MATHTECH
The Technical Research
and Consulting Division of
Mathematica, Inc.
BENEFIT AND NET BENEFIT ANALYSIS OF
ALTERNATIVE NATIONAL AMBIENT ATTt QUALITY
STANDARDS FOR PARTICOLATE MATTER
VOLUME III
-------
BENEFIT AND NET BENEFIT ANALYSIS OF
ALTERNATIVE NATIONAL AMBIENT ATB QUALITI
STANDARDS FOR PAKTICDLATE MATTER
VOLUME III
Prepared for:
Benefits Analysis Program
Economic Analysis Branch
Strategies and Air Standards Division
Office of Air Quality Planning and Standards
U.S. ENVIRONMENTAL PROTECTION AGENCY
Research Triangle Park, North Carolina
March 1983
-------
BENEFIT AND NET BENEFIT ANALYSIS OF ALTERNATIVE
NATIONAL AMBIENT AIR QUALITY STANDARDS FOR
PARTICDLATE MATTER
By:
Ernest H. Manuel, Jr.
Robert L. Horst, Jr.
Kathleen M. Brennan
Jennifer M. Hobart
Carol D. Harvey
Jerome T. Bentley
Marcus C. Duff
Daniel E. Klingler
Judith K. Tapiero
With the Assistance of:
David S. Brookshire
Thomas D. Crocker
Ralph C. d'Arge
A. Myrick Freeman, III
William D. Schulze
James H. Ware
MATHTECH, INC.
P.O. Box 2392
Princeton, New Jersey 08540
EPA Contract Number 68-02-3826
Project Officer:
Allen C. Basala
Economic Analysis Branch
Strategies and Air Standards Division
Office- of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
March 1983
-------
The analysis and conclusions presented in this report are
.those of the authors and should not be interpreted as necessarily
reflecting the official policies of the U.S. Environmental
Protection Agency.
-------
EPA PERSPECTIVE
There has been growing concern with the effectiveness and burden of
regulations imposed by the Federal government. In order to improve the
process by which regulations are developed, Executive Order 12291 was
issued. The order requires that Federal agencies develop and consider, to
the extent permitted by law, Regulatory Impact Analyses (RIA) for the
proposal and promulgation of regulatory actions which are classified as
major. According to the order, a significant component of the RIA is to be
an economic benefit and benefit-cost analysis of the regulatory alternatives
considered. Under the Clean Air Act, the Administrator of EPA may not
consider economic and technological feasibility in setting National Ambient
Air Quality Standards (NAAQS). Although this precludes consideration of
benefit cost analyses in setting NAAQS, it does not necessarily preclude
consideration of benefit analyses for that purpose.
In full support of the Executive Order, the EPA commissioned Mathtech,
Inc. to accomplish an economic benefit and benefit-cost analysis of some of
the alternatives that were thought likely to be considered in the development
of proposed revisions to the NAAQS for particulate matter (PM). The report,
entitled "Benefit and Net Benefit Analysis of Alternative National Ambient
Air Quality Standards for Particulate Matter," documents the results of the
contractor's study. One of the major objectives of the study was to give a
better understanding of the complex technical issues and the resource
requirements associated with complying with the spirit of the Order for the
NAAQS program. In order to achieve this objective, the contractor was
given a wide range of latitude in the use of data, analytic methods, and
underlying assumptions.
It is important to stress that the benefit analysis portion of the
Mathtech study has not had a role to date in the development of proposed
revisions to the NAAQS for particulate matter. Staff recommendations
currently under consideration are based on the scientific and technical
information contained in two EPA documents. They are the "Air Quality
Criteria for Particulate Matter and Sulfur Oxides" and the "Review of the
National Ambient Air Quality Standards for Particulate Matter: Assessment
of Scientific and Technical Information, OAQPS Staff Paper." These documents
have undergone extensive and rigorous review by the public and the Clean
Air Scientific Advisory Committee in accordance with the Agency's established
scientific review policy. Although the Mathtech study reflects the
"state-of-the-art" in particulate matter benefit analysis, the approach and
results have not been subjected to a comparable extensive peer review
process. In addition, some EPA staff have raised questions regarding the
approach taken in the analysis and the significance of the results for
standard setting purposes under the Act. These circumstances do not
necessarily preclude use of the benefit analysis in some manner after
appropriate peer review and further consideration of the questions that
have been raised.
-------
PREFACE
This report was prepared for the U.S. Environmental Protection Agency
by Mathtech, Inc. The report is organized into five volumes containing a
total of 11 sections as follows:
Volume I
Section 1:
Section 2:
The Benefit Analysis
The Net Benefit Analysis
Volume II
Section 3:
Section 4:
Appendix:
Health Effects Studies in the Epidemiology Literature
Health Effects Studies in the Economics Literature
Valuation of Health Improvements
Volume III
Section 5:
Section 6:
Section 7:
Section 8:
Residential Property Value Studies
Hedonic Wage Studies
Economic Benefits of Reduced Soiling
Benefits of National Visibility Standards
Volume IV
Section 9:
Section 10:
Air Quality Data and Standards
Selected Methodological Issues
Volume V
Section 11:
Supplementary Tables
IV
-------
ACKNOWLEDGMENTS
While preparing this report, we had the benefit of advice, comments
and other assistance from many individuals. Allen Basala, the EPA Project
Officer, and James Bain, former Chief of the Economic Analysis Branch
(EAB), were especially helpful. They provided both overall guidance on
project direction as well as technical review and comment on the report.
Others in EAB who assisted us included Thomas Walton, George Duggan, and
John O'Connor, the current Chief of EAB.
Others within EPA/OAQPS who reviewed parts of the report and assisted
in various ways included Henry Thomas, Jeff Cohen, John Bachman, John
Haines, Joseph Padgett, and Bruce Jordan.
Several individuals within EPA/OPA also provided comments or assis-
tance at various stages of the proj.ect. These included Bart Ostro, Alex
Cristofaro, Ralph Luken, Jon Harford, and Paul Stolpman.
Others outside EPA who reviewed parts of the report and provided
comments included V. Kerry Smith, Paul Portney, Lester Lave, Eugene Seskin,
and William Watson. Other Hathtech staff who assisted us in various ways
were Donald Wise, Gary Labovich, and Robert J. Anderson. We also
appreciate the assistance of Al Smith and Ken Brubaker of Argonne National
Laboratory who conducted the parallel analysis of control costs and air
quality impacts.
Naturally, it was not possible to incorporate all comments and
suggestions. Therefore, the individuals listed above do not necessarily
endorse the analyses or conclusions of the report.
The production of a report this length in several draft versions, each
under a tight time constraint, is a job which taxes the patience and sanity
of a secretarial staff. Carol Rossell had this difficult task and managed
ably with the assistance of Deborah Piantoni, Gail Gay, and Sally Webb.
Nadine Vogel and Virginia Wyatt, who share the same burden at EAB, also
assisted us on several occasions.
-------
CONTENTS
VOLUME III
Section Page
5. RESIDENTIAL PROPERTY VALUE STUDIES
Summary of Results 5-1
General Background 5-4
Methodology 5-6
Literature Review 5-16
Limitations of the Hedonic Technique 5-21
Benefit Estimation 5-26
Conclusion 5-44
References 5-47
Appendix 5A: Sources of Data 5-51
6. HEDONIC WAGE STUDIES
Summary of Results 6-1
Introduction 6-3
Hedonic Wage Models: Theoretical Construct 6-6
Empirical Estimates of Hedonic Wage Models 6-17
Benefits 6-30
Concluding Remarks 6-50
References 6-52
VI
-------
CONTENTS (Continued)
Section
Page
6. HEDONIC WAGE STUDIES (Continued)
Appendix 6A: Estimated Results from Hedonic Wage
Models 6~57
7. ECONOMIC BENEFITS OF REDUCED SOILING
Introduction • 7-1
Overview • 7-1
Scope of Analysis 7-2
Summary of Results 7-5
Models of Benefits from Reduced Soiling 7-9
Introduction 7-9
PM Measures and Soiling 7-11
Physical Damage Functions 7-13
Behavioral Models of Reduced Soiling Benefits 7-18
Summary of Models 7-56
Benefits Calculations 7-62
Introduction 7-62
Household Sector 7-63
Manufacturing Sector 7-97
References 7-132
8. BENEFITS OF NATIONAL VISIBILITY STANDARDS
Introduct ion 8-1
Estimating the Benefits of Urban Visibility
Improvements 8-5
Introduct ion 8-5
Property Value Studies 8-7
Direct Willingness-to-Pay Studies 8-10
Biases in Direct Willingness-to-Pay Studies 8-12
Prediction Relationships for Urban Areas 8-17
Conclusions 8-22
vn
-------
CONTENTS (Continued)
Section Page
8. BENEFITS OF NATIONAL VISIBILITY STANDARDS (Continued)
Recreation Benefits 8-22
Review of Studies 8-23
The Existence Value of Protecting Visibility 8-42
Studies of Existence Value 8-44
Tentative Estimates of Existence Values for
Visibility 8-49
Nonaesthetic Benefits of Visibility Improvement 8-53
National Benefit Estimate for Achieving a 13-,
20-. 30-Mile and a Nationwide 20 Percent
Standard 8-56
Visual Range Regions Utilized in the Benefit
Calculation 8-56
Procedures for Data Collection 8-62
National Benefits for a 13-, 20-, 30-Mile and
20 Percent Improvement Standard 8-63
References 8-66
Appendix 8A: How Visibility Changes Were Calculated
for the Property Value Studies 8-73
Vlll
-------
FIGUBES
VOLUME III
Figure No.
