e., the area between the
two demand curves, B-C-E-D.
11-44
-------
The net benefit can be divided into two categories. The
first is the increase in utility or consumer surplus to those 0-
Q-l users who were using the facility even when polluted. This is
the area B-C-F-D. This area represents their increase in willing-
ness to pay to maintain present use rates at this recreation site
rather than do without. In addition, the greater attractiveness
of this site relative to alternative recreation sites and alterna-
tive consumption activities, other than recreation, results in an
increase in recreation days at this site equal to Q-i-Q?- There is
a benefit associated with this increase in use which is equal to
the area C-E-F.
In utilizing such a measure of benefit, there is no need to
take into account changes in recreation use at other sites, sav-
ings in travel cost, or whatever. These are all reflected within
the B-C-E-D, benefit measure. When there are simultaneous changes
in quality or availability at several recreation sites, the theore-
tical and empirical analysis becomes a bit more complicated. (Burt
and Brewer, 1971).
As the analysis of Figure II-3 shows, changes in water quality
will be associated with both changes in willingness to pay and
changes in quantity of participation. This suggests that the simul-
taneous pursuit of two research paths might be most appropriate.
The two are a national recreation participation survey incorporating
socioeconomic variables as well as quality and quantity availability
variables, and micro efforts to measures shifts in the demand curve
for specific sites as water quality changes.
11-45
-------
G. CONCLUSIONS
Four questions were raised at the outset of this paper. The first
was: "Would national benefit estimates be useful?" In Section B, se-
veral important environmental policy issues were described which would
certainly be illuminated by better information on benefits. Even when
environmental policy issues are not made in accordance with a strict
benefit cost rule, benefit and cost magnitudes are s'ill important con-
siderations in environmental decision making.
The second question was: "Will the state of the art permit the
development of national estimates of benefits?" In my judgment, the
economic theory and analytical techniques for valuing benefits for most
classes of benefits are relatively well developed. There are still ma-
jor unresolved theoretical and analytical issues with respect to valu-
ing health and life benefits. But in other areas, for example, land
value, travel cost approach to recreation, net factor income and pro-
ductivity approaches, etc., the validity of the theoretical framework
is now well established.
However, major problems remain at the interface between the eco-
nomic dimensions of the benefits problem and those dimensions dealt
with by the other sciences. There are major gaps in our ability to re-
late changes in use of the environment to changes in the environmental
quality parameters. Examples include dose-response relationships for
air and water pollutants, and relationships-between water quality and
recreation and fishing benefits.
It is fair to say that we know much more now about the nature and
magnitude of air and water pollution control benefits than we did, say,
five years ago. A review of Ridker's (1967) valiarvt effort is instruc-
tive in this regard, both in terms of the very crude conceptual frame-
work and analytical technique and the almost non-existent data base he
had to work with.
Turning specifically to air pollution, it is now possible to use
existing techniques and data to obtain tentative estimates of air pol-
11-46
-------
lution health and property value benefits. Obtaining these estimates
will involve the explicit use of value judgments and assumptions about
certain magnitudes; and there will be some degree of uncertainty in
the final figures. But part of the present state of the art includes
a consistent framework and language for expressing this uncertainty in
probalistic terms. Furthermore, these estimates will be reasonable in
the sense that we are more likely to hit the correct figure by using
these techniques than we are by drawing numbers out of a hat.
In the case of water pollution, medical knowledge of potential
health benefits is simply inadequate. And for the major class of water
pollution control benefits, recreation, the data base for relating water
quality changes to changes in recreation use is not adequate. I would
conclude that the state of the art and data do not permit reasonable
water pollution control benefit estimates.
Are there any "hopeless cases?" For purposes of this discussion,
we can define "hopeless" as a situation where the marginal cost of ad-
ditional useful information is infinite. By this definition, there are
no hopeless cases in the area of benefit measurement. Consider the prob-
lem of valuing human life. One might argue that although individuals
do make decisions which imply money values for changes in life expectancy,
these decisions reveal a multiplicity of values for different individuals
and different circumstances. However, the search for a single unique
value of life applicable to all situations is unnecessary. The absence
of a unique single price or marginal value is a characteristic of all pub-
lic goods, such as the demand for clean air and the unit value of recrea-
tion experiences. The point is that our understanding of the valuation
problem can be improved by further research. It is in this sense that
the problem is not hopeless.
There are, however, some insoluble problems in the area of benefit
estimation. They are insoluble because they involve value judgments
and fundamental questions of equity, not primarily questions of economic
theory and measurement. We have uncovered three manifestations of this
fundamental problem in the course of our discussion. First, there is no
11-47
-------
objective criterion for judging or evaluating different distributions
of benefits and costs. Thus, although distributions of benefits and
costs can be described for decision makers, this information cannot
be incorporated into a comprehensive, objective decision 'rule with-
out the prior specification of some set of equity weights. Second,
our estimates of benefits are derived from the price information re-
vealed by existing markets. This price information derives its valid-
ity from the implicit acceptance of the existing distribution of wealth
which produced the existing set of prices. If the present distribution
of wealth is judged to be unacceptable on ethical grounds, no meaning-
ful welfare statements can be derived from the existing price informa-
tion. And third is the evaluation of potentially massive or catastro-
phic environmental damages which may occur at some more distant time
in the future, i.e., so-called intergenerational effects.
The third question was: "Do recently published estimates of na-
tional benefits adequately reflect the state of the art?" In other
words, do these estimates make the best use of presently available
analytical techniques and data? Although a detailed answer is beyond
the scope of this paper, in summary, the answer is "no". However,
the cost of generating estimates which do reflect the state of the art
would be relatively small; and in my judgment, the benefits of doing
so justify incurring the cost.
The final question was: "How can present national benefit esti-
mates be improved?" The recommendations for gathering additional
data and more effective utilization of existing data are presented
in the next section.
11-48
-------
H. , RECOMMENDATIONS
Recommendations for improvements in benefit estimation here fall
into three categories: those dealing with improvements in the basic
data and analytical techniques; those dealing with improved uses of
existing data; and one proposing the initiation of ex post measure-
ments of realized benefits.
1. Basic Data and Analytical Techniques
Earlier we distinguished between measuring effects and as-
signing of values, both necessary steps in the determination of
benefit estimates. But in the key areas of mortality and morbid-
ity from air pollution, and in the lesser areas of vegetation and
materials damages from air pollution, as well as sports and com-
mercial fisheries and production benefits from water pollution
control, the measurement of these effects lies largely outside
the realm of economics. And economists' efforts to assign values
and estimate benefits in these areas must await better informa-
tion on physical effects.
The apparent importance of health effects of air pollutants
suggests a high payoff for more information on dose-response re-
lationships. More effort should be devoted to large-sample care-
fully controlled epidemiological studies to examine both long-
term chronic and short-term acute effects of air pollutants. An
effort should be made to move beyond the inter-urban comparisons
of the sort done by Lave and Seskin. Perhaps the population at
risk data base developed for EPA can be used to carry out epidemi-
ological studies disaggregated within urban areas. Also, some at-
tention should be given to epidemiological studies of water pollu-
tion health effects. In particular, attention should be given to
the problems of persistent organic chemicals, and viral pathogens
in drinking water.
Research efforts to determine health and other kinds of ef-
fects should be designed so as to complement and support efforts
11-49
-------
to determine values. For example, sample surveys to determine
morbidity effects due to air pollution should include questions
to obtain socioeconomic data. This is important not only because
socioeconomic variables are themselves associated with health sta-
tus and must be controlled for in estimating health effects, but
also because if these data are part of the effects study, they
may facilitate the later assignment of economic values.
Research on the value of human life should be extended in
two directions. First, the individual choice/willingness to pay
model where one element of choice is a small change in the prob-
ability of death should be further developed. (Bergstrom, 1975;
Zeckhauser, 1975; Thaler and Rosen, 1974). One:important exten-
sion is to examine the implications of changes in the shape of
the probability distribution of expected life. Second, govern-
ment decisions and other collective decisions involving safety.
health, medical research, etc., should be carefully analyzed to
determine the implied or explicit values placed on human life or
changes in probability of death. This effort may give some in-
sight into collective attitudes toward life valuations, and it may
also stimulate efforts to achieve greater consistency among public
sector decision makers in this area.
The U.S. EPA should commission a series of micro region-spe-
cific efforts to estimate the demand curves for specific recrea-
tion sites using the Clawson-Knetsch travel cost approach and to
relate shifts in the demand curve to changes in various water qua-
lity parameters. This series of studies should be designed and
coordinated so as to control cross studies for socioeconomic vari-
ables, the availability of substitutes and alternative sites, and
other regional differences in recreation behavior. The objective
of these studies would be to identify the underlying functional
relationship between quality and willingness to pay for recrea-
tion experiences. If this effort is successful, it should be pos-
sible to generalize the results to additional sites and regions.
This series of studies would emphasize the determination of the
demand curve and evaluation.
11-50
-------
Another study should address changes in the quantity of use
on the basis of a large, carefully constructed survey of recrea-
tion participation. Earlier surveys have not adequately dealt
with the quantity and quality of the available water based recrea-
tion sites and the availability and quality of alternative forms
of recreation opportunity as determinants of water based recrea-
tion participation rates. If such a data base were established,
and participation rates could be successfully related to the quan-
tity and quality of recreation opportunity, then the prediction of
changes in participation rates associated with specific water pol-
lution control programs would be relatively straightforward. Of
course, this approach would not reveal values or shadow prices; nor
would it capture the change in utility of existing recreationists.
The model being developed by the National Planning Associa-
tion under contract to the National Commission on Water Quality
appears to be a very promising step in the right direction. How-
ever, it does not appear that that model can be adequately devel-
oped given existing time and resource constraints. That research
effort should be broadened and continued at a generous level of
support. Independently, a series of travel cost and similar studies
should be initiated in an effort to determine unit values for recre-
ation days, and, more importantly, the way in which they vary with
changes in water quality and other attributes of particular recrea-
tion sites.
There should be carefully constructed experiments with the in-
terview/questionnaire techniques for estimating willingness-to-pay
for reduction in pollution. These experiments should be coordina-
ted with studies based on other analytical techniques in an effort
to provide a cross check or validation of benefit estimates obtained
by different approaches.
2. Use of Existing Data
The Lave and Seskin air pollution-mortality regression equa-
tions should be used to predict changes in mortality by SMSA for
11-51
-------
changes in air quality necessary to meet national primary air qua-
lity standards. These estimates of changes in mortality should
be combined with alternative assumptions about the value of human
life to calculate a range of benefit estimates.
The Freeman-Rosen technique should be used to compute correct
benefit estimates from existing property value studies for indi-
vidual cities. Also, it may prove useful to commission a series
of new property value studies using improved measures of property
value and a consistent theoretical model and empirical technique
for all cities. These studies should be designed to identify the
demand curve or willingness-to-pay curve for clean air.
A group of experts should be convened, perhaps under the au-
spices of the National Academy of Sciences, for the purpose of de-
veloping a comprehensive estimate of national benefits of pollution
control. The group should be comprised of professional researchers
with backgrounds in environmental economics, the economics of pro-
ject evaluation, econometrics, epidemiology, environmental health,
limnology and marine biology, agricultural sciences, and engineer-
ing. The group should be asked to review all of the available lit-
erature and on-going research on the effects of all classes of pol-
lutants, and to compile their best estimate of national benefits of
pollution control for each of the major polluting substances.
The group should utilize a consistent framework for defining
and measuring benefits, and consistent techniques and methodologies
for moving from building block studies to national estimates. It
should utilize all available data, making use of their expert judg-
ments as to relative quality and accuracy of parameters and rela-
tionships from different studies. The group should be asked to
make recommendations to guide further research to close gaps in
information that are identified by the group. A new group should
be convened every three to four years in order to review new data
and to prepare new national benefit estimates.
11-52
-------
3. Ex Post Evaluation
A careful and comprehensive program of ex^ post analysis of
pollution control benefits should be developed and implemented.
The period of the seventies will have been characterized by a
major commitment of resources to air and water pollution control
It would be penny-wise and pound-foolish not to plan now to al-
locate one or two percent of the total funds to be spent on pol-
lution control toward measuring and evaluating what we will have
bought with that massive expenditure. In one sense, the nation
is embarking on a large-scale socioeconomic experiment in alter-
ing environmental conditions. We should take advantage of this
experiment with a carefully thought out and comprehensive pro-
gram of data gathering and analysis.
11-53
-------
I. REFERENCES
Bergstrom, Theodore, "Preference and Choice in Matters of Life and
Death," in press, 1975.
Burt, Oscar, and Durward Brewer, "Estimation of Net Social Benefits
from Outdoor Recreation," Econometrica 39(5):813-827, September
1971.
Crocker, Thomas, Urban Air Pollution Damage Functions, University of
California, Riverside, PB 197-668, 1970.
Freeman, A. Myrick, III, "A Survey of the Techniques for Measuring
the Benefits of Water Quality Improvement," in Henry Peskin and
Eugene Seskin, eds., Cost/Benefit Analysjs and Hater Pollution
Policy, The Urban Institute, Washington, D.C., 1975.
__, "On Estimating Air Pollution Control Bene-
fits from Land Value Studies," Journal of Environmental Econono-
mics and Management 1:74-83, 1974.
Haveman, Robert H., and Burton Weisbrod, "The Concept of Benefits in
Benefit/Cost Analysis with Emphasis on Water Pollution Control
Activities," in Henry Peskin and Eugene Seskin, eds., Cost/Bene-
fit Analysis and Water Pollution Policy, The Urban Institute,
Washington, D.C., 1975.
Kneese, Allen, and Orris Herfindahl, Economic Thoery of Natural Re-
sources , Charles Merrill, Columbus, 1974.
Lave, Lester and Eugene Seskin, "Air Pollution and Human Health,"
Science 169:723-733, 21 August 1970.
Lind, Robert C., "Spatial Equilibrium: The Theory of Rent and the
Measurement of Benefits from Public Programs," Quarterly Journal
of Economics 87(2):188-207, May 1973.
Mishan, E. J., "Evaluation of Life and Limb: A Theoretical Approach,"
Journal of Political Economy 79(4):687-705, July/August 1971.
National Academy of Sciences, Air Quality and Automobile Emission Con-
trol , A Report by the Coordinating Committee on Air Quality Stu-
dies, Vol. 4, The Costs and Benefits of Automobile Emissioii Con-
trol , Washington, D.C., 1974.
, Principles for Regulatory Decision Making
About Chemicals in the Environment, 1975.
Nelson, Jon, "Residential Choice, Hedonic Prices, and the Demand for
Urban Air Quality," Pennsylvania State University (in press), 1975
Ridker, Ronald G., Economic Costs of Air Pollution, Praegar, New York,
1967. ~'
11-54
-------
Rosen, Sherwin, "Hedonic Prices and Implicit Markets: Product Differ-
entiation in Pure Competition," Journal of Political Economy, 1974
Schelling, Thomas C., "The Life You Save May be Your Own," in Samuel
Chase, ed., Problems in Public Expenditure Analysis, Brookings In-
stitution, Washington, 1968.
Thaler, Richard, and Sherwin Rosen, "The Value of Saving a Life: Evi-
dence from the Labor Market," University of Rochester Discussion
Paper, 1974.
Waddel1, Thomas E., The Economic Damages of Air Pollution, U.S. Envi-
ronmental Protection Agency, EPA-600/5-74-012, 1974.
Zeckhauser, Richard, Processes for Valuing Lives, Harvard Public Poli-
cy Program Discussion Paper #29D., 19/5.
11-55
-------
III. BENEFITS OF AIR POLLUTION CONTROL
Thomas D. Crocker
University of Wyoming
A. INTRODUCTION
This section presents the purpose and scope of this paper, the
techniques for estimating air pollution control benefits, and finally,
the organization of this paper.
1. Purpose and Scope
The purpose of this paper is to review and assess the methods
available for modeling and measuring the economic benefits of air
pollution control. Models of the economic behavior of beneficiar-
ies of air pollution control could be increasingly used to untangle
the various financial, economic, and environmental issues relevant
to proposed control efforts. Nevertheless, in some instances, at-
tempts to apply these models may already have gone beyond the gen-
eral understanding of their analytical foundations and their appro-
priate roles in decision making. In order to evaluate the extent
to which this might be true, it is necessary to have a clear idea
of the possible purposes of a model of the behavior of beneficiar-
ies of air pollution control.
Fundamentally, a model can have any one or more of three ob-
jectives:
• Enhanced understanding of the phenomena being inves-
tigated
0 Improved ability to predict the future courses of im-
portant variables at various levels of disaggregation
• Improved basis for analyzing the economic consequences
of specific alternative control policies.
Generally speaking, all three objectives are served better by any
model encompassing a greater range of phenomena or one characterized
III-l
-------
by variables more closely corresponding to observable phenomena,
or one permitting the use of more powerful and readily applied es-
timation techniques.
Among these three objectives, the first is most likely to be
of interest to the professional researcher,, while the latter two
assume greatest relative importance for the decision maker. More-
over, the first requires more emphasis upon fortnal analytical com-
prehension, whereas the latter two emphasize empirical implement-
ability. In those parts of economics relevant to the assessment
of the benefits of air pollution control, there has frequently been
inadequate attention by analytical investigators to possibilities
of improved empirical implementation. On the other hand, decision
makers have overly encouraged the production of empirical efforts
lacking proper analytical foundations. Recognizing these deficien-
cies, there now appears to be a judgment that research into the
economic benefits of air pollution control can ultimately have lit-
tle to contribute to the improvement of decisions about air pollu-
tion control. This paper will attempt to temper this judgment.
The framework of analysis adopted throughout this paper is the
market, the interaction of buyers and sellers. The particular inter-
est is those markets in which air pollution is thought to have a sub-
stantive impact on buyer and seller valuations. The manner in which
the valuations expressed in these markets can be used to infer the
value of air pollution control is discussed at some length. Substan-
tial attention is given to the modeling complications introduced by
special features of markets such as health and real property in which
air pollution is thought to play a significant role. Among these fea-
tures are multi-dimensional heterogeneity (in which the combination of
characteristics embodied in each unit of a good is more or less unique
to that unit), costly information, costs of market participation, and
institutional and time constraints.
The neglect of these factors results in the omission of entire
classes of air pollution control benefits stemming from the ameliora-
tion of intertemporal externalities and uncertainty about future
III-2
-------
levels of air pollution and its effects. This may mean that sub-
stantial biases are built into current estimates of air pollution
control benefits, even if these current estimates have been estab-
lished in an analytically and empirically sound fashion.
2. Estimation of Benefits
Estimating techniques employ one of three alternative mea-
sures of benefits:
• Alternative cost
• Opportunity cost
• Willingness-to-pay.
The first of these has been employed most frequently because of its
great simplicity. However, its failure to consider adjustments
that the sufferer can make in response to the presence of air pollu-
tion means that it will nearly always yield estimates of benefits
that are biased upward. Moreover, because users generally do not
employ an explicit analytical model, the approach yields little, if
any, information about behavioral phenomena, the future course of
important variables, or the consequences of specific alternative
control policies.