5-1. Implicit price schedule and bid functions ............... 5-13
5-2. Marginal implicit price schedule and demand price
functions ............................................... 5-15
5-3. Alternative benefit estimates for a given change
in air quality ...... .................................... 5-32
6-1. Indifference maps and equilibria for two workers ........ 6—8
6-2. Iso-profit lines and equilibria for two firms ^ ........... 6-13
6-3. Illustration -of labor-market equilibrium ................ 6-15
6-4. Implicit marginal price schedule for air quality and
compensated supply functions for two workers ............ 6-16
6-5. Illustration of benefits calculation for non-marginal
changes in TSP levels ................................... 6-35
7-1. Processes leading to economic benefits ... ............... 7-10
7-2 . Example of economic surplus ............................. 7-33
7-3. Demand and supply curves in the WJ analysis ............. 7-37
7-4. Household decision process ..... . ........................ 7-40
8-1. Median summer visual range (miles) and isopleths for
suburban/non-urban areas, 1974-76 8-57
-------
TABLES
VOLUME III
Table No. Page
5—1. Property Value Benefits of Attaining Alternative
Particulate Matter Standards 5-3
5-2. Review of Property Value Studies 5-22
5—3. Studies Considered in Estimating the Benefits of
Particulate Matter Reduct ions 5-27
5-4. Comparison of TSP Elasticities Calculated from
Residential Property Value Studies 5-30
5-5. Alternative Particulate Matter Standards 5-36
5-6. Estimated Benefits for Residential Property Value
Studies - Benefits Occurring Bet-ween 1989 and 1995 -
Scenario: Type B PM10 - 70 AAM/250 24-hr 5-37
5-7. Estimated Benefits for Residential Property Value
Studies - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM 5-38
5-8. Estimated Benefits for Residential Property Value
Studies - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM/250 24-hr 5-39
5-9. Estimated Benefits for Residential Property Value
Studies - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM/150 24-hr 5-40
5-10. Estimated Benefits for Residential Property Value
Studies - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B TSP - 75 AGM/260 24-hr 5-41
5-11. Estimated Benefits for Residential Property Value
Studies - Benefits Occurring Between 1987 and 1995 -
Scenario: Type B TSP - 150 24-hr 5-42
-------
TABLES (Continued)
Table No.
5-12. Estimated Benefits for Residential Property Value
Studies - Benefits Occurring Between 1989 and 1995 -
Scenario: Type A PM10 - 70 AAM/250 24-hr 5-45
5-13. Benefits of Attaining Alternative Particulate Matter
Standards 5-46
6-1. Summary of Estimated Benefits for Hedonic Wage Studies -
Primary Standards 6-2
6-2. Summary of Estimated Benefits for Hedonic Wage Studies -
Secondary Standards 6-2
6-3. Summary of Selected Hedonic Wage Models 6-25
6-4. Estimated Benefits for Marginal Changes in Air Quality
Based on Results Reported by Rosen and Smith 6—34
6-5. Data Sources and Transformations 6-38
6-6. Air Quality Data: Original Models and Current
Benefits Analysis 6-40
6-7. Estimated Benefits for Marginal Changes in Air Quality .. 6-41
6-8. Estimated Benefits for Hedonic Wage Studies - Benefits
Occurring Between 1989 and 1995 - Scenario: Type B
PM10 - 70 AAM/250 24-hr 6-43
6-9. Estimated Benefits for Hedonic Wage Studies - Benefits
Occurring Between 1989 and 1995 - Scenario: Type B
PMlO - 55 AAM 6-44
6-10. Estimated Benefits for Hedonic Wage Studies - Benefits
Occurring Between 1989 and 1995 - Scenario: Type B
PMlO - 55 AAM/250 24-hr 6-45
6-11. Estimated Benefits for Hedonic Wage Studies - Benefits
Occurring Between 1989 and 1995 - Scenario: Type B
PMlO - 55 AAM/150 24-hr 6-46
6-12. Estimated Benefits for Hedonic Wage Studies - Benefits
Occurring Between 1987 and 1995 - Scenario: Type B
TSP - 75 AGM/260 24-hr 6-47
XI
-------
TABLES (Continued)
Table No. Page
6-13. Estimated Benefits for Hedonic Wage Studies - Benefits
Occurring Between 1987 and 1995 - Scenario: Type B
TSP - 150 24-hr 6-48
6-14. Estimated Benefits for Hedonic Wage Studies - Benefits
Occurring Between 1989 and 1995 - Scenario: Type A
PM10 - 70 AAM/250 24-hr 6-51
6-15. Summary of Biases in Benefits Calculations: Hedonic
Wage Models 6-53
6A-1. Hedonic Wage Model Estimated by Rosen: Effects of
Individual—Specific Characteristics 6-58
6A-2. Hedonic Wage Models Estimated by Rosen: Effects of
Site-Specific Characteristics 6-59
6A-3. Hedonic Wage Models Estimated by Smith . 1 6-60
7-1. Summary of Estimated Benefits for Alternative Primary
Standards 7-6
7-2. v Summary of Estimated Benefits for Alternative
Secondary Standards 7-6
7-3. Regression Results from Beloin and Haynie 7-14
7-4. Cleaning Activities in Cummings' Analysis 7-25
7-5. Total Per-Household Soiling Costs by Pollution Zone 7-27
7-6. SMSAs Included in MTH Analysis , 7-43
7-7. Goods Included in the MTH Analysis 7-44
7-8. Demand Equations with TSP 7-46
7-9. Biases in Models of Soiling Studies 7-57
7-10. Air Quality Scenarios 7-63
7-11. Estimated Benefits for Cummings Soiling Study -
Benefits Occurring Between 1989 and 1995 - Scenario:
Type B PM10 - 70 AAM/250 24-hr 7-66
Xll
-------
TABLES (Continued)
Table No.