In contrast, opportunity cost and willingness-to-pay measures
require the construction of explicit analytical models of the pheno-
mena being studied. Consequently, investigators with a background
in formal economic analysis who are not under severe time constraints
tend to employ these two measures. The research problem then becomes
primarily one of comprehending economic, rather than physical or bio-
logical relations. The investigator is able to describe the behav-
ior of the price or value of air quality with respect to various phy-
sical, biological, and economic parameters. Analytical models of
health markets, property markets, etc., employing opportunity cost
or willingness-to-pay measures are amenable to empirical treatment
III-3
-------
with either available data or data that can plausibly be obtained.
One, must, however, be moderately less sanguine about the avail-
ability of data to implement models in which uncertainty or inter-
temporal externalities have a substantive role.
Whatever the analytical and empirical quality of the benefit
estimates thus far obtained, they rarely, if ever, make clear the
uncertainty of the estimates themselves. All benefits assessment
efforts should generate probability distributions for whatever in-
put variables are relevant and then aggregate these distributions
to produce a probability distribution for the output measure, a'ir
pollution control benefits. Doing so would reduce the amount of
useful information that is thrown away and would reduce the suscep-
tibility of benefits estimates to criticism.
In addition, a number of short-term analytical and empirical
approaches could yield rapid improvements in the state-of-knowledge.
These include:
Development and empirical implementation of general
equilibrium models of property values
Adoption and empirical implementation of a household
production function framework for health effects stud-
ies
Input productivity studies with special emphasis upon
human factors of production
Adoption of economic, rather than physical or biologi-
cal, perspectives for the construction of environmental
quality indices
Adoption of new techniques from mathematical statistics
to test the functional forms of expressions employed to
estimate benefits
Testing of procedures intended to cause interviewees
to reveal their true preferences
Increased use of algorithms intended to enhance the
investigator's ability to discriminate among variables
111-4
-------
that might contribute to variations in the values
assumed by another variable
• Use of sophisticated econometric models of agricul
tural markets.
3. Organization
The next section presents the conceptual foundations of bene-
fits assessment for any air pollutant and discusses the issues that
must be resolved in applying these foundations to the analysis of in-
dividual air pollution control decisions. The third section describes
each of the primary methods that might be used to estimate the bene-
fits of air pollution control. An attempt is made throughout to re-
late these methods to a conceptual framework or a model and the prac-
tical questions, including data availability, associated with using
each method are raised.
The fourth section compiles the most recent empirical efforts
and reviews the content of these efforts in terms of the earlier com-
mentary on conceptual foundations and empirical procedures. The last
section reviews briefly the role and prospects of benefits assessment
in policy decisions and argues for an increased allocation of research
effort to the solution of specific analytical problems.
In order to provide the reader with some appreciation of the mo-
tivations underlying the occasionally more abstract discussions in the
preceding sections, the attachment contains a discussion of the econom-
ic benefits for the participants in a single activity of controlling a
single class of pollutants in a particular locale.
III-5
-------
B. CONCEPTUAL FOUNDATIONS
This section presents the conceptual foundations for assessing
the benefits of controlling any air pollutant and takes up the issues
that must be resolved in applying these foundations to the analysis
of specific control decisions. The presentation covers the motiva-
tions for measuring benefits, measurement techniques, and sources of
benefi ts.
1. Motivations for Measuring Benefits
Air pollution control benefit assessments are intended to aid
the public decision maker in formulating policy. If neither more
nor less meaning is to be attached to these assessments than their
underlying analytical foundations justify, then the decision maker
must be sensitive to the limitations as well as the strengths of
these foundations.
Cost-benefit analysis is an attempt to ascertain the socially
valuable resources and services to be gained and lost in the attain-
ment of any particular state of the world. In each state, market
prices alone are presumed to embody all relevant information about
relative scarcities and to be a sufficient means of allocating re-
sources to their socially most highly valued uses. Price is assumed
to be the sole mode of allocation used in voluntary interactions of
buyers and sellers. Markets are pervasive; that is, all gains from
exchange are quickly exhausted. Price Is thus a sufficient measure
of social as well as private value. In effect, the objective of
cost-benefit analysis is to ascertain what the price structure of
any particular state would be when markets are pervasive.
Sophisticated proofs are available in the economics literature
showing that states of tha world attained in a system in which mar-
kets are pervasive are Pareto-efficient: the net social value of
output is maximized because no one individual's welfare can be en-
hanced without reducing the welfare of another individual (Arrow
III-6
-------
and Hahn, 1971). The price structure to be ascertained via cost-
benefit analysis is therefore a Pareto-efficient price structure.
It follows then that, insofar as benefits assessments are used by
the decision maker as information inputs in policy decisions, he
is making the judgment that the benefit valuations established
through strictly voluntary exchange in a setting where markets
are pervasive are at least one useful means of weighting the alter-
natives he has before him.
In the practical world outside economics textbooks, all gains
from exchange are not exhausted because markets are, in fact, not
pervasive. Institutionally fostered monopolistic advantages, tech-
nological externalities, such as air pollution, and consumption and
production indivisibilities persevere even though their very exis-
tence implies the presence of potential gains from exchange. In a
general equilibrium setting in which the activities of all pairs
of buyers and sellers are interconnected, the presence of an inef-
ficient price structure in one market implies a similar presence
in all other markets. If efficiency is defined only for circum-
stances in which markets are pervasive, and if the activities of
all pairs of buyers and sellers are interconnected, then all real
observable price structures must be inefficient.
The activity of benefits assessment, when employed for pub-
lic policy purposes, is a direct recognition that markets for cer-
tain goods and services are either nonexistent or incomplete.
These goods and services are therefore valued inaccurately in the
sense that their market prices diverge from efficient levels as
defined by a world of pervasive markets. Benefits assessment em-
ploys various techniques to infer the equivalent prices of these
non-marketed goods and services from information embodied in the
observed market prices of marketed goods and services that are in
the same production or consumption process as the non-marketed en-
tity or which are substitutes or complements to it.
III-7
-------
The extent of inefficiency reflected in these observed mar-
ket prices is typically unknown. Unless an empirically implement-
able general equilibrium model of the economy is available, the
extent of inefficiency (remembering that efficiency is typically
defined only for a world of pervasive markets) is inherently un-
knowable. However, one does not do serious violence to accurate
benefits assessment if one imposes separability conditions that
permit dimissal from consideration of all markets whose goods and
services are plausibly only remotely connected to the good or ser-
vice for which benefits assessments are being performed.
Of course, a contribution to the description of the efficiency
benefits of air pollution control need not be the sole or even one
of the reasons why a decision maker wants benefits assessments per-
formed. After all, any benefits assessment is simply a description
of the price structure effects of a particular change in the state
of the world. It is therefore readily used also to describe the
effects of a change in air pollution control upon the distribution
of income, government tax revenues, or any of a number of other pe-
cuniary and economic questions of interest. Nevertheless, since
all of these effects are influenced by the price structure of the
market for the services of the atmospheric resource, as well as
the price structures of closely related markets, analytically sound
descriptions of these effects also require the construction and
empirical implementation of a general equilibrium model.
2. Measures of Benefits
Having recognized the type of world to which economic effi-
ciency is generally meant to refer and the implications of this
reference for the construction and empirical implementation of
models intended to measure air pollution control benefits, it re-
mains to make explicit exactly what is meant by the term "bene-
fits". First, it is necessary to be absolutely clear that the
parties to whom the benefits of air pollution control are presumed
III-8
-------
to accrue are individual persons. Social benefits are simply
some aggregate of individual benefits: there does not exist in
cost-benefit analysis a concept of the society having an exis-
tence independent of the individuals who compose it.
Second, there exists a consistent, discoverable relationship
between a change in ambient air pollution levels and the change
in the individual's utility or welfare. The latter is interpreted
as the benefits accuring to the individual. Central to the logic
of this relation is the idea of consumer's surplus. This states
that the individual continues to purchase additional units of the
ith good until the marginal utility of the additional unit is just
equal to the marginal utility of the money he must pay to obtain
the additional unit. That is, the individual will be indifferent
between keeping his money and consuming an additional unit of the
good when the marginal utility of the additional unit is equal to
its market price. The difference between what the individual is
willing to pay, as determined by the marginal utility he attaches
to the unit, and what he has to pay, as determined by market price,
is the consumer's surplus for the unit. The sum of the surpluses
for all these units is total consumer's surplus.
This concept is illustrated in Figure IV-1, where DD is
the demand function, the locus of points depicting the consumer's
Price
Z°
Z1 Air Quality
Figure IV-1. Representation of Consumer's Surplus
III-9
-------
marginal willingness to pay. The cross-hatched area, ABC, repre-
sents the consumer's surplus associated with an improvement in air
quality, Z, from Z° to Z .
In a very important paper, Willig (1974) has shown recently that,
in most applications, the difference between the observable consumer's
surplus area, ABC of Figure IV-1, and the theoretically correct con-
sumer's surplus of the income-compensated demand function is accurate-
ly described as trivial. Willig's (1974) results vindicate analyti-
cally the conceptual foundations and great practical value of the con-
sumer's surplus methodology so frequently employed in studies of the
benefits of air pollution control. It should nevertheless be noted
that there exist several different measures of consumer's surplus,
most of which are distinguished by the path the consumer is presumed
to follow from one equilibrium position to another. For presentation
of a set of criteria with which to evaluate these various measures,
see Mohring (1971).
The empirical establishment of the observable surplus becomes
considerably more complicated if alternative formulations of the con-
sumer's problem are employed. For example, rather than directly en-
tering the utility function, air pollution might reasonably be viewed
as influencing the cost to the consumer of using goods capable of pro-
viding utility. Thus, if the absence of respiratory problems is a
good from which a consumer derives utility, he will have to expend a
greater quantity of valuable resources attaining this absnece in New
York City than in Laramie, Wyoming.
An alternative formulation would have the consumer obtain uti-
lity directly from the absence of air pollution while also permit-
ting air pollution to affect the costs of other activities or goods
from which he obtains utility. Estimation of consumer's surplus
using this formulation is more complex than using the traditional
formulation, because the goods or activities the use of which air
pollution is presumed to affect (e.g., medical services) are treat-
ed as inputs into commodities (e.g., the absence of respiratory
111-10
-------
problems) that are the ultimate objects of the consumer's pre-
ferences. In general, even though these formulations will usually
increase estimation problems, failure to adopt them in circum-
stances where reality warrants their adoption would introduce spe-
cification errors and thus possibly result in serious errors in
the benefit estimates obtained.
As these progressively more complex formulations of the con-
sumer's surplus problem make intuitively clear the greater the
generality and realism of the formulation, the greater the diffi-
culty likely to be involved in empirically implementing the pro-
positions falling from the formulation. Nevertheless, because
both informational and modeling constraints prohibit the consid-
eration of the full range of commodities, activities, and condi-
tions influencing consumer behavior, some subset of these things
must be taken as given. An explicit statement of the formulation
on which the empirical effort is based is thus highly desirable.
Only then is the reader able to ascertain whether the estimation
techniques and the data base employed fit the problem. Equally
important, the reader is unable to assign a unique meaning to the
empirical results, unless he is made fully aware of the conceptual
framework from which these results come.
3. Sources of Pollution Control Benefits
Most discussions of air pollution control benefits explain
the existence of economically excessiye ambient pollution concen-
trations in terms of static nonpecuniary externalities. That is,
one economic agent's production or utility relations are said to
include levels of air pollution chosen by other economic agents,
without regard to the effects on the first agent's welfare. The
damaged agent in these discussions is implicitly presumed to know
with certainty either the magnitude of the effects caused by the
air pollution and/or the level of air pollution. Moreover, the
reader is customarily permitted to assume that the activities of
III-ll
-------
the pollution perpetrator and the responses of the sufferer to this
pollution are contemporaneous.
The restriction of the discussion of the analytical reasons
for the existence of air pollution control benefits to static non-
pecuniary externalities might explain why two plausibly important
causes of air pollution control benefits* intertemporal externali-
ties and uncertainty, have been rarely considered. Their neglect
may mean that substantial myopic biases are built into current esti-
mates of benefits.
There appear to be a number of air pollution effects charac-
terized by either or both of the above causes. For example, rain
containing acid sulfates may ultimately alter the forest ecosystems
of entire regions. Certain long-lived synthetic chemicals, once
widely dispersed throughout the economy, and thereby the natural
environment, may pose carcinogenic and mutagenic hazards for fu-
ture generations. An economic agent's activities at one point in
time impose constraints upon other economic agents in the future.
Insofar as there exists imperfect knowledge of future events, these
externalities are a problem worthy of empirical attention.
The air pollution effects associated with intertemporal exter-
nalities and uncertainty are characterized by some or all of the
following attributes:
• Information asymmetry
• Effects uncertainty
• Public-good type risks
• Low-probability catastrophic losses
• Irreversible effects.
Substantial discrepancies exist in available information
about the benefits of increasing an activity, as opposed to the
benefits of reducing it. The benefits of automotive travel are
III 12
-------
relatively easy to calculate; the benefits of reducing auto-
mobile travel and its associated air pollution, noise, and aes-
thetically loathsome landscapes are generally more difficult
and expensive to ascertain. For many air pollution phenomena,
there is a nearly complete absence of knowledge about the response
of physical and biological systems. For example, the overall ef-
fect on the earth's climate of increased particulate loadings of
the troposphere is highly uncertain. Similarly, the influence
on human health of increasing ambient loadings of acid sulfates
has not been adequately defined.
The actual or potential risks caused by certain activi-
ties are indivisible as between and among economic agents.
Moreover, they are frequently borne by very large groups of
these agents, including future generations. These groups are
unable to reject the risks because they are bound up in a common
property resource, e.g., the atmosphere. As used here, risk may
refer not only to a lack of complete certainty about one's future
wealth, but also to not knowing if one will be alive to enjoy this
wealth.
With current information, there exists a probabilistically
small possibility of catastrophic or large-scale losses. For ex-
ample, one potential result of burning large quantities of high
sulfur fuels in England is the possibility of greatly and detrimen-
tally altering the forest ecosystems of the Scandinavian countries
through formation of acid rain. There is also a possibility that
the effects of a particular activity are physically impossible to
reverse or cannot be reversed at reasonable cost. Mutagenic changes
probably have this characteristic.
The importance of issues embodying substantial uncertainty
or ignorance, and therefore the extent to which they are worthy
of substantial research effort to assess their benefits and costs,
depends in large part upon the decision strategy adopted. If a
once-and-for-all commitment is to be made to undertake or fail to
111-13
-------
undertake an activity, then the likely effects deserve to be spe-
cified and measured rather closely. But if a sequential decision
procedure is to be followed, the accumulation of knowledge can be
delayed. One can then live with a fairly high degree of uncertainty
in the present. The cost of an error in choice is not likely to
be as great, because it will be relatively less costly to reverse
course.
Setting intertemporal externalities and lack of basic knowl-
edge about air pollution effects aside, there is an additional way
in which uncertainty might plausibly enter calculations of air pol-
lution control benefits. In the more mundane localized materials
damage, property value, interview, and health effects studies of
common urban air pollutants that form the bulk of air pollution
control benefits studies, the uncertainty of potentially affected
individuals about future air pollution levels and the possible in-
fluence of this uncertainty upon air pollution control benefits is
usually not even mentioned (Crocker, 1971; National Academy of
Sciences, 1974).
The disutility the affected individual attaches to future air
pollution dosages has been treated as identical in situations where
perfect foresight is lacking and in situations where it is present.
An affected agent's behavior at a point in time has therefore been
viewed as invariant with respect to the probability of error in his
forecast of expected air pollution dosages and/or effects. Yet an
extremely voluminous analytical and empirical literature is avail
able in economics indicating that in most everyday circumstances,
individuals are risk-averse, i.e., uncertainty is costly.
When one admits that one component of air pollution damages
may be due to uncertainty of the affected individual about current
and future air pollution levels and effects, it is apparent that
individual damages from air pollution dosages are by no means in-
dependent of the process by which expectations about future dosages
and effects are formed. That is, realized damages are but present
111-14
-------
representations of positions taken in the past in response to
expected air pollution dosages and effects. Most empirical air
pollution control benefit studies have assumed that sufferer ad-
justments to changes in air pollution dosages and effects are in-
stantaneous. A more appealing hypothesis is that expectations
about future air pollution dosages and effects adapt to changes
in present dosages and effects only after some lag in time. Thus,
if a change has any permanent effect at all, the effect is not
registered all at once but is instead distributed over several
time periods.
111-15
-------
C. EMPIRICAL PROCEDURES
This section reviews some general formulation employed in estima-
ting the benefits of air pollution control. The topics include methods
of estimating benefits, availability of data, expression of uncertainty
in results, and aggregation procedures.
1. Methods of Estimating Benefits
The general formulations customarily employed in the esti-
mation of air pollution control benefits are:
• Alternative cost
9 Opportunity cost
« Willingness to pay.
The alternative cost formulation is employed most frequently
because of its simplicity and avoidance of complex economic analy-
sis. The procedure begins with determination of the physical ef-
fects of an activity. For example, an agronomist might construct a
dose-response function relating ambient concentrations of acid sul-
fates to the yield of a particular agricultural crop. These effects
are then valued in monetary terms. The assigned money value is that
minimum cost which would have to be borne in order to return the af-
fected entity to its original state. This implies that all substi-
tution possibilities and other adjustment processes available to the
owner of the affected entity are dismissed from consideration. Losses
therefore tend to be biased upward in the case of reductions in air
quality, and benefits tend to be biased downward in the case of im-
provements in air quality. If, and only if, no ready adjustments
are possible for the owner of the affected entity is the alternative
cost formulation a reasonable approximation of the actual losses or
benefits.
Wherever these substitution possibilities and adjustment
processes are accounted for, the researcher has adopted an
III 16
-------
"opportunity cost" formulation. In the case of a decline in air
quality, a net loss measure is thus obtained. For example, the
researcher studying the influence of smog on the behavior of out-
door recreationists would explicitly take into account the stock
and the unit price of substitute recreational opportunities. -By
taking these opportunities into account, the researcher is, in
effect, deriving the minimum amount the recreationists would ac-
cept to be willing to be subjected to a change in air quality.
The presumption is that title to the air resource is held by the
parties who would be affected by the change rather than by the
potential perpetrator of the change.
Finally, the "willingness-to-pay" formulation underlies the
discussion of the entire previous subsection. As noted earlier,
this formulation involves the determination of what individuals
would be willing to pay in order not to be subjected to a change
in air quality. Therefore, the researcher is assuming that title
is held by the perpetrators, rather than by those who will be af-
fected.
The ordering of the magnitudes of these formulations can dif-
fer from one set of circumstances to another. Frequently, estimates
of losses associated with reductions in air quality using the op-
portunity cost formulation exceed the estimates resulting from the
willingness-to-pay formulation. This is due to the difference in
fundamental entitlements and therefore the wealth positions of the
parties. Estimates obtained using the alternative cost formulation
as opposed to the opportunity cost formulation follow no consistent
pattern with respect to relative magnitudes. For example, one could
reasonably argue that the money necessary to compensate outdoor re-
creationists for increases in the amount of smog-damaged forest might
be relatively small when compared with the alternative of returning
the forest to its original state.