7-12. Estimated Benefits for Cummings Soiling Study -
Benefits Occurring Between 1989 and 1995 - Scenario:
Type B PM10 - 55 AAM/250 24-hr 7-67
7-13. Estimated Benefits for Cummings Soiling Study -
Benefits Occurring Between 1989 and 1995 - Scenario:
Type B PM10 - 55 AAM/150 24-hr 7-68
7-14. Estimated Benefits for Cummings Soiling Study -
Benefits Occurring Between 1989 and 1995 - Scenario:
Type B PM10 - 55 AAM 7-69
7-15. Estimated Benefits for Cummings Soiling Study -
Benefits Occurring Between 1987 and 1995 - Scenario:
Type B TSP - 75 AAM/260 24-hr 7-70
7-16. Estimated Benefits for Cummings Soiling Study -
Benefits Occurring Between 1987 and 1995 - Scenario:
Type B TSP - 150 24-hr 7-71
7-17. Estimated Benefits for Cummings Soiling Study -
Benefits Occurring Between 1989 and 1995 - Scenario:
Type A PM10 - 70 AAM/250 24-hr 7-73
7-18. Estimated Benefits for Watson and Jaksch Soiling
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 70 AAM/250 24-hr 7-76
7-19. Estimated Benefits for Watson and Jaksch Soiling
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM/250 24-hr 7-77
7-20. Estimated Benefits for Watson and Jaksch Soiling
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM/150 24-hr 7-78
7-21. Estimated Benefits for Watson and Jaksch Soiling
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM 7-79
7-22. Estimated Benefits for Watson and Jaksch Soiling
Study - Benefits Occurring Between 1987 and 1995 -
Scenario: Type B TSP - 75 AAM/260 24-hr 7-80
Xlll
-------
TABLES (Continued)
Table No. Page
7-23. Estimated Benefits for Watson and Jaksch Soiling
Study - Benefits Occurring Between 1987 and 1995 -
Scenario: Type B TSP - 150 24-hr 7-81
7-24. Estimated Benefits for Watson and Jaksch Soiling
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type A PM10 - 70 AAM/250 24-hr 7-83
7-25. Estimated Benefits for Hathtech Household Expenditure
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 70 AAM/250 24-hr 7-86
7-26. Estimated Benefits for Hathtech Household Expenditure
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM/250 24-hr 7-87
7-27. Estimated Benefits for Mathtech Household Expenditure
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM/150 24-hr 7-88
7-28. Estimated Benefits for Mathtech Household Expenditure
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type B PM10 - 55 AAM 7-89
7-29. Estimated Benefits for Mathtech Household Expenditure
Study - Benefits Occurring Between 1987 and 1995 -
Scenario: Type B TSP - 75 AAM/260 24-hr 7-90
7-30. Estimated Benefits for Mathtech Household Expenditure
Study - Benefits Occurring Between 1987 and 1995 -
Scenario: Type B TSP - 150 24-hr 7-91
7-31. Estimated Benefits for Mathtech Household Expenditure
Study - Benefits Occurring Between 1989 and 1995 -
Scenario: Type A PM10 - 70 AAM/250 24-hr 7-93
7-32. Extrapolation Biases in the Household Sector Models 7-95
7-33. Summary of Benefits from Reduced Soiling in the
Household Sector 7-98
7-34. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 344) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
70 AAM/250 24-hr 7-101
xiv
-------
TABUS (Continued)
Table No.
7-35. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 344) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
55 AAM/250 24-hr 7-102
7-36. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 344) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
55 AAM/150 24-hr 7-103
7-37. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 344) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 - 55 AAM . 7-104
7-38. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 344) - Benefits Occurring
Between 1987 and 1995 - Scenario: Type B TSP -
75 AAM/260 24-hr 7-105
7-39. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 344) - Benefits Occurring
Between 1987 and 1995 - Scenario: Type B TSP -
150 24-hr 7-106
7-40. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 354) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
70 AAM/250 24-hr 7-107
7-41. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 354) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
55 AAM/250 24-hr 7-108
7-42. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 354) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
55 AAM/150 24-hr 7-109
7-43. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 354) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 - 55 AAM . 7-110
xv
-------
TABLES (Continued)
Table No.
7-44. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 354) - Benefits Occurring
Between 1987 and 1995 - Scenario: Type B TSP -
75 AAM/260 24-hr 7-111
7-45. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 354) - Benefits Occurring
Between 1987 and 1995 - Scenario: Type B TSP -
150 24-hr 7-112
7-46. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 34) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
70 AAM/250 24-hr 7-113
7-47. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 34) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
55 AAM/250 24-hr 7-114
7-48. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 34) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
55 AAM/150 24-hr 7-115
7-49. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 34) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 - 55 AAM . 7-116
7-50. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 34) - Benefits Occurring
Between 1987 and 1995 - Scenario: Type B TSP -
75 AAM/260 24-hr 7-117
7-51. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 34) - Benefits Occurring
Between 1987 and 1995 - Scenario: Type B TSP -
150 24-hr 7-118
7-52. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 35) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
70 AAM/250 24-hr 7-119
xvi
-------
TABLES (Continued)
Table No.
7-53. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 35) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
55 AAM/250 24-hr 7-120
7-54. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 35) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 -
55 AAM/150 24-hr 7-121
7-55. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 35) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type B PM10 - 55 AAM . 7-122
7-56. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 35) - Benefits Occurring
Between 1987 and 1995 - Scenario: Type B TSP -
75 AAM/260 24-hr 7-123
7-57. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 35) - Benefits Occurring
Between 1987 and 1995 - Scenario: Type B TSP -
150 24-hr 7-124
7-58. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 344) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type A PM10 -
70 AAM/250 24-hr 7-126
7-59. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 354) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type A PM10 -
70 AAM/250 24-hr 7-127
7-60. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 34) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type A PM10 -
70 AAM/250 24-hr 7-128
7-61. Estimated Benefits for Mathtech Manufacturing
Expenditure Study (SIC 35) - Benefits Occurring
Between 1989 and 1995 - Scenario: Type A PM10 -
70 AAM/250 24-hr. 7-129
-------
TABLES (Continued)
Table No. Page
1-62. Summary of Benefits from Reduced Soiling in the
Manufacturing Sector 7-131
8-1. 1980 Annual National Benefits of Alternative
Visibility Standards .................................... 8-5
8-2. Annualized Property Value Changes Per Mile of
Visibility by Study and Area ............................ 8-9
8-3. Annual Willingness to Pay Per Mile of Visibility
By Study and Area ....................................... 8-11
8-4. Low, Medium and High Estimated Residential Benefits
Per Tear in 1980 Millions of Dollars for Visibility
Policies of 13, 20, 30 Miles and a Nationwide 20%
Improvement ............................................. 8-21
8-5. Mean Incremental Willingness to Pay in 1980 Cents
Per Mile ................................................ 8-34
8-6. Visual Range Valuation Elasticities (02)
8-7. Low, Medium and High Estimated Recreational Benefits
Per Tear in 1980 Millions of Dollars for Visibility
Policies of 13, 20, 30 Miles and a Nationwide 20%
Improvement ................................... . ......... 8-43
8-8. National Existence Value Benefits from S.B.W.K. Study ... 8-48
8-9. • Recreation Areas Used in Existence Value Analysis ....... 8-50
8-10. Estimated Existence Value Benefits Per Tear in 1980
Millions of Dollars Per Visibility Policies of 13,
20 and 30 Miles and a Nationwide 20% Improvement ........ 8-54
8-11. Low. Medium and High Visibility Visual Range Values
by Region ............................................... 8-61
8-12. Low, Medium and High Estimated Residential Aesthetic
Benefits Per Tear in 1980 Millions of Dollars for
Visibility Policies of 13, 20, 30 Miles and a
Nationwide 20% Improvement .............................. 8-64
XVlll
-------
TABLES (Continued)
Table No. Page
8-13. Low, Medium and High Estimated Recreational Benefits
Per Year in 1980 Millions of Dollars for Visibility
Policies of 13, 20, 30 Miles and a Nationwide 20%
Improvement 8-65
8-14. Estimated Existence Value Benefits Per Year in 1980
Millions of Dollars Per Visibility Policies of 13,
20, 30 Miles and a Nationwide 20% Improvement 8-67
8-15. Total Existence Value Index 8-68
8-16. Ranges for the Low, Medium and High Total Benefits Per
Year in 1980 Millions of Dollars for Visibility
Policies of 13, 20, 30 Miles and a 20% Nationwide
Improvement 8-69
xix
-------
SECTION 5
RESIDENTIAL PROPERTY VALUE STUDIES
-------
SECTION 5
RESIDENTIAL PROPERTY VALUE STUDIES
SDMMAHY OF RESULTS
One method that has been used to estimate the benefits of air quality
improvements has involved the analysis of residential property value
differentials. The underlying hypothesis in this method is that residen-
tial property values will reflect not only housing quality'but also site-
specific attributes such as location, neighborhood characteristics, avail-
ability of services, and environmental amenities including air quality.
Under this hypothesis, property value differentials attributable to air
quality differences reflect the household's economic valuation of air
quality and therefore can be used to estimate the economic value of
improvements in air quality.
The purpose of this section is to develop estimates of the benefits
resulting from reductions in the ambient level of suspended particulates
based on the results of representative studies employing the property value
technique. Because property value differentials can measure only the
effects of air pollution that are perceived by the household, it should be
mentioned that the types of benefits that are measured in this section are
any perceived health, physical property, aesthetic, or psychic benefits
that can be attributed to residing in an area with relatively clean air.
Consequently, it is likely that the benefits estimated in this section will
be larger than the benefits estimated in Section 7 for household soiling
and materials damage. It is unclear, however, whether the estimates pre-
sented in this section will exceed the health benefits presented in
Sections 3 and 4.
5-1
-------
The benefits in this section were estimated using the same air quality
data and air quality standards used throughout this study — implementation
of four particulate matter with a diameter of 10 um (PM10) and two total
suspended particulate (TSP) standards in counties where air quality data
are available. The resulting estimates under these alternative standards
are summarized in Table 5-1. The benefits reported in this table are
reported in terms of the discounted present value in 1982 of a stream of
benefits occurring between the year of standard attainment and 1995.*
These benefit estimates are stated in 1980 dollars and assume a 10 percent
discount rate. As Table 5-1 indicates, under the most lax PM10 standard of
an annual arithmetic average (AAM) of 70 ug/m and a 24-hour expected value
(EV) of 250 ug/m , the discounted present value of benefits range from $3.4
to $11.4 billion and include a point estimate of $6.9 billion. The most
stringent of the PM10 standards — an AAM of 55 ug/m3 and a 24-hour EV of
150 ug/m results in benefits ranging from $7.6 to $25.4 billion with a
point estimate of $15.3 billion. The current primary standard of 75 ug/m
annual geometric mean (AGM) and 260 ug/m 24-hour maximum value not to be
exceeded more than once a year for TSP are estimated to result in benefits
ranging from $11.2 to $37.3 billion and include a point estimate of $22.4
billion.
The benefits estimated in this section should be considered to be
general approximations of the household benefits resulting from the attain-
ment of alternative particulate matter standards for the following reasons:
• Since property value differentials only reflect the value
of perceived differences in air quality levels, any unper-
ceived benefits resulting from air quality improvements
will not be captured in the estimates reported in this
section. Consequently, the benefits reported in this
section are underestimates of the total benefits of reduc-
tions in the ambient level of particulate matter.
* The year of standard attainment for PM10 standards is 1989. The year of
standard attainment for TSP standards is 1987.
5-2
-------
Table 5-1
PROPERTY VALUE BENEFITS OF ATTAINING ALTERNATIVE
PARTICDLATE MATTER STANDARDS*
(Billions of 1980 Dollars)
Standard**
PM10 - 70 ng/m3
AAH and 250 (ig/m3
24-hour EV
PM10 - 55 fig/m3
AAH
PM10 - 55 (ig/m3
AAM and 250 fig/m3
24-hour EV
PM10 - 55 jig/m3
AAM and 150 (ig/m3
24-hour EV
TSP - 75 jig/m3
AGM and 260 |ig/m3
24-hour MAX
TSP - 150 fig/m3
24-hour MAX
Minimum
$ 3.4
6.0
6.0
7.6
11.2
14.9
Point Estimate
$ 6.9
11.9
12.0
15.3
22.4
29.7
Maximum
$11.4
19.9
20.0
25.4
37.3
49.6
* Benefits are stated in terms of the discounted present value in 1982 and
assume a 10 percent discount rate. Benefits for the PM10 standards are
accumulated from 1989 to 1995, while the benefits for the TSP standards
are accumulated from 1987 to 1995.
** AAM - annual arithmetic mean; AGM = annual geometric mean; EV - expected
value; MAX = maximum value not to be exceeded more than once a year.
The results of studies on specific cities in the early
1960's and 1970's are used to estimate the benefits of
pollution reductions occurring between 1987 and 1995 for
the counties included in this analysis. If the valuation
of air quality improvements has changed significantly since
that time, or if the valuation of air quality improvements
differs between cities and counties, the use of these
5-3
-------
studies' results can only be considered as approximations
of the benefits of attaining alternative particulate matter
standards.
The assumptions of the hedonic technique are assumed to
hold in the property value market. The effect of this
assumption is unknown.
The average of the particulate matter readings within a
county is taken as representative of the level of exposure
of all households within a county. The effect of this
assumption on the estimated benefits is unknown.
The marginal willingness to pay for air quality improve-
ments of households residing in single-family units is
assumed to be representative of all households. The effect
of this assumption is unknown.
The marginal willingness to pay for air quality improve-
ments is assumed to decline at a constant rate as air
quality improves. The effect of this assumption is
unknown.
GENERAL BACKGROUND
The analysis of residential property value differentials has been
widely used for estimating the benefits of reductions in air pollution
levels. This method assumes that the benefits of living in a clean air
environment are capitalized into property values. In other words, what
people are willing to pay for air quality improvements can be measured by
the observed differences in the value of residential properties that are
identical in every respect except air pollution exposure.
Since this method focuses on the decisions made in the housing market,
the household does not need to know the technical relationship between air
pollution and physical damage. The household, however, must be able to
perceive the effect of different levels of air quality, and make decisions
in the housing market based on that perception. Consequently, the types of
benefits that are measured through the property value method are any
perceived health, physical property, aesthetic, or psychic benefits that
are the result of residing in an area with relatively clean air. Because
some of the effects of air pollution are probably not perceived by
5-4
-------
households, one of the disadvantages of property value studies is that they
cannot provide estimates of all the benefits accruing to households that
result from air quality improvements. For example, health effects that are
not perceived by the household will not be captured by residential property
values. Another disadvantage is that residential property value studies
may only provide estimates of benefits that occur at home. For example,
benefits of air quality improvements that occur at recreation areas and the
workplace may not be measured by residential property value differentials.*
One of the major advantages of property value differential analysis is
the ability to capture the value that households place on the aesthetic and
psychic amenities of the place where they reside. Neither health studies
nor physical property dose-response functions measure the aesthetic
benefits of improvements in air quality. In addition, property value
studies can reflect the choice of substitute activities and goods that are
used as a means of offsetting the effects of pollution. For example, if
the members of a household substitute indoor activities for outdoor activi-
ties on certain days because of poor air quality, this reduced flow of
services from the property would be reflected in a lower property value.
It is possible, however, that the purchase of goods to offset the effect of
pollution may result in an enhancement of the property value in a polluted
area. If central air conditioning, for example, is bought by a household
in order to offset the effects of pollution, the value of that house is
higher relative to an identical home without air conditioning that is
exposed to the same level of pollution. This can be correctly reflected in
property value differentials if air conditioning is identified as one of
the attributes of housing.
* It is possible that property values, in addition to reflecting the value
individuals place on amenities at the home, may also reflect the value
placed on amenities at the workplace since once an individual makes a
residential location decision, the choice of other locational amenities,
such as those at the worksite, are limited. See Cropper (1).
5-5
-------
METHODOLOGY
The use of property value differentials as a means of determining the
willingness to pay for air quality improvements has its underpinnings in
the hedonic price technique. This technique was originally developed by L.
M. Court (2). Griliches and Adelman (3), Griliches (4), Ohta and Griliches
(5), Kain and Quigley (6), and others have used the technique to estimate
the value of changes in the quality of consumer goods. Generally stated,
the hedonic technique examines the functional relationship between the
price of a good and its characteristics.* It has been used extensively as
a means of estimating the marginal willingness to pay for environmental
quality [Harrison-Rubinfeld (7); Nelson (8)]. In these studies, housing
values are regressed on a set of housing characteristics which includes a
measure of air quality.
Before explaining the hedonic technique, it is necessary to address
the question of whether predicted changes in property values are accurate
measures of the total benefits of air quality improvements. Using a model
of locational choice, Polinsky and Shavell (9) have shown that predicted
property value changes are accurate measures of these benefits only under
certain rather stringent assumptions. Their explanation proceeds as
follows:
Assume that there is a city inhabited by individuals with identical
utility functions and equal incomes.** People work in the center of the
city and reside in the area surrounding the center city. Air quality (AQ)
at a specific location increases with distance (d) from the center city.
* See Chapter 1 of Griliches (4) for a summary of the hedonic price
technique.
** The model can be generalized to reflect the possibility that there is
more than one utility function and unequal incomes within the city. In
this case, there would be i consumer groups (i = l,n) where each member
of the i group would have identical utility functions and incomes.
This would only serve to complicate the analysis without changing the
results. If incomes are endogenous to the model, however, the following
results will be altered.
5-6
-------
Travel cost (T) to the center city is also an increasing function of d.
Utility in this city is a function of the consumption of household services
(H), a composite good (X), and the level of air quality [AQ(d)]:
U = U(H. X. AQ(d)) (5.1)
The consumer desires to maximize his utility subject to the budget
constraint:
Y = p(d)H + X + T(d) (5.2)
where Y = money income.
p(d) = per unit price of housing services at a location with
distance (d) from the center city.