The alternative cost method is nearly always used in the "tech-
nical coefficients" approach (Anderson and Crocker, 1971). Also,
111-17
-------
the widely quoted Lave and Seskin (1971) study on the economic as-
pects of the health effects of air pollution is best described as
an alternative cost study. Examples of studies that do not involve
air pollution, but show the empirical practicality of willingness-
to-pay studies applied to health effects are provided by Acton (1975),
Davis and Russell (1972), Feldstein (1971), Grossman (1972), Phelps
and Newhouse (1974), and Rossett and Huang (1973).
2. Availability of Data
The extent to which each of the three alternative measures have
been employed in investigations of air pollution control benefits
has been determined by three factors: the availability of data, the
availability of time, and the background of the investigator in for-
mal economic analysis. Use of the alternative cost measure typically
requires little background in economic analysis, while allowing the
use of accumulated skills in the natural sciences. The necessary eco-
nomic "analysis" involves little more than the weighting of physical
and biological relationships by readily observed market prices.
The investigator has the opportunity to draw upon the more so-
phisticated developments in economic analysis when he chooses to em-
ploy the opportunity cost or the willingness-to-pay measures. Ra-
ther than simply using observed market prices as determined by un-
known and presumed irrelevant forces, the opportunity cost and the
willingness-to-pay measures compel the construction of models of
the implicit market for air quality, thus permitting the investiga-
tor to describe the behavior of the price or value of air quality
with respect to various physical, biological, and economic para-
meters .
The extent to which observations must be available on the
actual values assumed by the elements of each of these three sets
of parameters will, of course, vary according to the problem being
111-18
-------
investigated. It is true, nonetheless, that the available secon-
dary data-and readily acquired primary data for the economic para-
meters of a great many problems have by no means been fully ex-
ploited. For example, great masses of data are available in pub-
lications of the U.S. Department of Agriculture and other sources
allowing the empirical implementation of opportunity cost models
of air pollution damage in agricultural markets. Similarly, large
secondary data collections from Federal, state and local sources
make possible the estimation of hypotheses derived from willingness
to-pay models of property markets.
Even opportunity cost and willingness-to-pay models of the
effect of a/ir pollution upon health are amenable to empirical
treatment with either available data or data that can be readily
obtained. The widely known CHESS studies of U.S. EPA would have
provided enormously more valuable information on the health im-
pacts of air pollution if they had initially been designed by
economists as well as by biomedical people.
In summary, analytically sophisticated and empirically im-
plementable models for which data are available or can be readily
generated do exist. This is not to say that all the economic bene-
fits of air pollution control are readily susceptible to empirical
estimation. The point is that a great deal more can be done. The
decision maker must judge whether the results are likely to be
more useful than estimates obtained via the simplistic alternative
cost formulation, or even the intuitive or experiential approaches
that are used frequently in political decisions.
The above comments about the availability of data allowing the
empirical implementation of opportunity cost and willingness-to-
pay measures apply mainly to static externalities. One must be
less sanguine with respect to data availability for empirical im-
plementation of economic analysis of intertemporal externalities
and risk aversion. The latter is that facet of human preferences
111-19
-------
responsible for the existence of benefits from the reduction of un-
certainty about air pollution effects and levels.
Two approaches have dominated economists' and psychologists'
attempts to measure risk aversion. In the first, lotteries are
established in a controlled laboratory setting and the investiga-
tor, usually a psychologist, attempts to extrapolate the results
to real-world situations. Certainly, one can generate a great
deal of data in this manner. However, the degree of correspon-
dence between the controlled laboratory and real-world situations
is not always apparent. (Lichtenstein, 1965; Pruitt, 1962).
The second approach has been fairly widely used by econo-
mists. Here, the observed behavior of individuals in their every-
day pursuits is used to infer the value they attach to the reduc-
tion of risk to fortune, health, and life. (Fama and MacBeth, 1973;
Oi, 1975; Brown and Enke, 1972). The approach is appealing to the
professional economist, because it permits the construction of mo-
dels of the behavior of economic agents, the formal derivation of
testable propositions, and the application of opportunity cost or
willingness-to-pay measures of benefits. However, its usefulness
is limited by data availability, at least with respect to health
and life.
A number of difficulties may be attributed to the public good
type risks that characterize air pollution. These are borne by
large groups of economic agents, including future generations.
That is, the population exposed to deleterious ambient levels of
air pollution typically involves a wide cross-section of ages, in-
comes, medical histories, employments, and other factors.
The most available data, permitting ready inferences about
economic agents' voluntary willingnesses-to-pay to avoid risk to
health and life are found in the area of industrial safety. How-
ever, since netiher the medical backgrounds nor other personal
111-20
-------
attributes of industrial workers are representative of the popula-
tion of a typical urban area, the extrapolation of these workers'
health and life risk valuations to entire human populations can be
done only with some trepidation. On the one hand, industrial work-
ers may, as a group, simply have less or more innate aversion to
risk than do other identifiable groups within the entire population.
In addition, some groups, such as university professors and govern-
ment or private bureaucrats, may have less accumulated experience
with situations posing risks to health and life. Given more ex-
perience, their attitudes may be altered.
There exists at least one source of readily available data
on the risk aversion of representative individual economic agents
that has not been fully exploited. This involves the voluntary
consumption or use of goods which carries obvious and generally
acknowledged risk to health and/or life. For example, certain
automobiles are widely known to be safer than other automobiles.
In effect, the safety features of the automobile are weighed by
the purchaser against its other features and its price relative to
other automobiles and perhaps other modes of transportation. The
question, nevertheless, remains as to how one extrapolates the
agent's valuation of risk to health and life in the context of an
automobile to his valuation of risk to health and life in the con-
text of air pollution.
Intertemporal externalities are, of course, intimately tied
to the presence of uncertainty and risk^aversion. These externa-
lities are complicated, however, by the integral part played in
their determination by irreversible effects and the passage of
time. Analytical efforts in this area have thus far concentra-
ted upon the nature of the adjustments that must be made in tra-
ditional benefit-cost evaluations of alterations of natural en-
vironments in order to account for irreversibilities, uncertain
future effects, etc. However, these efforts typically fail to
supply exact quantitative adjustments for these evaluations, nor
do they provide computational algorithms indicating the form,
111-21
-------
type, and quantity of data required for empirical implementation.
Without such knowledge, one cannot state whether the requisite
data is, in fact, already available or easily obtainable.
Nevertheless, it should be noted that, in particular cases
not involving air pollution, some investigators have managed to
implement empirically at least a part of the relevant analysis.
(Krutilla and Fisher, 1975). These partially successful empiri-
cal efforts have occurred in geographical areas where detailed
data have been available on likely future technological develop-
ments, environmental carrying capacities, and demand patterns for
uses based on the existence of a natural environment, as well as
for uses involving the transformation of a natural environment.
3. Expression of Uncertainty in Results
Even if the world is a purely deterministic place, it is
indeed unlikely that it would ever be worthwhile to accumulate
enough data about air pollution problems to eliminate all possi-
ble sources of error in benefit estimates. Errors in measure-
ment and simple absences of observations on key variables would
persist. In spite of this rather obvious point, estimates of air
pollution control benefits are, for the most part, compiled and
reported in terms of a single outcome that fails to take account
of valuable information about the extent of uncertainty in the
reported estimates. The commonly accepted procedure in benefits
assessment calls for the calculation of the benefits from each al-
ternative action and for the use of political and economic criteria
to choose among these on the basis of expected returns. But ex-
cept by chance, single-valued benefit estimates must always be
in error.
The uncertainty problem discussed in this subsection occurs
because many of the variables affecting an air pollution control
plan are neither subject to the control of the decision maker
nor fully captured in the empirical implementation of any model
III-22
-------
constructed by the benefit assessor. These variables include prop-
erty rights structures, technological change, prices, micro-climates,
etc. Hence, in contrast to commonly accepted procedures, benefits
assessment that takes full account of uncertainty requires: judg-
ments about the likely pattern of behavior of these nohcontrollable
variables; calculation of an entire set of outcomes or returns for
each ambient air concentration; and criteria for choosing among con-
centrations on the basis of sets of possible benefits for each level
of control.
On those occasions when uncertainty has been explicitly taken
into account, range sensitivity tests have typically been used to
account for lack of knowledge about the behavior of different rela-
tionships or the values assumed by particular parameters. Waddell
(1974), for example, includes upper and lower bounds and "best
guesses" for various air pollution damage categories. An alterna-
tive procedure is to generate probability distributions for what-
ever input variables are relevant and then aggregate these distri-
butions to produce a probability distribution for the output mea-
sure, i.e., air pollution control benefits. In an interesting ap-
plication of this latter approach, Mercer and Morgan (1975) have
successfully applied the Weibull (1939) family of distributions,
showing that the amount of valuable information made available to
the decision maker is enhanced significantly (see also Pouliquen,
1970).
The probabilistic approach to presentation of benefit esti-
mates has many advantages, not the least of which is that it per-
mits the investigator to incorporate accumulated wisdom and intui-
tion into the analysis in an explicit and communicable fashion.
If research resources are scarce, individuals who have had long
experience in working in an area frequently have a "feel" for the
structure of the problem, which permits them to assign subjective
probabilities to various outcomes. Some people argue that such
evidence should be dismissed on the grounds that it is not "objec-
tive", but this argument is badly mistaken. It is difficult to
111-23
-------
understand why the objective specification of uncertainty should
be so important, when the criteria used for choosing among alter-
native models and oftentimes the model choices themselves are sub-
jective.
Another major advantage of the suggested probability approach
is that it does not throw away useful information. For example,
in attempting to assess air pollution damages to commercial crops,
the biochemist might specify the "best" of several dose-response
functions relating crop yield to ambient concentrations of a par-
ticular pollutant. In the absence of a thoroughly coordinated
research effort in which the economist specifies the variables,
units of measure, and sampling procedures, it is possible that
the biochemist's conception of best does not accord with that of
the economist. It is then up to the economist, who usually is ut-
terly illiterate in biochemistry, to translate the biochemist's
results into something useful for purposes of economic analysis.
By requiring that a probability be assigned to the various plausi-
ble outcomes, the impact of this decision problem can be greatly
improved. (Crocker, 1975).
4. Aggregation Procedures
Any estimate of national air pollution control benefits is
an index number. That is, a variety of relationships of multiple
dimensionality are collapsed into a unique sealer measure. The
process of the collapse or aggregation involves an exchange of
detailed information about the form and structure of a set of
problems in return for tractability and ease of comprehension.
The best single source of information on the aggregation problem
in economics continues to be Green (1964). The portion of Marschak
and Radner (1972) dealing with the one-person team is highly in-
structive from a decision theory perspective. Fisher and Shell
(1972) provide an excellent treatment of the place in index num-
ber problems of taste and quality changes, both of which are
highly relevant to environmental quality problems.
111-24
-------
For a limited number of purposes, it may be of interest to
obtain estimates of the total nationwide benefits accruing from
air pollution control. An aggregate benefit estimate may provide
policy makers with an easily grasped idea of the magnitude of
the problem and may therefore give some indication that additional
or reduced research resources should be devoted to ascertaining
the economic effects across space and time of alternative control
strategies. As a possible component of the national accounts,
these estimates may also be useful for macroeconomic decision mak-
ing purposes since they provide some evidence on how the presence
of air pollution and attempts to control it alter the nation's ag-
gregate net productivity. An effort, currently underway at the
National Bureau of Economic Research to find ways to include en-
vironmental quality costs and benefits in the national accounts,
is partially described by Peskin (1974).
However, attempts to apply aggregated economic response func-
tions or aggregated magnitudes to local or regional air pollution
decision problems introduce sources of error that can lead to sub-
stantial discrepancies between expected and realized economic re-
sults. These sources of error may remain even if the aggregated
functions or magnitudes are used as no more than the points from
which disaggregation to the local level is initiated. The initial
process of aggregation over locales, and subsequent disaggregation
back to locales, may destroy knowledge of the various transforma-
tion and substitution possibilities and relative consumer valua-
tions of alternatives in each locale. That is, the weights used
to determine the contribution of each type of sufferer and each
locale to the aggregate measure may not correspond to the relative
weights sufferers in each locale attach. The application of the
same set of weights across different sufferers and locales assumes
that all sufferers and locales are identical and that existing
differences among sufferers and locales will remain unchanged as
air pollution levels change. These are indeed strong assumptions.
Existing estimates of the national benefits of air pollution
control combine a large number of heterogeneous disaggregated
111-25
-------
studies dealing with vegetation and materials damages, property
values, health effects, and opinion surveys. A majority of
these studies employ alternative cost measures, although will Ing-
ness-to-pay measures have been widely used in the property value
studies. The process of aggregation typically employed to obtain
these estimates is exemplified by the procedures Waddell (1974)
employed to obtain an estimate of aggregate national (residential)
property value damages.
Waddell first reviewed a collection of studies that had es-
timated marginal purchase price functions with respect to sulfur
oxides and/or suspended particulates for eight different cities.
Interpreting the values of the air quality parameters in these
several studies as measures for the average household in each
study of equilibrium marginal willingness-to-pay at given air qua-
lity states and with given demand functions for air quality, he
selected a value within the range of these estimated values. By
selecting this value within the range of values, he assumed that
what was interpreted as the equilibrium marginal willingness-to-
pay was the same for all households in all cities. Then, using
the further assumption that this assumed equilibrium marginal
willingness-to-pay was in fact the actual marginal willingness-
to-pay for all air quality states, he multiplied the constant
marginal willingness to pay by the number of households and the
number of air quality states to obtain an estimate of aggregate
national air pollution damages.
In effect, Waddell assumed that the decision problem
of each and every household in each urban area of the country
could be represented by Figure IV-2, where b is the marginal will-
ingness-to-pay, Q is an air quality level, D is the aggregate na-
tional air pollution damage, <9P/
-------
that willingness to pay involves only the multiplication of b by
whatever change in air quality is postulated. Thus the value to
the depicted household of an improvement in air quality (Q** - Q*)
is simply b(Q** - Q*), and the sole distinction one needs to make
among households in order to calculate aggregate national damages
is to account for the location of each household on the Q axis.
Figure IV-2. Willingness to Pay for Improvement in Air Quality
The above discussion has been devoted to a single aggregation
over individual households. Typically, however, estimates of na-
tional air pollution control benefits involve aggregation over mul-
tiple classes of pollutants, over time, as well as over households.
sealer estimates of the national benefits of air pollution control
may thus involve two or three distinct types of aggregation, each
of which embodies unique assumptions about the similarities among
the units undergoing aggregation. Thus,one must choose which type
of aggregation is to be performed first in arriving at a sealer
representing air pollution control benefits for households, for
pollutants, and for time intervals. Moreover, in deciding how to
perform the first aggregation, one must take into account how the
aggregation for the second and third steps will be carried out.
111-27
-------
Finally, it should be noted in passing that the entire dis-
cussion in this subsection has presumed that estimates of the ag-
gregate benefits of air pollution control must be constructed from
disaggregated, microeconomic studies. The presumption is that em-
pirically measurable microeconomic variables and relations have
been derived from microeconomic theory. Given these derivations,
aggregated or macro-variables representing national benefits of
air pollution control and national ambient air pollution levels
have then been defined, Thus far, there has been no intensive
analytical effort devoted to constructing an aggregate or macro-
theory that will serve to make the macro-variables consistent with
the micro-theory and micro-variables. At least in principle, the
development of such a theory is by no means a hopeless task. The
development of macro-economic theory and national income account-
ing is evidence of this.
111-28
-------
D. SHORT-TERM IMPROVEMENTS IN BENEFITS ASSESSMENT
Most of the discussion to this point has been directed at im-
provements in the long-term strategy of benefits assessment research.
This section seeks to make the reader aware of some theoretical con-
structs and empirical procedures holding tangible near-term promise.
The topics are arranged in order, from the more analytical to the more
empirical.
1. General Equilibrium Models of Property Values
Rosen (1974) and Freeman (1974) have recently developed a con-
ceptually improved method of obtaining aggregate national damage
estimates that neither assume that the same set of weights applies
to each locale and each individual, nor disregards the manner in
which burdens are distributed across space, time and individuals.
This method employs property value and attributes data for indi-
vidual urban areas to ascertain a distinct marginal purchase price
function for each urban area, using similar data and statistical
techniques for each area. Having estimated the marginal purchase
price function for each urban area, the data from each area are
merged into a single data set. From this data set the relation-
ship of air quality to marginal purchase prices and various house-
hold attributes such as income and age is estimated. The value
of the parameter attached to the marginal purchase price variable
can then be interpreted as the marginal willingness-to-pay for an
improvement in air quality.
If the expression relating air quality to marginal purchase
prices and various household attributes is nonlinear in the origi-
nal variables, one can readily obtain scalar measures of marginal
willingness-to-pay that can differ from one urban area to another.
If this expression is linear in the original variables, one obtains
a scalar measure of marginal willingness-to-pay that is constant
from one area to another. Nevertheless, this two-stage procedure
111-29
-------
does permit the distinctive attributes of different urban areas
and different households to enter into the determination of aver-
age marginal willingness-to-pay.
Summation over the marginal willingness-to-pay estimates for
each city in the sample and extrapolation to the nation of the
average sample marginal willingness-to-pay, if the sample is thought
to be representative of all the urban property markets in the coun-
try, provides an estimate of aggregate national air pollution con-
trol benefits. This estimation procedure has been recently applied
to a single urban area in an extremely interesting and innovative
paper by Nelson (1975).
The procedure is soundly grounded in economic theory and is
capable of capturing many elusive features of the market, such as
multi-dimensional heterogeneity and assorted institutional con-
straints. In contrast to many other approaches that could be em-
ployed, this one can be implemented by widely and thoroughly un-
derstood statistical techniques. Finally, a multitude of data,
most of which is isomorphic with the theoretical parameters, is
readily available.
The Rosen-Freeman estimation procedure forgoes far less in-
formation than studies in the other benefit categories when ag-
gregated to the national level. It thus becomes worthwhile to
invest substantial analytical and empirical research effort in
establishing the extent to which health effects, materials da-
mages, etc., are registered in property values. Insight can
then be gained into the decision value of the estimates likely
to be obtained by investing further research resources in the
development of the other benefit categories. A serious effort
should be made to implement empirically the Rosen (1974) and
Freeman (1974) estimation procedure in a single coordinated
research project, in order to obtain national estimates of
the benefits of air pollution control.
111-30
-------
2. Household Production Function
There is a common perception that the reduction of health
effects constitutes, both absolutely and at the margin, the great-
est benefit of air pollution control. Substantive opportunity to
test this perception is now available in the form of an analytical
framework commonly known as the household production function.
Fundamentally, this approach views the consumer as combin-
ing market-purchased goods and his own resources to produce a
state of health; that is, medical services and other remedial and
precautionary measures are not desired for their own sake but are
instead used because they reduce the probability that certain states
of the world will occur or because they reduce the loss to the con-
sumer, given the occurrence of certain states of the world. The
properties of these household production functions for states of
health can readily be perceived as being determined by the state
of the household's consumption technology in precisely the same
manner that the properties of conventional production functions of
firms are determined by the state of standard production technology.
Conceptually, air pollution is thus considered as an exogenous fac-
tor that detrimentally influences the state of health likely to be
attained with a given mix and magnitude of household inputs.