H - household services.
X = the composite good with price equal to 1.
T(d) = commuting costs to the center city.
By solving the first-order conditions of the utility maximization
problem, Equation (5.1) can be stated in terms of an indirect utility
function — utility as a function of the demand functions for H, X, and AQ:
U = I[p(d), Y-T(d), AQ(d)] (5.3)
Under the assumption of unrestrained and costless mobility throughout
the city, and identical utility functions and income, a common equilibrium
level of utility, U*, will 1
his utility level by moving.
level of utility, U , will be obtained. At U , no individual can increase
U* = I[p(d), Y-T(d). AQ(d)] (5.4)
Implicit in this relation is the equilibrium housing function:
p(d) = P[U*, Y-T(d), AQ(d)]. (5.5)
5-7
-------
It is important to note that the equilibrium price of housing is a
function of U* as well as Y-T(d) and AQ(d).
In Polinsky and Shavell's paper. Equation (5.5) is used to explain the
conditions under which a new property value schedule can be predicted from
a change in air quality. If it is assumed that the city is small and there
is perfect and costless mobility among cities, then U will be the same
across all cities and exogenous to the small city. If air quality improves
within the small city, U will not change and the change in property value
is only dependent on the characteristics of d. A regression equation
specifying the relationship between p(d) and d can be used in this case to
predict the change in property values resulting from a given change in air
quality. If the city is either large, or there is imperfect mobility among
cities, then U will be endogenous. If air quality improves within the
city, U will be affected. In this case, the new property value schedule
cannot be predicted without first using a general equilibrium model to
determine the new level of U resulting from the change in air quality.
In general, therefore, property value equations that estimate the
change in property value for a given change in air quality can be used to
predict the new property value schedule for such changes only if the
following assumptions are met:
• The geographical area under consideration must be small.
• There must be perfect mobility throughout, and into and out
of, the geographical area.
• There must be no changes in input and output prices.
The hedonic technique, however, does not attempt to predict a new
property value schedule resulting from a change in air quality, but rather
estimates the marginal willingness to pay for air quality improvements by
observing the housing market in equilibrium. In this method, the implicit
price of air quality is identified by examining the differentiated prices
within the housing market that result from variations in existing air
quality. Since the housing market is in equilibrium, the implicit price of
5-8
-------
air quality can be shown to be equal to the marginal willingness to pay for
air quality. The hedonic technique is therefore useful in predicting the
benefits of marginal changes in air pollution. Given certain conditions,
the implicit prices estimated by the hedonic technique and other relevant
variables can be used to estimate the inverse demand function for air
quality.* Through the estimation of this demand curve, the benefits of
non-marginal changes in air quality can also be predicted.
The general form of a hedonic equation relates the price of a good to
the characteristics of that good. As applied to the housing market, this
can be expressed as:
R£ = r(Si, 1^, Q£) (5.6)
where R^ = price of the i residential location.
S^ = a vector of structural characteristics of the i
location.
Nj = a vector of neighborhood characteristics of the i
location.
Q. = a vector of environmental characteristics of the i
location with one element in the vector being air quality
Note that housing price in the hedonic equation is a function only of the
characteristics of the house, not of the household.
The assumptions that are necessary in order for the hedonic equation
to estimate the marginal willingness to pay for air quality improvements
are:
* In a recent study, Palmquist (9) has found that without strong
assumptions, it is not possible to identify the demand curve for a
housing characteristic, such as air quality, with data from one city.
Palmquist has suggested that the demand curve can be identified by using
data from a number of cities.
5-9
-------
The housing market must be in equilibrium.
Individuals must be able to perceive the characteristics
and attributes of housing.
A complete range of houses with alternative characteristics
must be available.
The partial derivatives of property value with respect to the housing
characteristics are interpreted as the marginal implicit prices or the
additional amount that must be paid for a house with one more "unit" of a
particular characteristic, ceteris paribus. Since one of the assumptions
of the property value model is that the housing market is in equilibrium,
the marginal implicit price is therefore equal to the marginal willingness
to pay for that characteristic. In terms of the partial derivative of
property value with respect to air quality (dR^/dq^), an estimate of the
equilibrium willingness to pay for marginal air quality improvements is
obtained.
In order to see why the partial derivatives of the housing equation
variables are equal to the equilibrium marginal willingness to pay for
housing characteristics, it is helpful to develop a model of consumer
choice following Rosen (11).*
Assume that there is a consumer whose utility is dependent on the
consumption of a composite good (X) and a vector of housing characteristics
(H) where air quality (h,) is an element in the vector:
U = U(X, hx, ... . hn) (5.7)
The consumer has income (y) which can be expressed as:
* Rosen's paper dealt with both the consumption and production of a good
that could be defined in terms of its attributes and characteristics.
Since the hedonic price technique only reveals the equilibrium outcome of
demand and supply conditions and not the underlying demand and supply
functions, and since the purpose of this paper is to estimate the
willingness to pay for air quality improvements, we will limit our
discussion to the consumer allocative decisions made for housing.
5-10
-------
y = X + p(H) (5.8)
where p(H) = price of housing.
X = the composite good with price equal to 1.
Setting up the Lagrangian, the consumer maximizes U subject to his
budget constraint. The first order conditions can be expressed as:
3U/3X = X and 3U/3hi = XapfHJ/dhj (5.9)
where X = the Lagrangian multiplier.
By combining equations, it is found that in equilibrium the marginal
rate of substitution between each housing characteristic and the composite
good is equal to the partial derivative of the price of housing with
respect to that particular characteristic (i.e., the implicit price of the
characteristic estimated by the housing equation):
(3U/3hi)/OU/3X) = Op(H)/3hi) . (5.10)
If X is thought of as money (the price of a dollar is equal to $1.00),
equilibrium is achieved when the marginal rate of substitution between h-
and money is equal to the marginal implicit price of h... Since the
marginal rate of substitution between h^ and money can also be viewed as
the marginal payment for h. with money, the equilibrium conditions can be
expressed as equating the marginal willingness to pay for h- (with money)
with its marginal implicit price.
The equating of the marginal implicit price and the marginal
willingness to pay can also be explained by viewing equilibrium in terms of
a particular h.. Assume that there is a level of consumer utility, u, that
can be defined by the function:
U(y - 9; ht, ... , hQ) = u (5.11)
5-11
-------
where y - 9 = X.
For a given u, Equation (5.11) can be thought of as an indifference
curve relating the tradeoff between b^ and X. A bid function:
0(14, ... , hn; u,y) (5.12)
can be derived from Equation (5.11) which relates the alternative expendi-
tures a consumer is willing to make for h^ given a certain level of utility
and income. By totally differentiating (5.11), we find that:
aU/aX(dy - d9) + aU/ahj^dhj + ... + aU/ahndhn = du (5.13)
Given the assumption of a fixed level of utility and income, du = 0 and dy
= 0. If dhj, = 0 for k ^ i, Equation (5.13) reduces to:
ae/ahi = ou/ah^/ou/ax) . (5.14)
Viewing a particular h^h^ as air quality, Equation (5.14) shows that the
marginal rate of substitution between air quality and X (money) is equal to
the marginal implicit bid for air quality (39/dh^) at a given level of
utility and income.
Figure 5-1 shows the bid function of consumer j for air quality while
holding everything else constant, d-5^, l^. ... , hn; u , y ). There are
a number of different bid functions reflecting the different levels of
tastes, preferences, and income of consumers. This function shows the
willingness to pay (bid) for air quality in terms of the amount of money
(X) foregone, ceteris paribus. The minimum implicit prices revealed in the
market that must be paid for different levels of air quality while holding
h2 through hfl constant is shown by p(hj^, hj, ... , h*). Equilibrium is
reached when d-^hj, hj, ... , hQ; u*, y*) is tangent to p(h1, hj, ... ,
$ <
hQ); i.e., where the marginal willingness to pay for air quality is equal
to its marginal implicit price.
5-12
-------
9rP
.. hn*, u* , y*)
h. (air quality)
Figure 5-1. Implicit price schedule and bid functions
5-13
-------
Since the hedonic equation can be expressed by p(h^, .•• • hn)'
marginal implicit price schedule for air quality (apOD/dh^) can be easily
obtained by taking the derivative of the hedonic housing equation with
respect to air quality (see Figure 5-2). Assuming that the housing market
is in equilibrium and following the above explanation, this schedule will
also trace out the loci of marginal willingness to pay equilibria for
different levels of air quality by consumers with different bid functions.