The household production function approach to the theory of
the consumer is consistent with alternative cost, opportunity cost,
or willingness-to-pay measures of benefits of air pollution con-
trol. The alternative cost measure, for example, requires investi-
gation of the ways in which the consumer can, under given ambient
air pollution conditions, combine market-purchased goods and ser-
vices as well as his own resources to produce a particular state of
health. Development of opportunity cost or wi Hifigness-to-pay mea-
sures would, of course, require that the consumer's choice process
among alternative states of health be accounted for. The theore-
tical development of these latter two measures within the household
111-31
-------
production function framework is set forth in Rosen (1974). Evi-
dence is abundant that empirically sound opportunity cost and will-
ingness- to-pay measures can be obtained via the approach for such
nonmarket activities as church attendance, human fertility, educa-
tion, outdoor recreation, medical services, and a variety of other
topics.
Previous studies of the economic benefits of reducing the
health effects of air pollution neglect, among other things, the
value of the leisure time the individual loses while he is ill, and
fail to account for individuals who are nonparticipants in the work
force or who refuse medical services. Moreover, these studies do
not recognize that the consumer may be uncertain about the covaria-
tion of air pollution and health effects and about the "producti-
vity" of particular precautionary or remedial activities. Finally,
previous studies have abstracted from the possibility of the indi-
vidual learning about effects and about the productivity of various
activities as well as the possibility that he eventually becomes
acclimatized to air pollution and thus attaches less relative im-
portance to its presence.
The major advantage of the household production approach is
that it provides an analytical framework for the resolution of the
issues cited above. U.S. EPA should encourage research efforts
that attempt to establish in detail the empirical observations re-
quired to gain the necessary insights from the analytical perspec-
tive of household production functions.
3. Input Productivity Studies
Past studies of the economic effects of air pollution on pro-
duction inputs have focused entirely on specific inputs without
giving attention to the manner in which these inputs are involved
in a production process. This is due to the universal absence in
these studies of an explicit model of producer behavior. Moreover,
these studies have dealt with non-human inputs (e.g., materials
111-32
-------
damages). The economics of the effects of air pollution upon
the supply of worker effort and participation has apparently not
been studied at all.
The analytical perspective that studies of this sort might
adopt is exemplified at least in part by the perspective this au-
thor is employing in a U.S. EPA-funded study of the impact of air
pollution upon the work performance of citrus pickers in southern
California. In this study, the individual citrus picker is viewed
as a contractor who sells his labor services in response to various
combinations of piece-work wage rate offers and expected picking
and environmental conditions. The product the picker is selling
is the number of boxes of fruit he picks within a given time in-
terval. His returns are determined by his wage per box of fruit
picked and the relative ease of picking the fruit. It is thus pre-
sumed that the worker attaches utility to his working conditions
as well as his additional income and is willing to substitute one
for the other. The grove owner is aware of this. He thus pro-
duces a primary output (fruit) that he sells in the market and a
set of working conditions that he offers his pickers. The grower's
problem is then to choose a combination of working conditions and
fruit output, while the picker's problem is to choose a combina-
tion of working conditions and piece-work wage rates.
This analytical perspective coincides with the fundamental
view of voluntary coordinated production presented in Thompson
(1968), Alchian and Demsetz (1972), and Crocker (1973). It can
be empirically implemented by means of procedures similar to those
set forth in Rosen (1974). It seems desirable to invest some re-
search effort in additional attempts to assess these effects, be-
cause there is no existing information dealing with the impact of
air pollution upon labor supply and productivity, because an analy-
tical framework and empirical observations are readily available,
and because it is not implausible that the effects in question
may be substantial. It is likely that the analytical framework
and perhaps some of the empirical results will be transferable to
111-33
-------
labor productivity situations involving other environmental pol-
lutants in the work place.
4. Environmental Quality Indices
In the early 1970's, there was some hope that useful indi-
ces could be constructed capable of conveying in capsule form
some information about changes in benefits of improvements in
environmental quality. Enthusiasm for the possibilities of con-
structing these indices was evident in the 1972 Annual Report of
the Council on Environmental Quality, though subsequent reports
have been noticeably less sanguine about these prospects.
Attempts to construct such indices have viewed changes in
ambient pollutant concentrations as being nearly synonymous with
changes in air pollution control benefits. Factors indicative of
social valuations have entered only as more-or-less arbitrarily
chosen weights to be attached to the changes in ambient concen-
trations of each pollutant of concern. These weights were typi-
cally based upon the Federal primary or secondary air quality
standards, and a great many facets of the economic benefits of
air pollution control were neglected.
To improve upon earlier attempts to construct useful indices
of changes in the benefits of air pollution control, one needs
to initiate the exercise from within an analytical economic frame-
work rather than a physical taxonomy. In particular, the quality
of the atmosphere must be viewed as a factor of production that,
in combination with other factors, influences objects such as
states of health that individuals value. Thus, for example, one
aspect of the "quality" index of an urban atmosphere might be the
difference in some morbidity characteristic in the present rela-
tive to some predetermined standard atmosphere.
Even if dollar values cannot always be attached to the parti-
cular object of preference, a great deal more information about
111-34
-------
benefits would be conveyed in this manner than with a simplistic
recitation of ambient pollutant concentrations. Moreover, in
order to construct a useful index, one need not have near-complete
information on the relation connecting air pollution to the object
of preference. In many cases, the accumulated wisdom and knowl-
edge of researchers will permit the assignment of probabilities
to various possible forms of the relation. These probabilities
can then be aggregated to form the desired index, in accordance
with well-known principles of the mathematical theory of prob-
ability.
5. Form of Benefit Functions
Economic theory provides few clues to the functional forms
appropriate to the specification of economic relationships. Bene-
fit functions have been estimated using a variety of specifications
without noticeable concern as to the appropriateness of the func-
tional form chosen. However, the choice of specifications affects
one's estimates of the characteristics of the benefits function and
therefore the calculated benefits from a change in ambient air pol-
lutant levels. This has been recognized by Nelson (1975) in his
employment of the household production function approach. Recent
references on the problem of choosing functional forms include
Dhrymes et al. (1972), Pesaren (1974), Ramsey (1974), and Schmidt
(1974).
6. Interview Studies
The major reason for distrust of interview studies in which
the interviewee is asked his willingness to pay for alternative
ambient air quality states is that he has an incentive to bias
his responses depending on how he anticipates the information
will be used. Although statements of total willingness to pay
for given levels of air pollution control may be biased, the dif-
ferences in total willingnesses to pay for various levels may be
reasonably accurate representations of marginal willingnesses to
111-35
-------
pay. It is, of course, the latter that is of primary importance for
decision making purposes.
It is difficult to conceive of any way other than interview
studies of valuing such amorphous facets of air pollution problems
as visibility and other aesthetic effects (Randall et al., 1974;
Krutilla and Fisher, 1974). Moreover, certain techniques for elic-
iting truthful responses about preference orderings have recently
been developed by economists studying the demand for public goods
(Clarke, 1971; Bohm, 1972). Although most of these techniques have
not been empirically implemented, it seems worthwhile to study their
possible empirical biases in some detail, if only to ascertain whe-
ther the present state-of-the-art really justifies the continued
viewing with great skepticism of interview studies.
7. Studies of Agricultural Damages
Studies of air pollution damages to agricultural crops have fo-
cused upon the physical response, i.e., the influence of air pollu-
tion upon the physical magnitude of output and the mixes and magni-
tudes of physical inputs that are adopted. The calculated economic
damages reported in agricultural damage studies typically assume that
changes in these outputs and inputs have no influence whatsoever up-
on the unit price of the output or the unit costs of the inputs. The
determination of these price effects requires the construction of a
detailed econometric model of the agricultural market in question.
Models of agricultural markets that are capable of capturing these
price effects are generally available, and should be applied to air
pollution studies.
8. Testing for Exclusion of Variables
Many of the investigators of property value and health effects
studies have had to rummage through large data files,involving num-
bers of structural, neighborhood, household, and personal attributes
that, given the discriminatory abilities of the models employed,
111-36
-------
could plausibly be determinants of property values or health ef-
fects. Generally, no prestated strategy performed in a reproduci-
ble way has been used to choose the set of explanatory variables
for which estimated parameters are finally reported. This is true
in spite of the widely known fact that the exclusion of a relevant
variable from an expression to be estimated will usually bias the
estimated parameters of the included variables.
In order to ameliorate this problem, some writers have employed
factor analysis. However, the economic meaning of the grouped vari-
able then obtained is typically unknown. A program called AID (Au-
tomatic Interaction Detector) by Sonquist et al. (1973) simulates
the prestated, if complex, strategy followed by a good researcher
in searching for the explanatory variables that increase his power
to account for the variance of a dependent variable. Using this
program, it should be possible to substantially reduce the likeli-
hood that a relevant explanatory variable is inadvertently excluded
while simultaneously avoiding the construction of economically mean-
ingless grouped variables.
111-37
-------
E. STATUS AND PROSPECTS OF BENEFITS ASSESSMENT
This closing section reviews the current state of the art in assess-
ment of air pollution control benefits and weighs its role and prospects in
policy decisions. An increased allocation of resources is advocated for
implementation of a directed and meaningful research program.
1. Status and Role of Benefits Assessment
In my opinion, the current state-of-the-art in assessment of the
benefits of air pollution control is rather feeble. Entire classes
of benefits possibilities stemming from amelioration of uncertainty
and inter-temporal externalities are nearly completely neglected on
both analytical and empirical terms. Static, nonpecuniary externali-
ties have monopolized empirical attention. However, the results from
studies devoted to this class of benefits have rarely been stated in
a form most useful to the decision maker. That is, no probabilities
have been attached to the alternative results. Because of the fre-
quent absence of an explicit analytical framework grounded in econom-
ic theory, the economic meaning of many of these results is a matter
of guesswork in any case. Given these problems, attempts to establish
aggregate national estimates of the benefits of air pollution control
are subject to enormous error.
The prospects for attaining national damage estimates are by no
means entirely bleak, however. New analytical approaches and empiri-
cal techniques have become available that are soundly based in econom-
ic theory, are susceptible to empirical treatment, and for which data
are already available or at least are readily collected. The applica-
tion of these approaches and the empirical techniques necessary for
their implementation for estimation of the benefits of air pollution
control is feasible. Moreover, from the perspective of decision mak-
ers, the encouragement of these approaches and techniques should be
worthwhile and is likely to become increasingly so.
It is axiomatic that the set of obligations faced by a public de-
cision maker will usually differ greatly from the obligations faced
111-38
-------
by the sufferers from and perpetrators of air pollution, if only be-
cause the public decision maker is unable to appropriate directly any
pecuniary benefits his decisions may cause. Thus the benefits of one
activity relative to another will differ between the public decision
maker and the air pollution sufferers and perpetrators. Therefore,
if the public decision maker is to act as the embodiment of the suf-
ferers and perpetrators by choosing the set of activities maximizing
the combined value of the air resource to sufferers and perpetrators,
it must itself face a price structure, given its own obligations, that
will cause it to choose this set of activities. Use without comple-
mentary information of benefits assessments and associated cost assess-
ments (that is, the price structure relevant to sufferers and perpetra-
tors) may cause the public decision maker to choose an inefficient set
of activities. Nevertheless, benefits assessment can be of great as-
sistance in choosing the set of activities that, given the decision
maker's obligations, constitute the economically efficient set of ac-
tivities. The reasons are several.
First, the availability of sound benefits assessment can help set
the decision maker's thinking straight. It makes it more costly for
him to set a policy course that entirely neglects the pervasiveness
of economic scarcity. Second, it enhances the development of a frame-
work in which, as additional information is accumulated, more sophis-
ticated and knowledgeable analysts can insert their own results and
probability estimates to make more informed judgments. Third, it al-
so permits the decision maker to insert his own results and probability
estimates about air pollution control benefits. Fourth, it helps the
decision maker see which questions for possible further investigation
are of greatest impact in terms of changing policy choices. Fifth, if
put together in a coherent analysis, it may at least point toward a
class of desirable policy decisions and, on occasion, even a unique de-
cision. Finally, the chances of the decision maker erring are likely
to increase progressively through time if he disregards benefits assess-
ments in his policy choices.
The last of the above arguments for public decision making giving
more attention to benefits assessments rests on two considerations that
111-39
-------
have received a good deal of attention in the recent economics lit-
erature which has been synthesized by Krutilla and Fisher, 1975.
The first of these considerations is that for a great number of peo-
ple man-made environments do not substitute for natural environments
over a quite wide range of availability of the latter. This range
of nonsubstitutability or very low substitutability appears to in-
i. rease with respect to realistic increases in per capita incomes.
Furthermore, there are strong theoretical grounds for arguing that
improvements in man's production technologies tend to enhance the
supply of man-made goods and services while leaving the supply of
goods and services produced by natural environments either unaf-
fected or causing this supply to decline. The implication is again
Jhat the value of natural environments relative to man-made environ-
ments will tend to increase over time. A casual approach by public
decision makers to benefits assessment can therefore become increas-
ingly costly to the citizenry whose welfare the decision maker is
supposed to be enhancing.
2. The U.S. EPA Program
In view of its potentially important role in establishment of
environmental policy, one is moved to ask why benefits assessment
research has not progressed farther in the recent past and why, if
appearances are real, it has now been accorded a rather low research
priority by U.S. EPA. Several plausible reasons exist for this state
of affairs.
It's possible that skepticism by natural scientists about the
discipline of economics has been a factor in U.S. EPA's reluctance
to use economic models for benefits assessment while amicably greet-
ing models from biomedicine, meteorology, and other areas that pro-
vide estimates with larger errors and greater sensitivity to extreme
events than do economic models. Such a situation is unlikely to be
ameliorated unless someone who understands economic analysis is pro-
vided equal opportunity with natural scientists to formulate U.S.
EPA's research programs.
111-40
-------
An additional factor that may have contributed to the relative
lack of U.S. EPA concern about benefits assessment is the Clean Air
Act itself. Of the many objects of preference individuals have, this
Act can readily be read as elevating health to a place where little
else can be considered in research activities or control policies un-
til the public health is fully protected [see Sections 103(a) and 109
(a)]. Consideration of economic considerations is relegated to vague
statements about the public welfare while explicit reference to epi-
demiology, toxicology, etc. is made. The desirability of the vague
statements and the explicit references is not at issue. The point
is that the Act is probably interpreted as saying that economic con-
siderations are, at best, of secondary importance in air pollution
problems.
This, of course, is not true, for the Act nowhere inhibits eco-
nomic considerations with respect to whether to set, for example, an
ambient standard or a source performance standard, or whether or not
to set any standard whatsoever. It is here that benefits assessment
can be of greatest value to decision makers. Nevertheless, research
on the economics of air pollution control benefits has not yet pro-
duced a set of results regarded to be sufficiently compelling to war-
rant their serious use in air pollution control planning.
U.S. EPA appears to have lacked a successful research strategy
in the economics of air pollution control benefits. The two main fea-
tures of any existing strategy seem to have had speed of-acquisition
of results and accumulation of piecemeal, uncoordinated, isolated
studies by individual researchers. The first of these features im-
plies a failure to recognize that any serious analytical effort takes
a fair amount of time and that crash programs are not likely to be
successful. As someone has pointed out, a crash research program
bears at least a family resemblance to an effort to have a baby born
in one month by getting nine women pregnant. U.S. EPA's efforts in
biomedical and technological research have perhaps followed more nor-
mal and acceptable fertilization procedures.
Ill-'
-------
The result of the piecemeal feature has meant that a random col
lection of studies, highly variable in quality and significance, has
been produced. These isolated efforts do not serve effectively as
building blocks to be put together in a cumulative fashion for sub-
sequent studies, nor are they particularly useful for decision mak-
ing purposes.
In fact, such an effort can fruitfully be viewed as an invest-
ment-production process characterized by considerable uncertainty
and stochastic elements (i.e., surprises, disappointments, and er-
rors). Naturally, in this investment-production process, as high
a return as possible is desired. However, that which constitutes
a payoff and the means by which this payoff is to be maximized,
cannot be adequately characterized, without knowing the objectives
of those who will use any information that is generated. Other-
wise, one can't know whether the research emphasis given various
regions, pollutants, sufferer classes, and analytical and empiri-
cal refinements correspond to those most useful to decision makers.
In the absence of knowledge of explicit objectives, my major
research recommendation is that the U.S. EPA concern with aggregate
national control benefits should, for the moment, be limited to
development of a plan for microeconomic benefits research that is
explicitly tied to economic theory and which will permit consis-
tent aggregation within a framework accounting for the intrinsic
uncertainty of all benefits estimates. My views as to what consti-
tutes the possibly effective long- and short-term components of
this strategy have been set forth earlier in this paper.
111-42
-------
F. REFERENCES AND BIBLIOGRAPHY
Acton, J. P., "Nonmonetary Factors in the Demand for Medical Ser-
vices: Some Empirical Evidence," Journal of Political Economy
83:595-614, June 1975.
Alchian, A. A., and H. Demsetz, "Production, Information Costs, and
Economic Organization," The American Economic Review 62:777-795,
December 1972.
Anderson, R. J., Jr., and T. D. Crocker, "The Economics of Air Pol-
lution: A Literature Assessment," in P. B. Downing, ed., Air
Pollution and the Social Sciences, Praeger Publishers, New York,
133-166, 1971.
Arrow, K. J., and F. Hahn, General Competitive Analysis, Holden-
Day, Inc., San Francisco, 1971.
Bohm, P-, "Estimating Demand for Public Goods: An Experiment,"
European Economic Review 3, June 1972.
Brown, R. S., and S. Enke, "Economic Worth of Preventing Death at
Different Ages in Developing Countries," Journal of Biosocial
Science 4:299-313. July 1972.
Clarke, E. H., "An Approach to the Demand for Public Goods," Pub-
lic Choice 9:84-93, 1971.
Cobb, F. W., et al., "Photochemical Oxidant Injury and Bark Beetle
Infestation of Ponderosa Pine, II, Effect of Injury upon Phy-
sical Properties of Oleoresin, Moisture Content, and Phloem
Thickness," Hilgardia 39:147-149, 1968.
Council on Environmental Quality, Third Annual Report, Washington,
D.C., August 1972.
Crocker, T. D., Urban Air Pollution Damage Functions: Theory
and Measurement, U.S. Environmental Protection Agency, Office
of Air Programs, June 1971.
Crocker, T. D., "Climate and Household Expenditures," in T. A.
Ferrar, ed., Climate and the Urban Economy, Wiley Intersci-
ence, New York (forthcoming).~~
Crocker, T. D., "Contractual Choice," Natural Resources Journal
13:561-577, October 1973.
Crocker, T. D., "Benefit-Cost Analyses of Benefit-Cost Analysis,"
in H. M. Peskin and E. P. Seskin, eds., Cost-Benefit Analy-
sis and Water Pollution Policy, The UrbarPInstitute, Washing-
ton, D.C., 341-360, 1975.
111-43
-------
Davis, K., and L. Russell, "The Substitution of Hospital Care for
Inpatient Care," Review of Economics and Statistics 54:109-120,
May 1976.
Dhrymes, P. J.,^et al., "Criteria for Evaluation of Econometric
Models," Annals of Economic and Social Measurement 1/3:291-324,
1972.
Fama, E. F., and J. D. MacBeth, "Risk, Return, and Equilibrium:
Empirical Tests," Journal of Political Economy 81:607-636, 1973.
Feldstein, M. S., "An Economic Model of the Medicare System," Quar-
terly Journal of Economics 85:1-20, February 1976.
Fisher, F. M., and K. Shell, The Economic Theory of Price Indices.
Academic Press, New York, 1972.
Freeman, A. M. Ill, "On Estimating Air Pollution Control Benefits
from Land Value Studies," Journal of Environmental Economics
and Management 1:74-83, May 1974.
Green, H. A. J., Aggregation in Economic Analysis: An Introductory
Survey. Princeton Univ. Press, 1964.
Grossman, G., The Demand for Health: A Theoretical and Empirical
Investigation, Columbia University Press, 1972.
Grossman, M., "On the Concept of Health Capital and the Demand for
Health," Journal of Political Economy 80, March/April 1972.
Jaksch, J. A., Outpatient Medical Costs Related to Air Pollution in
the Portland-Vancouver Area, Department of Agricultural Econo-
mics, Oregon State University, Corvallis, Unpublished Ph.D.
Dissertation, 1973.
Krutilla, J. V., and A. C. Fisher, The Economics of Natural Envi-
ronments, The Johns Hopkins Press, Baltimore, 1975.
Lave, L., and E. Seskin, "Air Pollution and Human Health," Science
j69_:723-733, 21 August 1970.
Lichtenstein, S., "Bases for Preferences Among Three Outcome Bets,"
Journal of Experimental Psychology 69:162-169, 1965.
Lind, R. C., "Spatial Equilibrium, Rents, and Public Program Bene-
fits," Quarterly Journal of Economics 87:188-207, May 1973.
Marschak, J., and R. Radner, Economic Theory of Teams, Yale Uni-
versity Press, 1972.
Marshall, A., Principles of Economics, 8th edition, McMillan and
Co. , London, 1961.
111-44
-------
Mercer, L. J., and W. D. Morgan, Reassessment of the Cross-Florida
Barge Canal: A Probability Approach,, Department of Economics,
University of California, Santa Barbara, June 1975.
Miller, P. R., et al., "Photochemical Oxidant Injury and Bark Beetle
(Coleoptera scolytidae) Infestation of Ponderosa Pine III, Ef-
fect of Injury upon Oleoresin Composition, Phloem, Carbohydrates,
and Phloem pH," Hilgardia 39:135-140, 1968.
Mohring, H., "Alternative Welfare Gain and Loss Measures," Western
Economic Journal 9:349-368, December 1971.
National Academy of Sciences, Coordinating Committee on Air Quality
Studies, "The Effect of Uncertainty on Project Evaluation and
Design," Air Quality and Automobile Emission Control, Vol. 4,
The Costs and Benefits of Automobile Emission Control, prepared
for Committee on Public Works, U.S. Senate, Washington, D.C.,
427-470, September 1974.
Nelson, J., Residential Choice, Hedonic Prices, and the Demand for
Urban Air Quality, a paper presented at the Third World of Con-
gress of the Econometric Society, Toronto, Canada, 16-20 August
1975.
Niskanen, W. A., and S. H. Hanke, On the Use of Land Values in Bene-
fit-Cost Analysis, Department of Economics, University of Cali-
fornia, Berkeley (undated).
Oi, W., "An Essay on Workman's Compensation and Industrial Safety,"
Supplemental Studies for the National Commission on State Work-
man's Compensation Laws, Washington, D.C., 41-106, 1975.
Pesaran, M. H., "On the General Problem of Model Selection," The
Review of Economic Studies 41, April 1974.
Peskin, H. M., Accounting for the Environment: A Progress Report,
paper prepared for the Second National Symposium on Corporate
Social Policy, Chicago, Illinois, 3-5 October, 1974.
Phelps, C. E., and J. Newhouse, "Coinsurance, the Price of Time,
and the Demand for Medical Services," Review of Economics and
Statistics 56:334-342, August 1974.
Pollak, R. A., and M. L. Wachter, "The Relevance of the Household
Production Function and Its Implications for the Allocation of
Time," Journal of Political Economy 83:255-278, April 1975.
Pouliquen, L. Y., Risk Analysis in Project Appraisal, The Johns Hop-
kins Press, Baltimore, 1970.
Pruitt, D. C., "Pattern and Level of Risk Taking in Gambling Deci-
sions," Psychological Review 69:187-201, 1962.
111-45
-------
Ramsey, J. B., "Classical Model Selection Through Specification
Error Tests," in P. Zarembka, ed., Frontiers of Econometrics,
New York: Academic Press, New York, 1974.
Randall, A., et al., "Benefits of Abating Aesthetic Environmental
Damage," Agric. Exp. Sta. Bull. 618. New Mexico State Univer-
sity, Las Cruces, New Mexico, 1974.
Rosen, S., "Hedonic Prices and Implicit Markets: Product Differ-
entiation in Pure Competition," Journal of Political Economy
132:34-56, January/February 1974.
Rosett,.R. N., and L. Huang, "The Effects of Health Insurance on
the Demand for Medical Care," Journal of Political Economy 81:
281-305, March/April 1973.
Schmidt, P., "Choosing Among Alternative Linear Regression Models,"
Atlantic Economic Journal 1, July 1974.
Sonquist, J. A., et al., Searching for Structure, Institute for
Social Research, University of Michigan, 1973.
Stark, R. W., et al., "Photochemical Oxidant Injury and Bark Beetle
Infestation of Ponderosa Pine, I, Incidence of Bark Beetle In-
festation of Injured Trees," Hilgardia 39:121-126, 1968.
Thompson, E. A., "The Perfectly Competitive Production of Collective
Goods," Review of Economics and Statistics 5JD:1-12, February 1968
U.S. Environmental Protection Agency, The Clean Air Act, As Amended
June 1974, Washington, D.C., 1974.
111-46
-------
ATTACHMENT. AN ILLUSTRATION OF SOME ISSUES IN BENEFITS ASSESSMENT
This attachment seeks to illustrate the application of the fore-
going discussion to a real world problem by analyzing the benefits ac-
cruing from controlling a single class of pollutants in a specific
locality.
1. Introduction
Conventional microeconomic theory constructs an extremely
general conceptualization of the consumer or producer decision mak-
ing process. For example, in the case of the consumer, households
are assumed to allocate their budgets so as to maximize the utility
they derive from the purchase and consumption of goods and services
subject to the constraints imposed by the resources to which they
have access as well as the prices of the commodities available to
them. Since the conceptualization is extremely general and,there-
fore,abstract, many issues that are of importance to particular ap-
plications are assumed to be irrelevant in order to focus attention
on the key principles governing consumer choice.
However, once the framework is applied to a specific set of
decisions, such as those involving the behavior of sufferers from
air pollution, then an array of details must be reviewed. These
details are indeed relevant to the usefulness of the models on
which the decisions are to be based and to the behavioral outcomes
that these models would predict. Inclusion of these details con-
siderably complicates the analytic, and even more the econometric
treatment of the benefits of air pollution control; but outright
omission in the interests of tractability may grossly misrepresent
the phenomena involved. By focusing upon a particular type of mar-
ket in an identifiable locale, the reader should become more sensi-
tive to the most important of the aforementioned details and the
difficulties one faces in trying to account for them.
111-47
-------
The natural framework of analysis for the economist is the
market, the interaction of buyers and sellers. Air pollution af-
fects behavior in a wide variety of markets, each of which has its
own unique structure. As used here, the term "markets" is meant
to apply to all air pollution control benefits studies employing
market-established prices or some surrogate as a means of estab-
lishing benefit assessments. The term is thus meant to encompass
each of the methods of assessment (technical coefficients, market,
opinion surveys, litegation, political expressions, and Delphi)
that Waddell (1974) distinguished. The presumption is that a
model, whether implicit or explicit, of a market is present when
any one of these methods is employed in a given situation. That
is, at its roots, any use of one of these methods involves some
perspective of the structure of buyer and seller interaction.
The benefits of air pollution control in the particular mar-
ket to be discussed are probably trivial relative to the benefits
in markets such as human health and life, materials, housing, and
assorted other buyer and seller interactions in which air pollu-
tion is thought to have a significant role. Nevertheless, this
case is very illustrative for each of its peculiarities are pre-
sent in one or more of the markets that are typically studied for
purposes of estimating the benefits of air pollution control. In
addition, this case illustrates the difficulties encountered in
analyzing the recreation benefits of water pollution control which
have been mentioned in the other papers of the report. Recreation
benefits are, it is frequently asserted, a major portion of the to-
tal benefits of water pollution control.
2. Presentation of the_P_rob1em
The San Bernardino Mountains of Southern California serve as
one of the primary outdoor recreational resources available to the
approximately 11 million people of the Los Angeles Basin. On the
11-48
-------
slopes of these frequently extremely steep mountains grows a
highly varied set of plant communities ranging from hot desert
to sub-artic. The lower portions are dominated by a fairly dense
coniferous forest consisting primarily of ponderosa pine and white
fir. Here are found the most heavily used outdoor recreational
facilities, including hiking trails, campgrounds, lakes, small
mostly intermittent streams, and arrays of publicly and privately
provided outdoor recreational amenities, such as cabins and res-
taurants.
Most users of these facilities would probably agree that the
ponderosa pine, an esthetically pleasing tree that is large, tall,
shady, and straight-stemmed, contributes significantly to their
outdoor recreational experiences. Recently, however, a relation
has been found between ozone, or ambient oxidants, and suscepti-
bility of this pine to attack by insects. It is thought that
the combined action of insects and pollutants results in higher
rates of needle drop and death than the action of insects alone.
(Miller et al. , 1968; Stark et a!., 1968; Cobb and Stark, 1970).
Let us presume that the biological relationship between photo-
chemical oxidants and ponderosa pine morbidity and mortality has
been reasonably well established. The problem facing the econo-
mist, then, is to tie to this biological relation a welfare mea-
sure that simultaneously registers the values of outdoor recrea-
tionists and is meaningful to air pollution control decision makers.
That is, the economist's task is to assess the value outdoor recrea-
tionists place upon the presence of healthy ponderosa pine.
This task is not easy, since neither air pollution control,
the services of ponderosa pine forests to outdoor recreation ex-
periences, nor recreation experiences themselves are, for the most
part, exchanged in organized, explicit markets permitting direct
observation of the relative values households attach to such goods
and services. One must therefore infer the relative values house-
holds attach to ponderosa pine in various states of health from
111-49
-------
observations on the rates at which these households voluntarily
substitute one state for another and the market values of the
resources they expend in performing these substitutions.
However, in order to make these inferences, all other fac-
tors that influence the voluntary rate of substitution must be
accounted for. This accounting requires the construction of a
model of the nature of buyer and seller interaction in outdoor
recreation. Some of the structural features of the outdoor
recreational market that, ideally, should be integrated into
this model are described below. With a bit of reflection, the
reader will recognize that similar features (in kind, if not in
degree) are to be found in many of the actual markets studied in
the past to ascertain the benefits of air pollution control. For
example, whenever the term "outdoor recreation" appears, the term
"housing" or -"state of health" would generally serve just as well.
(Vars and Sorenson, 1975).
3. Multi-Dimensional Heterogeneity
The sites a household considers in planning its outdoor re-
creational activities are heterogeneous rather than homogeneous.
That is, they exhibit great differences in a number of different
dimensions to which the household typically attaches value. Dif-
ferences in the physical recreation sites themselves are obvious
enough. For example, these differences may be manifested in size,
number of campsites, topography, forest type and condition, and
other attributes.
But an outdoor recreational site embodies more than the phy-
sical site itself. It is a composite consisting of the site and
its accessibility with respect to alternative recreational sites
and the household's residence. Accessibility of sites differs
greatly in terms of the pecuniary and temporal outlays required
and in terms of the quality of transport mode and route. More-
over, outdoor recreational sites, as well as the transport modes
111-50
-------
and routes that provide access differ greatly in the cultural
backgrounds of the people who frequent them, the assortment of
privately owned goods available, the quality and quantity of
common property and publicly owned goods provided, and conges-
tion.
The state of affairs described above is really not alto-
gether unusual in applied economic analysis. Just about any en-
tity a household values can be viewed as a composite of several
attributes. Frequently, these components will differ substan-
tially among various sellers and among various buyers. Econo-
mists nevertheless habitually aggregate these heterogeneous enti-
ties into units of analysis they term goods, industries, etc.
The aggregate unit of analysis is usually probably a good approxi-
mation of the most informative unit for the purpose at hand. How-
ever, in the economic analysis of outdoor recreation markets, the
question of the most informative units of anlaysis must be ap-
proached with a good deal more caution than is customary. The
reasons are several.
First, at least to the casual observer, households appear to
exhibit great differences in the utility they derive from outdoor
recreation. They differ greatly in the value they attach to any
form whatsoever of outdoor recreation, as well as in the relative
importance they associate with various forms of outdoor recrea-
tion. In addition, for any particular form (e.g., viewing scenery),
they differ in their preferences for the various components of
the form. All this implies that the differences observed among
outdoor recreational sites influence the market. To some extent,
particularly in the case of accessibility, the observed differences
might not be matters of chance but are instead demand-induced.
Second, the consumer and the producer are able to exercise
much less discretion as to the components of the commodity in the
case of outdoor recreation. The household that consumes outdoor
recreation experiences has substantial discretion with respect to
111-51
-------
the set of alternative recreational opportunities, but, once the
choice is made, many subsequent events are outside its control.
For example, when arriving in the mountains, the household is un-
able to control the congestion at a site it does not own. There-
fore, it is unable to control the extent of deterioration in the
quality of services it had expected to obtain at the site. Since
this can greatly affect the utility of recreation and typically
occurs without the direct intermediation of the market, there
exists a large potential for substantial externalities in out-
door recreational markets.
Third, from the perspective of the individual household, the
great variety of outdoor recreational sites available in Southern
California means that they are imperfect substitutes for one an-
other. Any fairly large collection of sites is likely to exhibit
a range of substitutive relations extending from rather close to
nonexistent. Is the beach a close substitute for the mountains?
In addition, given the near impossibility of untying the packages
into which the components of most sites are bound and, given the
limited number of sites the household is likely to consider, there
may exist a major difference in the degree of substitutability that
pairs of households assign to the same pair of sites. This implies
that a single outdoor recreation market does not exist. Instead,
there seems to be a set of markets that are more or less related to
one another.
One serious consequence of this market segmentation is that
the supply and demand relations that constitute each market seg-
ment can behave in different fashions across segments. Price
elasticity of demand for a particular activity may be high in
one segment and low in another; one segment may have excess de-
mand for a certain activity, while another may have excess sup-
ply. All this implies that prices, whether implicit or explicit,
may vary across the segments. Moreover, an exogenous change that
impacts forecefully on one segment may have no impact whatsoever
in another segment.
111-52
-------
This segmentation of the outdoor recreation market is nour-
ished by the high cost for households of information about market
conditions. Information is clearly an important factor in house-
hold decisions when one recognizes the great variety of outdoor
recreational opportunities and the differences in preferences
among households for these opportunities. However, it seems un-
likely that households can afford to survey all or even a major
part of their potential opportunities. Instead, they limit their
search to some portion, a delineation probably based on various
rules of thumb reflecting prevailing knowledge about opportunities
that are good substitutes for one another.
4. Costs_of Market Participation
To a perhaps greater extent than for many other commodities,
the act of participation in the various segments of the outdoor
recreation market is costly. For several reasons, the cost of
the act of participation tends to harden the discrepancies in
prices across market segments. First, as previously noted, since
heterogeneity in available outdoor recreational opportunities is
pervasive and considerable and tastes regarding these differences
are quite important, substantial search by a household may be re-
quired to assure a satisfactory recreational experience.
Second, the consumption of outdoor recreation generally re-
quires that the household remove itself from its permanent resi-
dence and selectively travel to one or more sites at which the
chosen combination of the components of the recreation package
afe thought to exist. Outdoor recreation does not have the port-
ability of the typical commodity. Since the commodity cannot be
carried from one location to another, the household experiences
pecuniary and temporal costs in overcoming the distance necessary
to match household to site. Third, even after the household's
search has been terminated and the distance to the site overcome,
actual participation usually requires the purchase of complemen-
tary inputs.
111-53
-------
These costs imply that the prospective gains from becoming
an active participant in the market may have to be quite substan-
tial in order to warrant such participation. The values of those
who do not participate are not registered in the market. The
prospective gains the household perceives from participation can
arise from changes in the household's preferences with respect to
different combinations of outdoor recreation packages and with
respect to recreational versus non-recreational commodities. They
can also arise from changes in the components and in the implicit
and explicit prices of the different packages available on the mar-
ket.
Barring changes of these sorts, the high cost of market par-
ticipation suggests that participation is likely to be intermittent
rather than continuous for most households. This, in turn, im-
plies that the set of households one observes participating, and
thus determining the outcome of market transactions in any fairly
short time period, might not be representative of the entire pop-
ulation of households.
5. Institutional and Temporal Constraints
Apart from the problems that multi-dimensional heterogeneity,
information, and the costs of market participation introduce, the
outdoor recreational market bears little semblance to a perfectly
free market. Indeed, the outdoor recreational market in Southern
California has been subjected to a number of public institutional
controls, including assorted health and safety regulation, non-
price rationing., such as queuing, and zoning regulations intended
to inhibit congestion. In effect, these controls reduce the do-
main over which the market is able to exercise discretion in ad-
justing to changing situations.
Although the reductions in negative externalities which these
controls permit may well exceed any losses in efficiency which the
controls engender, such losses are nevertheless undoubtedly present
111-54
-------
They involve the losses present when every household is forced
to consume and/or produce in more or less the same manner and
the flexibility that is lost when the only parameter subject to
occasional adjustment is the value of the regulation. Moreover,
to the extent they reduce substitutability among sites, the con-
trols intensify whatever tendency toward market segmentation al-
ready exists.
The constraints the household faces while recreating are
not only budgetary and those publicly designed to shape its ac-
tivities. Most households are unable to allocate their time in
any arbitrarily desired way. Instead, the household's free time
for outdoor recreation is structured and limited by inalterable
work schedules. In the case of family units, work and education
schedules may interact, so as to curtail family recreational time
futther. Even individuals who are self-employed are constrained
to the extent they must interact with other individuals and busi-
nesses .
6. Conclusion
All of the issues discussed above are relevant to the assess-
ment of effects that changes in the health state of the ponderosa
pine forest have upon the choice behavior of outdoor recreation-
ists in Southern California. The discussion is, in fact, germane
to most, if not all, markets that are studied to assess the bene-
fits of air pollution control. Nevertheless, a model capable of
capturing the full impact of the multi-dimensional heterogeneity,
cost of information and market participation, and institutional
and temporal constraints of the outdoor recreation market upon
the household's decision problem would indeed be unusual. In
any model that is to have real application, the treatment of many
and sometimes most of the issues raised above must of necessity
be conjectural. That is, in his applications of a model of bene-
ficiary behavior, the investigator must be willing to take a sub-
set of these issues as already settled for the application in
111-55
-------
question. Both the limited availability of suitable data and
the limited abilities of investigators in model construction dic-
tate the failure to comprehend in each instance the range of these
issues.