Note that unless all consumers have identical bid functions (i.e.,
identical utility functions and incomes), the hedonic technique yields only
the equilibrium marginal willingness to pay of consumer j with bid function
9J. This is only one point on consumer j's demand price function for h^
while holding utility constant (i.e., the inverse compensated demand
function). Consequently, dp(H)/dh« is not the inverse compensated demand
function for air quality and, in most cases, can only be used to approxi-
mate the benefits of marginal improvements in air quality. In order to
accurately estimate the demand curve for air quality and predict the
benefits of non-marginal changes in air quality when consumers do not have
identical utility functions and incomes, additional information and steps
are needed.
In this section, the hedonic studies that have measured the relation-
ship between residential property values and particulate matter will be
reviewed. The ultimate purpose of this review is to determine which of
these studies can be used to estimate some of the economic benefits
resulting from decreases in the ambient level of particulate matter. The
studies used in this section meet the following specific criteria:
• Each study uses a properly specified hedonic model and
attempts to include as many of the characteristics of the
residential property as possible.
• The relationship observed between particulate matter and
residential property value is plausible in terms of the
underlying theoretical construct.
• The results can be used to estimate the benefits of parti-
culate matter reductions.
5-14
-------
h, (air quality)
Figure 5-2. Marginal implicit price schedule and
demand price functions.
5.15
-------
The next subsection will contain a review of the pertinent residential
property value studies. This will be followed by a subsection discussing
the limitations of the hedonic price technique. The next subsection will
contain the methods used to calculate the benefits of reductions in the
ambient level of particulate matter. It will also contain the benefits
under alternative particulate matter standards. Finally, the section will
be ended with conclusions.
LITERATURE
As previously mentioned, all of the studies that will be reviewed in
this section examine the relationship between residential property values
and some measure of particulate matter. The measure of particulate matter
most commonly used in these studies are sulfates (SO^), total suspended
particulates (TSP), and dust fall. Consequently, studies including any of
these pollutants will be examined.
The first study undertaken to measure the relationship between
property values and the level of air quality was done by Ridker and Henning
(12). Using 1960 cross-sectional census tract data from the St. Louis
metropolitan area, the effect of air pollution levels on property values
was estimated using regression analysis. The dependent variable was the
median value (estimated by owner) of owner— occupied single— family housing
units, and the independent variables included those reflecting location
characteristics (e.g., accessibility to highway, travel time to central
business district), property characteristics (e.g., median number of rooms,
houses per mile), neighborhood characteristics (e.g., school quality,
persons per housing unit), median family income, and an air pollution
variable (an index indicating the presence of S02, SOg, HjS, H2S04, and in
some cases dustfall). Different linear specifications were tried and a
significant negative relationship was found between the dependent variable
and the air pollution variable. From these results, they concluded that
property values could be expected to rise at least $83.00 and more probably
5-16
-------
$245, if the measurement of SOj were to drop by 0.25 (ig/100 cm /day.* The
•
elasticity of the pollution variable could not be computed from the
information contained in the study.**
Zerbe (19) estimated a property value equation for Toronto and
Hamilton, Ontario. The annual averages of sulfur dioxide and dustfall were
the two pollution measures used. Both linear and log-linear specifications
were employed; in the log-linear specification, the elasticity of property
values with respect to sulfur dioxide ranged from -0.061 to -0.121 for
Toronto and -0.081 for Hamilton."1"
Crocker (20), in a study of the relationship between home sale price
and the annual arithmetic means for sulfur dioxide and total suspended
particulates in Chicago, found a consistently significant negative
relationship between sale price and particulate matter. The elasticity of
home sale price with respect to particulate matter ranged from -0.25 to
-0.83 and was generally significant. The coefficient of SC^ was inconsis-
tently signed and sometimes insignificant. However, when SOj was entered
separately into the property value equation, it was generally negative and
significant.
Anderson and Crocker (25) estimated the relationship between air
pollution and median property values (estimated by owner) for three cities:
St. Louis, Washington, D.C., and Kansas City. Using separate equations for
* Freeman (13) concluded that Ridker and Henning's results were over-
interpreted and could not be used to predict changes in property values
when air quality changed because the demand curve for air quality had
not been identified. This led to quite a debate in the literature over
the proper interpretation of the derivative of the air quality
variables. [See Anderson and Crocker (14); Freeman (15); Polinsky and
Rubinfeld (16); Small (17); and Harrison and Rubinfeld (18).]
** In this section, the elasticity measures the percentage change in resi-
dential property value that can be expected from a 1 percent change in
the pollution variable. In a log-linear property value equation, the
elasticity is equal to the coefficient of the pollution variable.
+ Information on the studies by Zerbe (19), and Steele (21) is taken from
Freeman (22), Waddell (23), and Appel (24).
5-17
-------
owner-occupied and renter-occupied housing, they found a significant
negative relationship between the annual arithmetic means of air pollution
and property values while- controlling for median family income, percentage
of old units, percent of run-down units, percent of non-white population,
distance to the central city, and median number of rooms. The annual
arithmetic means of sulfur oxides and suspended particulates were the
pollution variables considered in this study. Using a log-linear specifi-
cation, they found that the elasticity of sulfur oxides ranged from -0.07
to -0.12, while the elasticity of total suspended particulates ranged from
-0.06 to -0.17. It is interesting to note that by using plausible interest
rates (e.g., 8 to 12 percent), the results for the owner and renter
equations were roughly consistent. This indicates that the marginal
willingness to pay for air quality improvements estimated for households
living in owner-occupied residences may be similar to the marginal
willingness to pay of households that rent their residences.
Steele (21) did not find a significant relationship between property
values, as measured by mean value per room, and SO* and particulates for
residences in Charlestown, South Carolina. However, the relationship
between property values and pollution was plausibly signed.
Wieand (26) regressed per-acre housing expenditures in St. Louis (a
proxy for land values) on property characteristics, neighborhood charac-
teristics, income, and pollution as measured by annual mean sulfation and
annual mean particulates. Neither pollution variable was significant.
Deyak and Smith (27), using a log-linear specification, found a signi-
ficant relationship between median property values of representative SMSAs
and total suspended particulates. Other variables included in their best
equation were median family income and percent of inferior housing units.
The elasticity of particulate matter was quite consistent and ranged from
-0.083 to -0.088. In a later study on the owner- and renter-occupied
housing market for 85 cities which included measures of local public
services and taxes, Smith-Deyak (28) did not find a significant negative
relationship between air pollution and property value. Subsequent analysis
5-18
-------
by Deyak showed that if cities with relatively low levels of air pollution
*
were excluded from the analysis, there was a significant negative relation-
ship between air pollution and property value.
Using the data from Anderson and Crocker's St. Louis study (25),
Polinsky and Rubinfeld (29) empirically estimated the equilibrium housing
market function developed by Polinsky and Shave11 (9) using a Cobb-Douglas
form of utility function. Log-linear equations were developed for both
owned and rented properties. The annual arithmetic mean of TSP was
negative and significantly different from zero at the 0.05 level for both
equations, while the annual arithmetic mean of the sulfation variable was
negative and significantly different from zero at the 0.10 level for the
homeowner equation. In the owner—occupied property value equations, the
elasticity of TSP was equal to -0.132 and the elasticity of sulfates was
equal to -0.063.
Nelson (8) also found a significant relationship between air pollution
and median census tract owner—occupied property values in Washington, D.C.
Several different specifications were employed with the semi-log and log-
linear forms giving the best results. It was concluded that an increase of
10 ug/m in the average monthly geometric mean from February to July of
total suspended particulates would reduce the mean value of property by
$576 to $693. In the most representative equations, the elasticity of TSP
ranged from -0.078 to -0.116. The estimated marginal willingness to pay
was then used to calculate a bid price function (i.e., an inverse demand
curve for air quality). The price elasticity of demand for TSP in this
equation ranged from -1.2 to -1.4, indicating that the implicit price of
TSP is quite responsive to changes in the level of TSP. The bid price
function estimated by Nelson, however, is only a first attempt at
estimating the demand curve for air quality. In fact, Palmquist (10)
suggests that unless some rather stringent assumptions are met, it is not
possible to identify the demand curve for air quality using the data from
one city. Consequently, the true bid price function is probably more
complicated than the one estimated by Nelson.