Most fundamentally, in any attempt at benefit assessment one
must examine the role the assessment is to have in the decision
making process. In effect, one must know with some precision the
criteria to be used to evaluate the assessment as well as the ob-
jectives of the assessment effort. Such knowledge permits the in-
vestigator to isolate those facets and issues with significant
implications for decisions from the less important ones. (Crocker,
1975).
111-56
-------
IV. BENEFITS OF WATER POLLUTION CONTROL
Joe B. Stevens
Oregon State University
A. INTRODUCTION
Despite the richness of the literature in applied welfare econo-
mics and benefit-cost analysis, there is currently very little concrete
knowledge of the extent of economic benefit which would result if water
quality were brought up to state and Federal standards. The develop-
ment of what knowledge does exist has been very asymmetric. Estimates
of production losses, including conventional waterborne disease, are
reasonably firm; unfortunately, only a small fraction of the potential
benefits appear to be in these areas. The predominant share of the
potential benefits appear to be in the categories of recreation, aesthe-
tic uses, and option demands, where the methodologies for simulating
consumers' willingness to pay for quality improvements are much less
well developed.
The lack of adequate, general izable behavioral relationships be-
tween water quality and consumer behavior, and hence, willingness to
pay for quality improvements, is thus the predominant problem in as-
sessing pollution control benefits. There is simply inadequate knowl-
edge concerning why people "consume" leisure-type environmental acti-
vities, and especially how changes in water quality (objective and/or
perceived) affect these consumption activities. Some creditable work
has been done in estimating recreation benefits from pollution control
in small areas, but extrapolation of these studies to national aggre-
gates is not currently possible with any degree of precision. The mon-
etization of aesthetic and option demand is, at present, even less well
developed than for active recreation uses.
Increased precision in the measurement of consumption benefits
(recreation, aesthetics, option demand) will come slowly and probably
fortuitously in the absence of an expanded U.S. EPA role in benefit
estimation. This is particularly true for the evaluation of efficiency
IV-
-------
at an aggregate level. Decentralized academic researchers do not have
an incentive system which encourages them to derive "premature" esti-
mates of national benefits p_£ to structure their partial equilibria
analyses to facilitate this derivation by U.S. EPA or other institu-
tions. Asking even the best consulting firm to extrapolate existing
partial analyses when these were not designed to be extrapolated is a
hopeless strategy.
Thus, U.S. EPA should assume an expanded role in designing and
funding high-quality analyses of consumption benefits, where these
analyses have the express objective of being extrapolated to national
levels. As a complement to this activity, a basic commitment should
be made by EPA to establish a long-term "behavioral monitoring system"
analogous to that established by the U.S. Geological Survey for assess-
ing water quality. Common socio-economic and physical data sets need
to be maintained over time at sites where water quality changes are, in
fact, occurring. The rationale is that a much more adequate knowledge
of behavioral response with respect to water quality changes must be
established before the efficiency of water quality enhancement can be
assessed with any substantial precision.
IV-2
-------
B. CONCEPTUAL FOUNDATIONS
This section is intended only as an overview of the welfare eco-
nomic foundations of benefit estimation in water pollution control,
since several other sources treat this essential topic in much more
detail. It identifies and synthesizes the implications of general
and specific problems affecting current benefit estimates, by cate-
gory of damage (e.g., health, recreation), and offers a set of sug-
gestions for improving the quality of future estimates.
1. Role of Benefit-Cost Analysis
Historically, benefit-cost analysis has been a valuable
tool for analyzing the economic worthwhileness of alternative
publjc policies. Although its origin can be traced to legisla-
tive mandates (e.g., the Flood Control Act of 1936), rather than
theoretical modeling, substantial theoretical and empirical lit-
erature on benefit-cost analysis now exists. The relevance of
this literature to the analysis of benefits and costs of water
pollution policy has been documented in a 1973 EPA Symposium,
the proceedings of which are now in print and available to admin-
istrators and researchers (Peskin and Seskin, 1975).
A particular policy can be supported by two antagonists,
each of whom views the policy as contributing to a different ob-
jective. Also, it is uncommon in the political world that the
real objectives of the policy's promoters will be explicitly sta-
ted, if in fact, the promoters are actually capable of articula-
ting these objectives. Indeed, the "muddling through" or "dis-
jointed incremental" approach to public policy (Lindblom, 1959)
rests largely on the notion that agreement on "means" (i.e., poli-
cies) is primary, and that agreement on "ends" (i.e., objectives)
is secondary and perhaps even unnecessary.
Benefit-cost analysis, on the other hand, presupposes that
objectives d_p_ matter in the public policy process. Simply stated,
IV-3
-------
the benefit-cost analyst says: "If X (or Y or Z) is your objec-
tive, then that particular objective is enhanced (or impeded)
by the policy or program under consideration." The variety of
objectives to which benefit-cost analysis can be addressed is
limited, yet the tools are powerful if economic efficiency is
a relevant objective.
Three possible objectives of water pollution control poli-
cy are:
• To make efficient use of resources associated with
water use
• To distribute goods and services, including those
which are water-related, in an "equitable" manner
• To avoid ecological irreversibilities.
Benefit-cost analysis can make a potential contribution to
each of these. However, a "state of the art" assessment of bene-
fits from water quality improvement can most productively be
focused on "national efficiency" benefits, or those which would
arise from the increased value of goods and services made possi-
ble by water pollution policies. Economic science is far more
advanced in contributing to this objective than to the two other
objectives.
Differential incidence of benefits and costs among various
people or groups may arise from different water pollution poli-
cies, and decision makers may choose to look at the problem par-
tially or solely in this context. The state of economic science
at present is inadequate to define distributional benefits in
a normative sense, but instead is limited to a positive descrip-
tion of "who gains," "who loses," and "how much." Such descrip-
tions are inherently useful, and should be included in any assess-
ment of pollution control benefits and/or costs. Finally, benefit-
IV-4
-------
cost analysis may be used to identify the costs of maintaining a
designated water quality standard, even though the benefits from
maintaining this standard may not be fully quantifiable or even
perceived at this time (Krutilla, 1967).
2. Estimation of Benefits
Efficiency benefits should be defined in terms of the total
value to users of increases in the national output of goods and
services, where total value is defined in terms of willingness of
users to pay for the additional output. Because it is often not
feasible to measure individual demand functions, and hence, the
willingness to pay by users, four alternative techniques for as-
sessing these benefits have been devised.
If markets for the goods and services exist, and if the
additional output does not have a significant effect on price,
then market prices can be used to assess benefits. This technique
will be of little value in assessing water pollution control bene-
fits, since markets are seldom used directly to determine water
quantities or qualities. Simulated markets, including demand func-
tions, supply functions, and/or simulated market outcomes, can be
used as a proxy for real markets. While the methodology for this
technique is rudimentary in its development, it appears to be the
only relevant technique for evaluating water quality benefits
where outputs are "consumed" directly by final users. In particu-
lar, benefits to recreation and aesthetics will need to be evalua-
ted by this method.
If the output of investment is an intermediate good, rather
than one intended for final consumption, then changes in the net
income of producers (or consumers), with and without pollution,
may be used to reflect benefits. This method will be useful pri-
marily for determining benefits from controlling industrial
IV-5
-------
and agricultural pollution, and from reduced waterborne disease.
Finally, if computation of net income changes is not feasible,
the cost of the most likely alternative for attaining the same
objective can be used to measure benefits. The rationale behind
this method is that benefits are obtained to the extent that re-
sources are thereby released for the creation of other goods and
services.
A second category of benefits, often termed "secondary bene-
fits," are those which might result from the use of resources
which would otherwise be unemployed or underemployed (Haveman and
Krutilla, 1968). In general, one would not expect to find sub-
stantially improved employment opportunities for otherwise under-
employed resources, if national water quality were restored to
meet Federal and state standards. The reason for this expectation
is that water pollution has probably occurred gradully enough that
the market has had time to re-allocate resources away from polluted
areas.
3. Special Considerations
If market data could be relied on extensively for calculating
benefits, care would need to be exercised to see that the markets
were competitive and to account for project-induced changes in price
Under such circumstances, welfare economic theory provides a basis
for asserting that prices reflect social values, provided that one
accepts the prevailing income distribution as "proper". In the
case of water pollution, however, much reliance in benefit evalua-
tion must be placed on simulation of markets for non-market goods
such as outdoor recreation, and calculation of net income changes.
Therefore, special attention would need to be given to several
criteria which relate to the internal consistency of the analysis.
Thus, inferences as to technical relationships should be
at the highest "state of the art" possible within the relevant
IV-6
-------
discipline. For example, the nature and extent of man-made pollu-
tion should be identified, along with marginal treatment costs for
these pollutants. To the extent feasible, inferences of technical
relationships should be both complete (e.g., no major omissions
in types of waterborne diseases) and capable of replication.
Although direct market prices for goods and services produced
by water pollution control will be of limited value in the direct
assessment of benefits, some reliance must of necessity be placed
on real or synthesized price data in the simulation of markets for
certain goods and services. Examples include:
• Health
cost of medical treatment
value of time lost to illness
value of human life (present value
of future earnings)
• Outdoor Recreation
costs of vehicle operation and travel
value of time spent in travel
• Production
marginal costs of waste treatment for
various types and levels of effluents.
Use of price data where market prices are artifically set,
either by imperfect competition or through public policy, should
theoretically be discouraged. This is easier said than done, when
one considers this varied range of contrived market outcomes:
t Health
restricted supply of medical personnel
t Recreation
OPEC policy on U.S. oil imports
auto prices administered by U.S.
manufacturers
IV-7
-------
Production
degree of price competition among sup-
pliers of pollution control equipment
price support levels of farm crops lost
to siltation and brackish water.
About the only reasonable expectation is that the analyst be able
to identify major price imperfections where knowledge of such has
been documented (e.g., farm price supports).
Where price data are relevant in the synthesis of proxies for,
"willingness to pay," the expected level of such prices over the
lifetime of the project needs to be identified. Use of current
prices such as, for example, the current cost of travel, may lead
to inflated or deflated estimates of benefits, depending on what
constitutes a "reasonable" long-term price expectation.
To the extent possible, the analyst should identify signifi-
cant technological externalities which would be alleviated by ex-
panded policies of water pollution control. In fact, this should
form the substance of benefit calculations for improvements in wa-
ter quality; these improvements would reduce the level of negative
externalities to users (e.g., disease, degraded recreation oppor-
tunities). National efficiency benefits, regardless of method of
calculation, should be net of pecuniary externalities (income trans-
fers). Especially if the "net income" approach is used in benefit
calculations, care should be taken that offsetting income losses are
included in the analysis. Also to be questioned or excluded are
windfall net rents which may accrue to fixed assets without corres-
ponding increases in outputs of goods and services.
Where the investment project for augmenting water quality lev-
els is expected to produce benefits extending over several years,
the annual increment (or decrement) to benefits should be identi-
fied, and the present value of current and future benefit streams
should be computed. Goods and services provided directly to con-
IV-8
-------
sumers through water quality restoration are of particular concern
here, in that the income elasticity of demand for different services
may vary widely.
With respect to human health, research indicates that the
value of the human agent has risen markedly over time (Schultz,
1968). So, too, has the demand for services to maintain the hu-
man agent, including health services. One would suspect a high
and rising income elasticity, then, for avoidance of waterborne
diseases. Within the category of outdoor recreation, there is
reason to suspect that a substantial range exists in the income
elasticities for different activities, including some "inferior"
goods (Stevens, 1966). Conceptually this poses a weighting problem
for assessment of benefits at the national level, where annual
changes in benefits are a function of the composition of current-
year benefits (by recreation activity), the expected income growth
in the economy, and the income elasticities of the different acti-
vities.
These time-related changes should be identified in computing
annual benefit streams. Inclusion of such calculations and break-
down of the requisite investment into installation cost and operat-
ing cost components would improve both the benefit and cost esti-
mates, since investment projects for improving water quality would
be capital-intensive and long-lived in nature.
IV-9
-------
C. GENERAL PROBLEMS
This and the following section are meant to be synoptic statements
on the state-of-the-art, rather than a detailed review of empirical
studies on benefit estimation.
1. State of Damage Functions
The most complete and useful review of empirical attempts to
estimate water pollution control benefits is Tihansky's chapter
in Peskin and Seskin (1975). To quote selectively from his con-
clusions: "This paper reviews an extensive compilation of major
benefit studies. From an original collection of almost two hund-
red studies, approximately sixty are selected as contributing most
to the conceptual framework and empirical insights of benefit eval-
uation. Of the selected studies, less than 30 percent are theore-
tically valid, but even fewer seem cognizant of the applicability,
let alone the existence, of welfare economics ... If only recrea-
tional studies are considered, fewer than 10 percent follow from
theory ... Although there is no theoretical basis for most benefit
values, equally discouraging is the paucity of empirically derived
damage functions. Only 20 percent of the surveyed literature de-
rives dose-response relations from on-site data. With recreation
studies, this proportion decreases to 10 percent."
Despite some allegations to the contrary, water quality data
are relatively abundant. The real problem, however, is the lack of
conceptually valid functional relationships between water quality
parameters and associated human behavior. We need to know:
• How economic activity affects the nature and con-
centration of waterborne pollutants
• How the latter affects behavioral perceptions of
water quality
• How this perception is translated into behavioral
decisions.
IV-10
-------
Welfare economics is a useful and integrative conceptual base
for benefit estimation, but existing facts often have multiple in-
terpretations since functional relationships well-founded in theory
have usually not been established. Generally speaking, welfare
economics is a "non-problem" in that the accumulation of theore-
tical insights has far outrun the empirical implementation of the
theory.
The current "state-of-the-art" in assessing water pollution
control benefits strongly reflects an asymmetric development in
terms of establishing such functional relationships. The reasons
are partly economic and partly institutional. At an earlier stage
of economic development in the U.S., the primary concern with water
quality was for human health and safety. The whole historical thrust
of the Public Health Service has to do with developing functional
relationships between pollutants and waterborne disease. Especially
with the more virulent and contagious diseases, a fairly simple be-
haviral response was assumed, i.e., avoidance.
As economic development progressed, natural resources were
increasingly regarded as items for consumption, rather than produc-
tion. At an institutional level, the Public Health Service con-
tinued to be concerned with the production of health. Elsewhere in
government, water-based leisure (and most other forms of recreation)
were evidently regarded as something to be done, not studied. In-
creased awareness of this knowledge gap led to the creation of the
Outdoor Recreation Resources Review Commission and subsequent efforts
elsewhere in government.
Within this set of poorly understood leisure-type phenomena
are those which relate changes in water quality to changes in con-
sumer behavior. In particular, damage functions which relate to
active recreational uses of water or to more vaguely defined aesthe-
tic uses have not been established to any substantial degree. Thus,
there is inadequate knowledge of the whole set of relationships
IV-11
-------
concerning why people "consume" leisure-type environmental activi-
ties, and how real and/or perceived water quality affects these con-
sumption activities. Individual researchers have done creditable
work in specifying particular water quality parameters and relating
these to particular recreational activities, but there is nothing
resembling a comprehensive research effort in this area.
The knowledge which does exist about behavioral responses with
respect to changes in water quality is fragmentary. More crucial
to the task at hand is that it deals with partial equilibria, i.e.,
variables which are crucial in the user's decision framework are
often assumed as constants in existing analyses. In particular,
most of the existing work on recreation benefits assumes a constant
level of water quality in bodies of water other than the particular
one being researched. In that this group of other sites probably con-
tains some which are substitutes for the one under study, the analy-
tical models are incompletely specified and, hence, less useful than
would be desired for purposes of national aggregation. The genera-
tion of national benefit estimates implies solutions which are more
general with respect to competing sites than is permitted by the
current literature.
This is not surprising, in that the immediate objectives of
research have usually been to measure costs and benefits of control-
ling specific water pollutants in specific bodies of water. Ten
years ago, when estimation of recreation benefits was in its first
primitive stages, it was a real challenge to perform the analysis,
let alone worry about extrapolation to higher levels of aggregation.
It is still generally acknowledged that consumers have a broader
preference set than just the particular river or lake under study.
The state-of-the-art is to assume away the influence of competing
streams and how recreationists might react to more general water
quality changes.
The needed knowledge of behavioral responses to changing
water quality levels has not progressed to the point where
IV-12
-------
reasonable extrapolation can be made with any degree of precision.
Although estimates of national benefits have been derived in the
past (e.g., Unger et a]_.. 1974), these estimates are based on ex-
trapolations which have neither a sound conceptual footing nor an
underlying accumulation of coherent empirical work. Tihansky's
(1975) review of such studies reveals that the bulk of water qua-
lity benefits seem generally to be in those areas where the "state-
of-the-art" is least developed, particularly in the areas of recrea-
tion, aesthetics, and option demands. The same conclusion is also
reached in this present paper.
2. Classification and Aggregation of Benefits
That proportion of specific pollutants which are man-made,
rather than naturally occurring, and the approximate control costs
for both man-made and natural pollution are often known with some
degree of precision. The general tendency has been to impute aver-
age (proportional) costs to control of man-made pollutants. How-
ever, marginal costs (over and above those required to treat natural
pollutants) would be a more appropriate measure. This can lead to
overstating the benefits of pollution control.
There is only a weak consistency in the quality of estimates
among the various benefit categories. Although health and produc-
tion benefits are relatively firm, or at least susceptible to re-
finement through further analysis, this is not the case with consump-
tion benefits (recreation and aesthetics). These estimates must
currently rely on highly simplifying assumptions in order to utilize
existing research not intended for purposes of extrapolation.
As the "state-of-the-art" develops in benefit estimation, knowl-
edge of the growth rates of benefit streams becomes increasingly im-
portant. Technological change in the future is likely to be greater
for production activities than for consumption activities. Hence,
the costs of pollution to production activities (and hence, potential
benefits) may be reduced over time as new technology is developed
IV-13
-------
and implemented to alleviate those problems. Considering health
as a "production" activity in a manner consistent with human capi-
tal theory, the per unit benefit values may rise over time as the
value of human capital rises relative to that of physical capital.
Some consumption benefit streams may grow and some may decline,
depending in part on the income elasticity of demand for water-
based services.
The decision context implicit in an assessment of national
benefits and costs is one of aggregate efficiency in water resource
use. The broad question is "Do the national benefits of cleaning
up our water outweigh the costs?" In reality, the issue of whether
or not this is the most productive question to be asked must be
faced. Obviously, a good answer to the larger question would be
immensely valuable to policy makers.
On the other hand, asking the broad question about an over-
all benefit-cost ratio may conceal the fact that there are a host
of individual benefit-cost ratios (some high, some low) for cleaning
up specific pollutants and for cleaning up or protecting specific
areas. If the latter is a more appropriate decision context, then
criticism of the state-of-the-art has to be mitigated to some degree.
Problems with the totality of behavioral responses become less cru-
cial, and a re-focusing on the more pressing responses is facili-
tated.