5-19
-------
In a study of single-family owned residences in the Los Angeles area,
Brookshire e_t .al.. (30) found a significant negative relationship between
the sale price of homes and air pollution measures as measured by nitrogen
dioxide (N0~) and total suspended particulates. Actual market transactions
2t
for individual homes were used as the unit of observation. Both the linear
and nonlinear specifications employed showed a significant relationship
between home sale price and pollution. The average sale price differential
attributable to a change in the level of pollution from "poor" to "fair"
ranged from $5,793 per home to $6,134 per home. Elasticities could not be
calculated from the information contained in the study. Brookshire .et, al.
also estimated an inverse demand function for air quality, but used NC^ as
the air quality measure.
In a study of Philadelphia property values, Peckham (31) found a
negative relationship between 1969 air pollution levels as measured by
particulates and sulfates, and 1960 owner-occupied property values. In the
log-linear equation, the elasticity of the sulfate variable was -0.096,
while the elasticity of the particulate variable was -0.116.
Spore (32) analyzed the effect of air pollution on property values in
Pittsburgh. Sulfation and dustfall were the two pollution variables
included in the analysis. A significant negative relationship between
property values and air pollution was generally exhibited in the log-linear
equations that were estimated. The elasticity of dustfall in these
equations ranged from -0.092 to -0.149.
The concentration of nitrogen oxides (used as a proxy fox air pollu-
tion) was found to be negatively related to median property values in
Boston by Harrison and Rubinfeld (18). Besides air pollution, the housing
characteristics that were included in the equation were: two structural
variables, eight neighborhood variables, and two accessibility variables.
With all the independent variables at their mean levels, a change in
nitrogen oxides (NO^) of 1 pphm was associated with a change in median
housing values of $1,613. When the variable for NOX was replaced by a
variable measuring particulate matter, particulate matter exhibited a
5-20
-------
negative and significant relationship with property value. Evaluated at
the mean levels of the variables, the elasticity of particulate matter was
-34.24. No explanation was given for the large magnitude of this number.
In a study of the New York metropolitan area, Appel (24) found a
significant negative relationship between the annual geometric mean of
total suspended particulates and mean property values. No other pollution
variables were included in the study. The hedonic equation that performed
best was one in which the TSP variable was entered in exponential form.
This form conforms to .a priori expectations that the marginal damages of
pollution increase as pollution increases. Evaluated at the mean, the
elasticity of TSP was equal to -0.039. Other variables included in the
best equation were the mean number of rooms, the crime rate, the percent of
non-white persons, and minutes of time to the central business district.
It is interesting to note that when the property value equation was
estimated in log—linear form, the elasticity of the TSP variable was equal
to -0.117. This elasticity was not significant, however.
A summary of the studies that have found a significant negative
relationship between residential property values and some measure of parti-
culate matter is given in Table 5-2.
LIMITATIONS OP THE HEDONIC TECHNIQUE
Before preceeding with the calculation of benefits, it is important to
reiterate the limitations of using the hedonic technique and the effect
these limitations have on the ability to predict the benefits of improve-
ments in air quality. The hedonic technique is capable of estimating the
implicit price of the characteristics of a good that the consumer is able
to perceive accurately. Since most characteristics of a good are tangible
and easily perceived, this is not unreasonable. Air quality, per se. is
not a tangible characteristic of housing and it is possible that households
are unable to accurately perceive the effect of air quality on their resi-
dential property. Even if households are cognizant of some of the effects
of air quality, it is doubtful that they will be aware of all of its
5-21
-------
Table 5-2
REVIEW OF PROPERTY VALUE STUDIES
i
ro
Study
Ridker «
Henning
(12)
Zerbe
(19)
Crocker
(20)
Peckham
(31)
Anderson i
Crocker
(14)
Spore
(32)
Pollnsky «
Rublnfeld
(29)
City!
Dependent Variable
St. Louis i HPVa
Toronto and
Haul 1 ton, Canada |
HPV
Chicago | hone value
given by sale price
Philadelphia! HPV
Washington, DC, St.
Louis and Kansas
Cltyi HPV, median
gross rent and
median contract rent
Pittsburgh) HPV
St. Louis | HPV,
median gross rent
and median contract
rent
Pollutants Measured,
Hethod of Measurement
Index of sulfatlon
Annual average and median of
averages for sulfatlon (lead
candle) In both cities.
Annual average and median of
averages for dustfall In
Toronto .
Annual arithmetic mean for
sulfatlon (lead candle) and
participates
One-month average for sulfa-
tlon and arithmetic mean for
partlculates
Annual arithmetic mean for
concentrations of SO
(measured by lead caftdle)
and partlculates
Annual geometric means and
maximum monthly values for
sulfatlon (lead candle) and
dustfall
Same as Anderson i Crocker
(14)
Form
Linear
Linear and
"log-linear
Log-linear
Linear and
log-linear
Log-linear
Log-linear
Log-linear
Elasticities of Pollution
Variables (At Heans)
Not Available11
Log-linear equation:
Sulfatlont
Toronto -0.061 to -0.121
Hamilton -0.081
Equation with both
pollutants!
Partlculates:
-0.2 to -0.5
Sul ration i 0.06°
Log-linear equation:
Sulfatlon: -0.096
Partlculates: -0.116
Owner-occupied equation:
Sulf. Part.
D.C. -0.07 -0.06C
K.C. -0.08 -0.09^
S.L. -0.10 -0.12d
Sulfatlon: 0.03d
Dustfalli -0.12
Owner-occupied equation:
Sulf at Ion i -0.063°
Partlculatesi -0.132
Estimated Benefits;
Base Year
If sulfatlon falls by 0.25 /tg/100 cmZ/day,
MPV Increases by $186.50 to $245.00; 1960
At the mean, a decrease of 1 mg SO^/
100 cm /day would Increase mean property
value by a maximum of $97 In Toronto; 1961
At the mean, a reduction of 10 fig/m In
partlculates and 1 ppb of SO would lead
to a $350 to $600 Increase in the mean
value of property! 1964-1967
At the mean, a decrease in SO of 0.1 mg/
100 cm /day and a 10 /ig/ra /day In
suspended partlculates leads to an
Increase In the mean property value of
$600 to $750| 1960
At the mean, a reduction of 1 mg SO,/
100 cm /day and a 10 pg/ni /day In
suspended partlculates would Increase mean
property value of owner-occupied housing
by $300 to $700 In Washington, DC) 1960
A reduction of 0.005 ppm/day In SO^ and
5 tons/ml /month In dustfall Increases the
value of mean property by $150 to $200;
1970
A 5* reduction In sulfatlon i partlculate
levels In all areas of St. Louis would
lead to a predicted change In aggregate
property values of $55 million; 1960
(continued)
-------
Table 5-2 (Continued)
Ul
i
Study
Deyak &
Smith (27)
Nelson
(8)
Brookshlre
et al.
(30)
Appel
(24)
Harrison 4
Rubinfeld
(18)
City,
Dependent Variable
SHSAj HPV
Washington, OC| MPV
South Coast Air
Basin In California!
sale price of
Individual houses
New York Metropoli-
tan Area; mean
property value of
single family owner-
occupied housing
Boston) MPV
Pollutants Measured,
Method of Measurement
Annual arithmetic average of
total suspended partlculates
Monthly geometric mean of
partlculates and arithmetic
average of means of oxldant
levels
Arithmetic average for
nitrogen dioxide and
partlculates
Geometric mean of suspended
partlculates
Mean concentrations for'
nitrogen oxides and partl-
culates calculated by a
dispersion model
Form
Log-linear
Log-linear
Linear and
semi -log
exponential
Exponential
Exponential
semi -log
Elasticities of Pollution
Variables (At Means)
Partlculates: -0.083
Partlculates:
-0.078 to -0.116
Oxldants:
-0.007 to -0.019
Not Available
Partlculates: -0.039
In separate equations —
Nitrogen oxldei -0.39
Partlculatesi -34.24
Estimated Benefits;
Base Year
Not reported
At the mean, a decrease In partlculates of
10 /ig/m Increases the value of mean
property by $576 to $693. At the mean, a
decrease In oxldants of 0.001 ppm
Increases the value of mean property by
$141 to $152 | 1970
At the mean, a decrease of 1 pphm In
nitrogen dioxide would result In an
Increase In the mean sale price of housing
of $2,010; 1977-1978
The average benefit (weighted by the
number of households exposed to alterna-
tive pollution levels) for a 1 /ig/n
reduction In suspended partlculates Is
$42.13; 1970
Mean value of property would Increase by
$1,613 If the oxldant concentration
decreased by 0.01 ppm; 1970
MPV Is median property value of single-family owner-occupied housing units in a census tract.