3. Other Problems
Even if the question of aggregate efficiency of water quality
enhancement could be answered definitively, this would be suffi-
cient information only if policy makers were interested solely in
economic efficiency. This is obviously not realistic, as most
policy makers are also interested in distributional issues. Hhe-
ther access to goods and services would be redistributed in a pro-
gressive or regressive manner among individuals is information
which is almost totally lacking in the benefits literature.
IV-14
-------
Similarly, pollution control might be progressive, if
the costs of pollution, to producers now paid through higher
product prices and/or utility rates, were met largely through
Federal expenditures financed by a reasonably progressive
tax system. On the other hand, pollution control could be
regressive if the dominant beneficiaries are white-collar,
urban recreationists. In addition, information is also
needed on redistribution at a more aggregate level, e.g.,
urban versus rural states, metropolitan vs. rural areas.
It is crucial that benefits and costs be estimated for a
variety of levels of water quality; this is not generally found
in the literature. Such a procedure would permit marginal analy-
sis by allowing water quality standards to be viewed as variables,
rather than basic parameters to be met at all costs. In light of
current concern about the high costs of meeting existing standards,
the estimation of benefits and costs for a range of water quality
standards becomes very crucial information.
IV-15
-------
D. SPECIFIC PROBLEMS
This section discusses problems inherent in estimation of speci-
fic benefit categories, including health, production, recreation, aesthe-
tic, and ecological benefits, and property values.
1. Health Benefits
It would appear that extensive pre-treatment of drinking water,
especially chlorination, has already greatly reduced the incidence
of conventional waterborne disease outbreaks, and that further defi-
nition of benefits depends on how accurately the following can be
quantified:
Extent of such diseases as hepatitis, gastroenteritis,
salmonellosis, and typhoid
Extent to which these outbreaks can be identified as
waterborne, at both current and standard levels of
water quality
Incidence of the outbreaks by morbidity and mortality
classifications
Assignment of unit values, in monetary terms, to the
above categories.
Current estimates of benefits in the above categories include
1 billion (Unger, et al. , 1974), 373 million (Liu, 1970), 356 mil-
lion (Lackner, 1973), and 120 million (Sokoloski, 1973). While
the variation in these estimates appears to be substantial, it is
largely due to different working assumptions about facts and func-
tional relationships.
Determination of the real extent of damages from conventional
diseases is tractable, requiring additional time and money. Such
an allocation is not of high priority, given the softness of the
benefit estimates in other categories, particularly recreation and
IV-16
-------
aesthetics. If such an allocation should be deemed desirable,
more attention should be given to the following:
• Assignment of mortality values according to human
capital theory, i.e., using expected present value
of future earnings as the measure of benefit
• Recognition that both the incidence of mortality
and expected lifetime earnings are age/sex/race-
specific rather than constant over the population
• Recognition that health services are generally
income-elastic in demand
• Recognition that costs of medical treatment have
risen secularly, as compared to other commodities.
The key problem in quantifying health benefits is how best
to deal with uncertain, long-term, and largely unknown water qua-
lity-health relationships, such as the presence of carcinogens
and certain inorganic compounds in water. Although a consumer's
perception of the importance of these events is subject to con-
siderable emotion, misunderstanding, and lack of an informed base
from which to make judgments, option value constructs should be
explored within this category as well as in the categories of re-
creation and aesthetics. The reasons for this view are that will-
ingness to pay can arise from desires for future as well as cur-
rent consumption, option demand can be viewed as willingness to
pay for future supplies of a specified good and non-carcinogenic
drinking water is surely one such good.
Although it was suggested above that conversion to monetary
values be based on a productivity approach (present value of future
earnings), an alternative willingness-to-pay framework is now being
developed in the literature (Mishan, 1971). An individual's willing'
ness to pay to avoid contracting cancer through drinking water, for
example, has an upper bound in terms of his own future productivity,
if one makes the value judgment that the individual should be re-
sponsible for his own actions and liabilities. On the other hand,
society may be willing, explicitly or implicitly, to expand the
IV-17
-------
effective willingness to pay of an individual beyond that which he,
himself, Would be able to pay (i.e., his future earnings). The
possibility of such revealed preferences, i.e., private and public
trade-offs between the probability of death and other goods and
services, need to be better understood in the context of evaluating
serious health hazards with low probabilities.
2. Production Benefits
The principal specific problem within this category is the
assumption that the marginal costs of water treatment due to man-
made pollutants, and hence the benefits from pollution control,
are equal to'the prorated portion of the average cost of treatment
due to both man-made and natural pollutants. If a substantial
amount of natural pollution must be treated in any event, then the
marginal costs of treating man-made pollution might be less (sub-
stantially or moderately) than average treatment costs. If so,
using average cost data would overstate the benefits from pollu-
tion control. An extreme case is that of boiler feed water, which
is usually demineralized and given tertiary treatment even in the
absence of man-made pollutants. In this case, the benefits to re-
duction of man-made pollutants would be zero, since the marginal
treatment costs are zero.
For municipal water supplies, Barker and Kramer (1973) have
attributed about 39 percent of total pollution to man-made sources.
For damages to household items from water supplies, Tihansky (1973)
has estimated that about 17 percent of dissolved solids and hardness
are due to man's activities. Also, it has been estimated (U.D. De-
partment of Interior, 1974) that about 40 percent of total salinity
in agriculture results from man-made sources.
For agricultural damages due to siltation and salinity, ob-
served prices are often distorted by price supports for farm pro-
ducts. If price supports were to prevent price declines which
might result from expanded farm production (due to siltation and
IV-18
-------
salinity control), then use of the price support level to compute
benefits would result in overestimation of benefits. This is of
particular concern for desert irrigated crops such as cotton grown
in the Southwest, where both salinity and siltation are production
problems, and where prices are highly influenced by public policy.
3. Recreation Benefits
The inescapable conclusion is that the state-of-the-art is
simply inadequate to allow generalizations of existing research
to arrive at national benefits. To cite just one example, Unger
et al. (1974) derived an estimate of 9.4 billion in recreation
benefits from water pollution control. Their estimate involved
extrapolation of national values from two small-area studies.
One of these involved a high man/water ratio, the other involved
a low ratio. Although this was an imaginative approachs the pre-
cision with which it can define true national benefits is neces-
sarily undefined, and hence, suspect. The model is inadequately
specified with respect to water quality dimensions of the two
sites, and the measure of value at either site is not clearly de-
fined.
The above study is symptomatic of the site-situation-specifi-
city of past research on outdoor recreation benefits related to
changes in water quality. Whether or not this is viewed as "good"
or "bad" depends on one's views of the appropriate decision con-
text, that is, whether or not information on aggregate efficiency
of water pollution control is more valuable than information about
the relative efficiency of improving water quality in selected
bodies of water.
If benefits are to be derived at the national level , research
must be sponsored explicitly for that purpose, or one must, wait for
the decentralized efforts of researchers to produce conceptual
frameworks and empirical relationships which can be properly ag-
gregated to the national level. The latter alternative must be
IV-19
-------
viewed as a "long-shot", as the implicit decision context (national
efficiency gains and losses) is not one which attracts decentralized
research effort when the state-of-the-art is not well developed.
The incentive system of most researchers stimulates performance of
relatively "safe" projects, such as introducing methodological in-
novations in analysis of situation-site-specific cases.
Especially relevant for purposes of aggregation is the need
for the analyst to consider the entire set of recreational options
as perceived by the consumer, and then to analyze how water pollu-
tion at one or more sites affects consumption behavior at all sites.
Analysis at present (even for small studies) is usually quite par-
tial; no attempt at "netting out" is made. This approach, inherent
in small-scale analysis, must be explicitly abandoned in any re-
search attempts to move toward aggregate estimates.
Whether viewed from a perspective of assessing national or
local water quality benefits, the principal limitation is the in-
adequate documentation of a variety of damage functions which re-
late changes in water quality parameters to changes in consumer be-
havior. Moreover, in most of the damage functions which have been
documented, a key water quality parameter is assumed by the analyst,
and its relationship to consumer behavior is derived (e.g., Stevens,
1966). Additional attention needs to be paid to perception of water
quality by the recreationist, which may be considerably at odds with
the scientist's perception.
Most studies consider only the recreational value of a body
of water in its current state, where one or more water quality para-
meters are not up to standard, and its recreational value if the
quality standards were to be met. Such two-point analyses must pre-
sume a linearity of value between points, which is probably not
valid. For example, bringing the DO levels in a certain river up
to 3 ppm might have positive net benefits, whereas at 5 ppm, the
incremental costs might far outweigh the benefits. Unless a range
IV-20
-------
of water quality levels is considered, the existing standard is
assumed as a parameter, when in fact it might more usefully be
considered as a variable.
One area of needed research is the estimation of the income
elasticities associated with various water-based activities. Not
all leisure or water-based activities have positive income elasti-
cities. One might hypothesize that the income elasticity for swim-
ming in rivers which run through major industrial centers is quite
low, and perhaps even negative. In fact, it might further be hypo-
thesized that what appears to be a substantial cost of pollution
(i.e., the abandonment of swimming in industrial rivers) is a phe-
nomenon which largely would have occurred in any event through in-
come growth and the substitution of other leisure activities, parti
cularly those which involve travel to less congested and more at-
tractive sites.
If this hypothesis is valid, and it appears tractable to test-
ing, then the implications are substantial. Rather than reclaim-
ing these rivers on efficiency grounds, as assumed in national
benefit-cost computations, the reclamation could be justified on
distributional grounds (as compensation to those whose incomes do
not allow them to recreate elsewhere), and/or to preserve the op-
tion to go back to older forms of recreation at some time in the
future, is worth the cost, and/or on the basis of "national pride".
The benefits of pollution control can be regarded as the
extra travel costs imposed on consumers by water quality con-
straints which are exogenous to their control. To infer that
these benefits from pollution control are the same as what appear
to be the damages of pollution may seem valid intuitively, but
the symmetry between the two needs to be examined closely.
Unusual income elasticities may cause pollution control bene-
fits to be less or more than the apparent damages of pollution.
Recreational preferences are not acquired instantly, and hence are
IV-21
-------
subject to private and public influences, some of which are both
strong and pervasive. One such influence is the encouragement
of travel through production of both super-highways and super-
autos. Hence, there may be substantial lags in returning to ur-
ban water-based activities even if all bodies of water were brought
back up to standard. Higher energy prices do, of course, pose a
constraint to travel, as dc the "psychic costs" of driving crowded
freeways to escape the crowded city.
One can find micro cases in which the reduction in travel
cost would lead to negative benefits. For example, consider two
sites of unequal quality, the one farther away being the more de-
sirable. Then if pollution destroys the (more valuable) distant
site, the recreationist turns to the less desirable, closer site,
and saves travel costs. While this does in fact happen (e.g., in
over-use of some high mountain lakes), the model may be appropriate
in a macro sense. That is, in terms of absolute magnitudes, water
pollution* has generally been urban-oriented and people have been
forced to travel more, rather than less, to find comparable sub-
stitutes.
An ambivalence can be reached on incorporation of the costs
of time in recreation benefits in that the opportunity cost of time
spent in travel is more likely to be other leisure activities for-
gone than income generation. To value time at zero, however, might
imply that alternative leisure activities are of zero utility; this
is not a reasonable argument. It would seem that from an ex ante
viewpoint, the expected gains from recreating would at least equal
the expected costs, including opportunity costs of forgoing other
recreational alternatives, some of which are substitutable with
money and, hence, have a positive monetary value.
The net effect of all these considerations is not obvious, and
hence, is deserving of research. The conclusion at this point is
that the terms "pollution damages" and "pollution control benefits"
should not be equated lightly.
IV-22
-------
A dynamic, simultaneous setting is the reality with which
analysts must deal. In this setting, the economy and its system
of distribution provides a driving force in re-shaping consumer
preference, both through allowing greater discretionary income
and in motivating consumers on how and where to spend it. More-
over, economic growth, in part through uncompensated negative ex-
ternalities, imposes constraints on the range of goods and ser-
vices which the recreationist can purchase (e.g., "no swimming"
signs).
Two conceptual models, presented in Attachments A and B, may
be useful in making aggregated assessments of recreation benefits
from water pollution control. Model 1 demonstrates the conceptual
links between quality, quantity, and price of the recreational ex-
perience. Model 2 demonstrates the usefulness of a "reduction in
travel cost" construct and specifies data requirements for its im-
plementation.
The typical recreationist finds that water pollution reduces
the quality of his recreation experience, causes him to recreate
less often, and causes him to incur higher travel costs to enjoy
substitute sites. The objective of Model 1 is to demonstrate
that these three components (quality, quantity, and price) are
conceptually linked, and that addition of the three as separate
components is not defensible.
Model 2 (Attachment B) moves toward establishing theory
which may be useful in specifying data requirements for assess-
ing national benefits. As a corollary, it may also be useful in
interpreting whether or not existing empirical studies and data
are of value in attempting to generate estimates of national bene-
fits. The implications of Model 2 are that decreases in travel
costs due to pollution abatement represent potential benefits from
abatement, existing data on total costs of recreation should be
explored as a proxy for travel cost data, and the empirical rela-
IV-23
-------
tionship between pollution levels and travel costs must be estab-
lished before the model can be implemented empirically.
4. Aesthetic and Ecological Benefits
We have seen that the methodology for simulation of changes
in market values for active recreational activities, as a function
of changes in water quality, is still being developed and as yet,
has not been adequately generalized to a national aggregate. The
methodology for monetizing the less active (e.g., viewing) or more
latent (e.g., future consumption) uses is even less developed. At
present, there is little conceptual or empirical support for any
available estimates of national aesthetic benefits from water pol-
lution control.
A specific problem which links the categories of outdoor re-
creation and aesthetics is whether to believe what people say, or
to believe, instead, what they do. Economists have long had a pre-
ference for the latter, fearing the lack of appropriate decision
context in the former method. In short, people may overstate their
willingness to pay if they think it will help secure a good, or
they may understate their willingness to pay if they fear that
their pocketbooks will somehow be affected. Although economists
have long recognized these tendencies, there is relatively little
in the literature in the way of constructive analysis on the avoid-
ance of under or overstatement. The work by Bohm (1971) and Tide-
man (1972) are exceptions, but the reliance on assumption is still
substantial there.
The constraint imposed by inadequate methodology is especially
apparent when one considers the dilemma that exists where direct
questioning appears to be the only available methodology. In the
case of option demands on the part of the current non-users, the
questioning must be with respect to desire for future consumption.
Hence, there is no way to check results of direct questioning
IV-24
-------
against observed behavior. In hypothetical cases of improved
water quality, the general question is "What would your recrea-
tion preference set be if certain water quality parameters were
varied?". In addition to problems that the recreationist might
have in visualizing the improvement, there is again the absence
of empirical cross-checks.
Despite this apparent dilemma, economists have had little
propensity to seek help from other disciplines. Even in cases
where the wording and setting of particular questions are cru-
cial to the results, economists have usually preferred to "go
it alone" rather than seek inputs from psychologists, social
psychologists, or communication theorists. On the other hand,
the non-economic literature on "environmental perception" is
flourishing, but usually with no handles by which to impute
monetary values.
Analysis of cross-section and time-series data is much pre-
ferred to direct questioning, if the data are available. Cross-
sectional and time-series analysis can be designed to reveal con-
sumer behavior in terms of what people actually do in response to
changes in water quality. The basic problem is the lack of data
with which meaningful analyses can be made.
Although the theoretical literature is far ahead of the em-
pirical work, the theory itself is plagued with ambiguities which
only time and intellectual effort can unravel. Its evolution dates
back only a few years and the implication for pollution control
benefits is that something in excess of consumers' surplus of cur-
rent users is ususally appropriate and legitimate. The extent of
this "something" is the crucial issue.
In spite of this evolutionary condition, several vital dimen-
sions of the problem are now being addressed in the literature:
IV-25
-------
» The role of risk-preference versus risk-aversion
• Policy outcomes as public goods (and the concomir
tant implications for direct questioning)
• Uncertainty as to future supply and/or demand
« Relevance of contingent claims markets as analogs
• Mixed-good (private and social) conception of en-
vironmental goods.
Current empirical work is limited in its applicability for
generalization, in that it focuses on high quality, unique envi-
ronmental services, rather than medjurn or low quality services.
This is not to dispute the future methodological value of such
work, only to point out that the ratio of option value to (cur-
rent) consumers' surplus may be quite different for Hells Canyon
than for more mundane activities on the Delaware River.
One would hypothesize that the ratio of option value to con-
sumers' surplus would be higher for the more unique sites, and
that aggregation based on current empirical work would overstate
total benefits. The rationale behind this hypothesis lies in the
differential income elasticities for high and low-quality environ-
ment services. One would suspect the demand for high-quality ser-
vices is more income-elastic than that for services of lower quality.
Thus, the former will become relatively more valuable in the future,
and consumers would be more likely to desire "insurance" of future
supplies, should contingent claims markets exist. Individual in-
vestigators appear most likely to analyze unique, high-quality
sites. In order to test the above hypothesis, one may need to pro-
mote similar research efforts for non-unique, lower-quality sites,
rather than wait for decentralized researchers to ask this question.
Economists and administrators are in a rather obvious dilemma
as to the extent to which option value estimates should be added to
current user values. A market failure symptom is evident in the
sense that contingent claims do not exist for environmental goods
IV-26
-------
in the sense that they exist for some other goods. On the other
hand, the current ability to estimate option values is admittedly
weak, and may cast doubts on the validity of the more concrete
estimates in other categories.
5. Property Values
The academician's traditional wisdom in benefit-cost analysis
has been to exclude from consideration as efficiency benefits
any increments in property values arising from public investment
for which there is no corresponding increase in the output of goods
or services. In other words, pecuniary externalities or simple in-
come transfers have been netted out of the analysis. On the other
hand, if there is in fact an increase in the output of goods or
services, a more direct measure than property value increments is
usually sought. For example, if crop damages were averted by flood
control, both higher farm incomes and higher land values, due to
higher farm incomes, would result. In this case, one would prefer
to use the increase in net farm incomes as the measure of benefit.
Nevertheless, increments to property values do arise as a re-
sult of water quality improvement. This has been substantively
documented by Dornbusch and Barrager (1973) and David and Lord
(1969), although both the capability for generalization and the
adequacy of their model specification must be questioned. David
and Lord's measurement of water quality was much less precise than
desired for purposes of extrapolation, and the Dornbusch and Barrager
effort failed to measure water quality explicitly. These are in-
adequacies that can be rectified with further effort; the real ques-
tion is the meaning, in the context of national efficiency, of
even more adequately derived estimates. That is, what do increments
in property values really reflect? The recent conclusion by Lind
(1973) is that, in a general equilibrium model, most increments (or
decreases) in land values can be ignored for benefit computation.
Property value increments may reflect an actualized option de-
mand which should be included as an efficiency benefit. Pursuit of
IV-27
-------
the relationships between water quality and property values should
be viewed seriously from a strategic point of view, considering
that existing behavioral damage functions for recreation and aes-
thetics are currently not capable of generalization to national
levels.
The methodologies for estimating both recreation demands and
option demands often rely on direct elicitation of responses,
whereas property markets tend to reflect what people do, rather
than what they say they will do. Since public access to water is
increasingly difficult, property values may increasingly reflect
an actualized option demand wherein buyers seek to assure, for
private use, a continued supply of particular environmental ser-
vices. To the extent that these option demands are satisfied, a
service is produced by pollution control.
If sufficient relationships between water quality and property
values were to become available, these might be used to derive gen-
eralized benefit estimates in terms of increments to property values
Caution would be needed to avoid "double-counting". On the other
hand, those to whom property value increments would accrue are prob-
ably a small fraction of total active users.
More refined lines of questioning could be developed to explore
option demand, including the degree to which people seek to secure
greater certainty over resource services through property acquisi-
tion. To the extent that property acquisition does act as a hedge
against uncertain future supplies, there is a "corresponding in-
crease in the output of goods or services," and the criteria for
benefit computation are not violated.
IV-28
-------
E. CONCLUSIONS AND RECOMMENDATIONS
This section lists the conclusions drawn from the foregone discus-
sion and recommendations for future work.
The analytical base of welfare economics and U.S. EPA's data col-
lection capability both provide hope for better benefit estimates. Al-
though numerous other problems exist, the analytical foundations of wel-
fare economics and the data collection capability of U.S. EPA both pro-
vide a basis on which to expect that improved benefit estimates can be
obtained. As Tihansky (1975) points out in his detailed literature re-
view, welfare economics provides a significant conceptual base which has
not been rigorously adhered to in much of the empirical work. The theory
for estimation of benefits is far more advanced than the empirical work;
hence, welfare economics, per se, is not a limiting factor.
The estimates of national benefits, by category of benefit, now have
vastly different precisions. Until these categories are made more nearly
comparable, there is little capability for judging to what degree invest-
ment in water pollution control would be efficient. Estimates of produc-
tion losses, including traditional waterborne disease, are reasonably
firm, although they could be improved at a cost. However, estimates of
consumption losses, including recreation, aesthetics, and survival dimen-
sions of health, are much less firm.
The general prospects for improving the precision of estimates also
differ greatly over the various categories. Additional analytical effort
along conventional lines can resolve much of the variation in the produc-
tion estimates. Also, technological change (biased toward production ra-
ther than consumption activities) may retard the rate at which these bene-
fit streams grow over time. Increased precision in estimates of recrea-
tion benefits and option values will come slowly and probably fortuitously
in the absence of a coordinated effort in benefit estimation.
Highest priority shou_ld_be_ directed to generation of behaviora_l_
response functions which are more conipjete and useful than those few
IV-29
-------
which now exist. Principal attention should be directed to generation
of meaningful behavioral response functions with respect to water qua-
lity changes, which improves their capability for aggregation and pro-
vides a more general framework. It seems almost a hopeless strategy for
U.S. EPA to continue retaining even the best consulting firms to period-
ically make efforts toward extrapolating existing studies, when the base
studies themselves are not designed to be extrapolated.
Criteria need to be established to identify and fund high priority
research on behavioral response functions. The ambitous nature of the
current decision context, that of weighing aggregate national benefits
against aggregate national costs, must be reconsidered; suboptimization
in research design should become a pragmatic reality. Decision criteria
then need to be set to identify and fund high-priority research settings
in which to generate behavioral response functions to changes in water
quality. A matrix relating types of pollutants to geographic regions
should be identified, and priorities for future research established.
The criteria which would be needed to set priorities, among different
elements of the matrix, should include consideration of the current le-
vel of pollution control benefits, considering both mean values and dif-
ferential precision of the estimates, the rate at which the benefit
streams appears to be growing or declining, the likely cost (in money
and time lag) of establishing the behavioral relationships.
The set of behavioral response functions should include services
for which "option demands" might be exercised. Within the set of behav-
ioral relationships to be identified are those which relate to willing-
ness to pay to increase the certainty of future provision of selected
consumption goods and services. The latter may include access to or
availability of recreation activities or aesthetic uses where water qua-
lity is a consideration and avoidance of waterborne health hazards (e.g.,
carcinogens and toxic chemicals).
A "reduction in travel cost" model offers potential for assessing
recreation benefits at a level of aggregation higher than that which
now exists. Empirical implementation of such a model will be more
IV-30
-------
difficult than specification of theory. Crucial elements include iden-
tification of the entire set of water-based recreation activities with-
in the preference set of the consumer and specification of subjective
substitutability or complementarity between sites. Also, measurement
of how water quality differences are related to these elements, ability
to measure this relationship on a continuum, rather than for "clean"
water and "polluted" water, and ability to test the equality of degrada-
tion versus restoration of water quality.
A basic commitment should be made by EPA to establish a long-term
"behavioral monitoring system" analogous to that established by the
U.S. Geological Survey for assessing water quality. The data sets for
a carefully selected variety of sites would include measurement of ba-
sic economic and ecological data, objective and perceived water quality
parameters, recreation activity, by type, intensity, and other charac-
teristics, and property values; particularly for residential and recrea-
tional properties. The monitoring system would be longitudinal in that
trends in both water quality and user behavior would be recorded, allow-
ing for time-series analysis.
Selection of a variety of sites would also allow cross-sectional
analysis. Periodic cross-sectional sampling could be done among recrea-
tionists and potential users. This sampling would not need to be done
necessarily on a large-scale basis, but it would need to be done well.
Samples of users could also be followed longitudinally; a substantial
literature now exists on this methodology. Criteria would need to be
established for selection of sites. Consideration should be given to
selection of sites where significant changes (upward and downward) are
occurring or are likely to occur in the water quality parameters and to
selection of sites which differ significantly in socio-economic variables
(so as to allow cross-sectional estimates of income elasticities).
The rationale behind the proposal to establish a monitoring system
are:
• More adequate knowledge of behavioral responses with re-
spect to price, income, and (perceived or objective) water
IV-31
-------
quality changes need to be established before national
benefits can be estimated
Current inferential capacity of analyses based on di-
rect questioning is largely unknown
Changes in water quality are in fact occurring, and
researchers should be taking advantage of this
The conceptual and empirical dimensions of "option de-
mand" are still evolving; the theoretical models ten
years hence may well be more powerful than can be
tested against the data unless steps are taken to ac-
cumulate a large set of data
A monitoring system provides a continuing flow of in-
formation to decision makers and to their constituency;
over time, behavioral relationships can be determined
and made available, thus augmenting this flow of infor-
mation as well as facilitating much more precise esti-
mates 'of national benefits from pollution control.
IV-32
-------
F. REFERENCES
Barker, Bruce, and Paul Kramer, "Water Quality Conditions in Illinois,"
in Statewide Water Resource Development Plan 1972, Illinois Depart-
ment of Transportation, Division of Water Resource Management, 1973.
Bohm, Peter, "An Approach to the Problem of Estimating Demand for Pub-
lic Goods," Swedish Journa1_of Economics 73(1):94-105, March 1971.
David, Elizabeth L., and William B. Lord, Determinants of Property
Value on Artificial Lakes, University of Wisconsin, Department of
Agricultural Economics, May 1969.
Dornbusch, David M., and Stephen M. Barrager, Benefit of Water Pollu-
tion Control on Property Values, U.S. Environmental Protection
Agency, EPA-600/5-73-005, October 1973.
Freeman, A. Myrick III, "A Survey of the Techniques for Measuring the
Benefits of Water Quality Improvement," in Peskin, Henry M., and
Eugene P. Seskin (eds.), Cost Benefit Analysis and Water Pollution
Policy, The Urban Institute, Washington, D.C., 1975.
Haveman, Robert H., and John V. Krutilla, Unemployment, Idle Capacity,
and the Evaluation of Public Expenditures, Johns Hopkins Press, Bal-
timore, 1963.
Haveman, Robert H., and Burton A. Weisbrod, "The Concept of Benefits
in Cost-Benefit Analysis: With Emphasis on Water Pollution Control
Activities," in Peskin, Henry M., and Eugene P. Seskin (eds.), Cost
Benefit Analysis and Water Pollution Policy, The Urban Institute,
Washington, D.C., 1975.
Krutilla, John V., "Conservation Reconsidered," American Economic Review
57_(4): 777-786, September 1967.
Lackner, Jack, Safe Drinking Water Act of 1973: Estimated Benefits and
Costs, U.S. Environmental Protection Agency, 1973.
Lind, Robert C., "Spatial Equilibrium, The Theory of Rents, and the
Measurement of Benefits from Public Programs," Quarterly Journal of
Economics 37(2):188-207, May 1973.
Lindblom, Charles E., "The Science of Muddling Through," Public Adminis-
trative Review, pp. 79-83, Spring 1959.
Liu, Oscar C., Illnesses Associated With Human Enteric Viruses, U.S.
Environmental Protection Agency, Northeast Water Supply Research Lab-
oratory, 1970.
Meyer, Philip A., Recreational and Preservation Values Associated With
the Salmon of the Fraser River, Fisheries and"Marine Service, Van-
couver, B.C., Canada, Information Report Series, No. PAC/N-74-1 , 1974
IV-33
-------
Peskin, Henry M., and Eugene P. Seskin (eds.), Cost Benefit Analysis
and Water Pollution Policy, The Urban Institute, Washington, D.C.,
1975.
Schultz, T. W., "Institutions and the Rising Economic Value of Man,"
American Journal of Agricultural Economics 50:1113-1122, December
1968.
Sokoloski, A., Economic Benefits of an Adequate Supply of Safe Hater,
U.S. Environmental Protection Agency, 1973.
Stevens, Joe B., "Recreation Benefits from Water Pollution Control,"
Water Resources Research 2(2):167-182, 1966.
Tideman, T. Nicolaus, "The Efficient Provision of Public Goods," in
Selman J. Mushkin (ed.) Public Prices for Public Products, The
Urban Institute, Washington, D.C, 1972.
Tihansky, Dennis P., "An Economic Assessment of Marine Water Pollu-
tion Damages," Proceedings of the Third International Conference
on Pollution Control in the Marine Industries, June 5-7, 1973.
Tihansky, Dennis P., "A Survey of Empirical Benefit Studies," in Pes-
kin, Henry M., and Eugene P. Seskin (eds.), Cost Benefit Analysis
and Water Pollution Policy, The Urban Institute, Washington, D.C.,
1975.
Unger, Samuel G., et al., National Estimates of Water Quality Benefits,
Development Planning & Research Associates, Inc., November 1974.
U.S. Department of the Interior, Westwide Study Report on the Critical
Hater Problems Facing the Eleven Western States, 1974.
IV-34
-------
ATTACHMENT A. MODEL 1
Assume that a recreationist has separate demand functions (D-, and 02)
for two non-priced sites which have, in his subjective judgment, some de-
gree of substitutability. Assume also that these two sites constitute his
"demand for outdoor recreation" (D3). The demand function for each site
is a function of both price (travel cost) variables and the quality para-
meter (Qu) of the site itself. Hence:
Q = f(Pr Qur P2)
Qd = f(P, Qu, P.)
Diagrammatical ly:
Price
C=k,
ql+q2
Quantity
Assume further that the relative price structures are such that the price
of Site I (P, ) is low relative to the price of Site 2 (Pp), so that the
quantity, q, , exceeds q^
implicit "before pollution" price is k]p^
weights determined by optimum quantities,
surplus which he derives is equal to CAFG.
IV-35
Viewing the aggregate demand function, D,, the
+ k2P2> where the k.'s are
q^ and q2> The total consumer's
-------
Now assume that water quality at Site 1 deteriorates to the point
that the recreationist no longer has any effective demand for this site.
Such an event obviously reduces the utility of the consumer. When Site 1
disappears from the consumer's utility function, his equilibrium position
changes from point A to point B. There is a reduction in his total quan-
tity taken, an increase in the unit price paid (though not in the total
bill), and a decrease in his utility. His consumer's surplus is now
PpBFG, and the double cross-hatched area (CAFBP^) is his reduction in
consumer's surplus due to pollution.
If the loss of Site 1 to pollution causes his demand for Site 2 to
shift outward and become more inelastic (fewer substitutes), as might be
expected, then the above decrease in willingness to pay has been over-
stated. Conversely, if the pollution of his "favorite" stream causes
him to quit fishing, for example, altogether (due to inter-re!atedness
of sites and complementarity in terms of fixed factors such as tackle,
etc.), then the reduction in willingness to pay is equal to CAFG, less
the salvage value of his equipment.
IV-36
-------
ATTACHMENT B. MODEL 2
Consider that in a "no-pollution" situation, a recreationists's
utility mapping for Site 1 or 2 contains indifference curve UQ which
passes through his current income position, Y . His willingness to
substitute income to obtain additional units of recreation along U
is identical between Sites 1 and 2. That is, the two sites are per-
fect substitutes for each other.
The cost of travel to the two sites, however, differs; Site 1 can
be reached for a fixed outlay of (Y -k-j), which is essentially the cost
of travel, lodging, and/or food required to reach and return from the
t
site. Once the fixed costs of reaching the site are incurred, the vari
able costs (k-,X-,) per unit of time (e.g., on-site lodging and/or admis-
sion fees) become relevant in determining his optimum length of stay.
Site 2 is located farther from his home, thus the fixed costs of
travel (Y -k?) exceed those for Site 1. Assume, for simplicity, that
the variable or on-site costs of the two sites are equal, or that (k
is parallel to (k,X,).
Income
or
Money
m
m
IV-37
Quantity of Days
-------
In that the two sites are defined (initially) as perfect substitutes
and their relative prices are unequal, a corner solution is reached. The
recreationist will reach a higher (than U ) indifference curve if he sacri-
fices (Y -m,) income and consumes q, days at Site 1 (i.e., point A on in-
difference curve IU).
Assume now that water quality at Site 1 deteriorates to the point
where he is no longer willing to sacrifice income to enjoy Site 1. Un-
der the assumption that the demand for each site is a function of both
site prices and its own quality dimension (i.e., IU retains the same cur-
vature as U and U,, but is now relevant only for Site 2), the recrea-
tionist will now turn to Site 2, which can be reached only at higher tra-
vel costs. His optimum level of consumption is now q? days (all taken
at Site 2) with a money outlay of (Y0-mp)i i-e., point B on indifference
curve Up-
Note at this point that the model is probably of general relevance
in that:
• Pollution leaves the consumer "worse-off" than before
• Pollution causes the distance traveled to increase
t Some recreation is still consumed, even though a site is
"lost" to pollution
• The total money outlay for recreation increases as pollu-
tion and per-unit prices increase (i.e., the demand for
outdoor recreation, in general, would seem to be inelas-
tic, as with most broad categories of commodities)
• Variable or on-site costs probably do not differ greatly
between non-polluted and previously non-polluted areas.
The model is simplistic, however, in that the two sites are initially
defined as perfect substitutes, differ only in distance, and constitute
his entire recreational demand. (The limiting nature of his assumption has
important implications, to be discussed below.)
The "costs" of pollution, to the recreationist, can be defined as
the reduction in willingness to pay which is incurred through pollution.
IV-38
-------
Similarly, the benefits to restoration of clean water (or prevention of
damage) can be seen as the same quantity. Each is a mirror image of the
other, ceteris paribus. The Hicksian concept of "compensating varia-
tion" allows us to approximate the monetary value of this averted utility
2
loss or potential utility gain. It asks: "How much would the consumer
have to be compensated to leave him as well off as before?" In this model
utility level U-j could be regained by an outward shift in the price line
from ^Xp) to (k,X,); i.e., an income transfer of the magnitude (k,^),
or the difference in the fixed costs of travel, could leave the recrea-
tionist as well off as he was prior to pollution, given that on-site costs
are the same in both cases.
This model illustrates two things. First, if the recreation sites
are initially perfect substitutes, the relevant measurement of benefit is
isolated in terms of additional travel costs alone. Any measures of bene-
fits from quantity and quality changes are clearly duplicative, since they
are already incorporated in the reasoning and diagram above. The model
says this; if pollution causes a recreationist to incur $10 in additional
travel cost to reach a more distant site, a $10 income transfer to him
would theoretically leave him as well off as before, and hence, is a mea-
surement of values foregone by pollution -- or benefits created by restor-
ation. Second, the model identifies crucial needs for (a) extent to which
pollution has caused substitutions between sites, and (b) data on additional
per unit travel costs due to pollution. While these may or may not be avail-
2
able, the differences in total outlays (ru-ni) might be a suitable proxy.
The critical limiting nature of this ceteris paribus assumption is examined
on pages IV 21-22.
n
The analysis is shown for those who participate in at least some recreation
both with and without pollution. Other recreationists, however, would cease
this type of consumption altogether with the demise of Site 1 (i.e., move to
Y0). In any empirical treatment, the willingness to pay of these (marginal)
consumers would need to be added to that of existing (intramarginal) consumers.
I would note, however, a difference in the ratios / 'o'"'1 | , and/ 'o'.M \ . If,
/Y°'kA
Iv^J
for example, moving from B to A represents a 40 percent reduction in total costs
/ Q'^1 = n A l > tnen this mi9nt involve (say) a 60 percent reduction in travel
= o
IV-39
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
i. REPORT NO.
EPA-600/5-78-014
3. RECIPIENT'S ACCESSIOf*NO.
4. TITLE AND SUBTITLE
5. REPORT DATE
CRITICAL REVIEW OF ESTIMATING BENEFITS OF AIR
AND WATER POLLUTION CONTROL
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
A. Hershaft (ed.); A. M. Freeman III;
T. D. Crocker, J. B. Stevens
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Environ Control, Inc.
11300 Rockville Pike
Rockville, Maryland 20852
10. PROGRAM ELEMENT NO.
1HA094
11. CONTRACT/GRANT NO.
Contract No. 68-01-2821
12. SPONSORING AGENCY NAME AND ADDRESS
Office of Research and Development
U. S. Environmental Protection Agency
Washington, D. C. 20460
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
EPA-ORD
15. SUPPLEMENTARY NOTES
Project Officer Thomas E. Waddell
16. ABSTRACT
This report provides a critical review of the current state-of-the-art and future
prospects of estimating benefits of air and water pollution control. This
report represents three independent critiques by three experts of benefit
assessment methodologies. Specific aspects discussed include the nature
and role of benefits, damage functions, valuation of effects, aggregation of
results, and representation of uncertainties. The conceptual foundations of
estimating pollution control benefits were presented and compared with empirical
studies. It was concluded that while available estimates often do not adequately
reflect the state-of-the-art, estimates of pollution control benefits would
potentially be very useful to decision makers. The conceptual basis provided by
economic theory for benefit estimation is adequate in most respects and far ahead
of the corresponding empirical effort. A number of studies are guilty of failing
to list explicitly critical assumptions or to express adequately uncertainty in
the results while other studies have employed conceptual models that are
inappropriate to the problem at hand or the available data. Damage functions
underlying benefit estimates are frequently based on insufficient data and/or
inadequate characterization of exposure and effects. National benefit estimates
were found to be based on regional studies which are frequently inadequate in
number and/or quality.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS C. COSATI field/Group
Economic Analysis
Economic Surveys
Air Pollution
Water Pollution
Environmental Economics
Environmental Impacts
Benefit/Cost Analysis
Economic Efficiency
Pollution Abatement
Benefits
Pollution Effects
5C
20A
13. DISTRIBUTION STATEMENT
Distribution Unlimited Available through
National Technical Information Service
Springfield, Va. 22151
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
20 SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
VUS GOVERNMENT PRINTING OFFICE 1978 - ':•_-
------- |