The mean values of pollution and property values were not reported; consequently, the elasticities could not be estimated.
Not significantly different from zero at 0.05 level.
Not significantly different from zero at 0.01 level.
-------
effects. Consequently, all of the effects of air quality may not be
capitalized into residential property values. Application of the hedonic
technique in order to estimate the effects of air pollution may therefore
result in an underestimate of the "true benefits" accruing to residential
properties. Although there has been some criticism that the hedonic
technique is invalid for predicting the benefits of air quality improvement
because households are unable to accurately perceive any of the effects of
air pollution, the studies in Table 5-2 appear to support the hypothesis
that households perceive at least some of the effects of air pollution and
these effects are capitalized into property values.
Benefits estimated through hedonic property value equations may only
provide estimates of the perceived benefits that occur at the residential
property. Some of the benefits from air quality improvements that occur
away from home (e.g., at the workplace, recreational areas) may not be
capitalized into residential property values. Since a portion of the
household's time is spent away from home, it is possible that only a
portion of the .total perceived benefits accruing to a household may be
predicted from the hedonic property value equations. This must be kept in
mind when comparing benefits estimated by the hedonic technique to benefits
estimated by other methods.
As mentioned in the last subsection, the assumptions that are
necessary in order to use the hedonic equation to estimate the marginal
willingness to pay for air quality improvements are:
• Individuals must be able to perceive the characteristics
and attributes of housing.
• The housing market must be in equilibrium.
• A complete range of houses with alternative characteristics
must be available.
It is very unlikely that these conditions will hold in the housing
market. In this study, we are mainly concerned with how the violation of
these assumptions will affect the estimated air pollution-property value
5-24
-------
relationship. The violation of the first of these three assumptions has
already been addressed in the discussion on the difficulty of applying the
hedonic technique to a good possessing a characteristic such as air
quality.
In order for the second assumption to be met in the housing market,
households must be just willing to hold the existing stock of housing at
the prevailing prices. Equilibrium will be achieved only if: 1) all
households have complete information on the prices and characteristics of
housing, 2) transactions and moving costs are equal to zero, and 3) housing
prices adjust instantaneously to changes in demand and supply. According
to Freeman (33), divergencies from equilibrium, in most cases, will only
result in random errors in marginal willingness-to-pay estimates. Freeman
mentions, however, that less than instantaneous adjustment in the housing
market to changes in demand or supply may result in biased estimates of the
air quality variable. For example, if equilibrium is disrupted due to an
air quality change, and transactions and moving costs are non-zero, house-
holds will not move unless the benefit is at least as great as the costs
involved in moving. If air quality is consistently changing in one direc-
tion and households consistently lag in their adjustment to that change,
the observed marginal implicit price will diverge from the true marginal
willingness to pay. In this case, the marginal implicit price of air
quality identified in the hedonic property value equation is a biased
estimate of the equilibrium marginal willingness to pay.
Freeman also mentions that future expectations on housing prices may
result in biased estimates of the implicit price of air quality. If house-
holds perceive that an improvement in air quality will take place in the
future and housing prices are affected by that perception, the market has
adjusted to the air quality change before the change actually takes place.
If a hedonic price equation is specified for the housing market in an area
that has already adjusted to a future air quality change, the marginal
willingness to pay for air quality would be underestimated. This possi-
bility can be tested by entering the pollution variables in lagged form.
5-25
-------
It is quite possible that the third assumption may be violated due to
the nature of the housing market. Given the time necessary for the supply
of housing to adjust to changes in demand, it is likely that some house-
holds will not be able to find housing with all of the characteristics that
the household finds desirable. In these households, utility cannot be
maximized. Whether this is a problem that will seriously affect the
estimates of the marginal implicit price of air quality has not been inves-.
tigated at the present time. It is doubtful, however, that the existence
of an incomplete range of homes in the study area will make the estimated
relationship between air pollution and property values totally unreliable.
Segmentation in the housing market can also affect the estimates of
the marginal implicit prices of housing attributes. Housing market segmen-
tation exists when the purchasers of housing participate in distinctly
separate housing submarkets even though the purchasers are technically
participating in the same housing market. The submarkets may exist because
of racial discrimination, cultural differences, or geographic immobility.
Where housing market segmentation exists, the structure of the prices of
housing in each submarket will be different. The specification of a
hedonic price function for one housing market when submarkets exist will
result in incorrect estimates of the marginal implicit prices of housing
attributes. In order for the implicit prices of housing characteristics to
be correctly estimated, separate equations for each submarket must be used.
Nelson (34) did not find that stratified samples for the Washington, D.C.
area affected the hedonic price functions. Further research investigating
this problem is needed before anything more conclusive can be said
regarding the effect of market segmentation on the hedonic price functions.
BENEFIT ESTIMATION
The studies reviewed in the Literature Review subsection provide
estimates of the willingness to pay for air quality improvements. All of
these studies have included some form of particulate matter in estimating
hedonic property value equations. The standards under consideration in
this report are stated in terms of particulate matter with a diameter of 10
5-26
-------
|im (PH10) and total suspended particnlates (TSP). Since it is assumed that
PM10 is equal to approximately 0.55 of TSP (see Section 9), only the
studies that specifically include TSP will be considered in estimating the
benefits accruing to households as a result of these standards.
Because of the relatively large number of property value studies that
have included TSP as an explanatory variable, it was possible to limit
further the selected studies to those that have included at least one other
pollutant besides TSP. Limiting the selected studies in this way minimizes
the possibility that particulate matter is proxying for the effects of the
general air pollution phenomenon. Table 5-3 lists the studies that will be
used for estimating benefits.
For the purposes of this analysis, it is assumed that the results of
these studies are representative of the relationship between particulate
matter and property values in the counties examined in this study, and that
these results can be used to estimate the benefits of achieving alternative
particulate matter standards. It should be noted that there are a'number
of reasons why a strict comparison of the results of these studies is not
possible. These reasons can be explained by referring to Table 5-2.
Although the majority of these studies use the median property value of a
census tract as the dependent variable in their equations, Crocker (20)
uses the sale price of individual homes. Most studies have concentrated on
Table 5-3
STUDIES CONSIDERED IN ESTIMATING THE BENEFITS OF
PARTICULATE MATTER REDUCTIONS
Crocker (20)
Peckham (31)
Anderson and Crocker (14)
Polinsky and Rubinfeld (29)
Nelson (8)
5-27
-------
only the owner-occupied housing market, while some studies have also
estimated equations for the rental market using median gross rent and
median contract rent as dependent variables.
The results of the studies in Table 5-3 are based on data collected in
different years (e.g., 1960 and 1970 Census data). This is not likely to
be a problem in comparing the studies because the demand for air quality
probably has not changed significantly over the years in which these
studies were done. However, the benefits estimated by the studies using
data from different years are obviously not comparable because of the
differences in property values due to increases in the price level. For
this reason, the comparison of the results of the property value studies
will be based on the estimated elasticities of the TSP variables.
As can be seen in Table 5-2, all of the studies examining the
relationship between TSP and residential property values have used a non-
linear functional form to estimate this relationship. The choice of a
nonlinear specification in these studies can be viewed as being twofold.
The hedonic equation need not be linear if costless repackaging of the
characteristics of the good is impossible. In the housing market, it is
unlikely that costless repackaging of the characteristics of housing will
be common. For example, two rooms with four sides are not equal to one
room with eight sides. In addition, the hedonic equation may be nonlinear
in the air quality variable depending on the assumptions regarding the
marginal implicit price of air quality. If the hedonic equation is linear
in the air quality variable, its marginal implicit price is constant over
the entire range of air quality. Since there is no variation in the
implicit price of air quality, it is not possible to identify the demand
for air quality. In those hedonic studies where both linear and nonlinear
specifications have been tried in order to measure the implicit price of
air quality within a specific geographic area (e.g., SMSA), the nonlinear
specifications have given more satisfactory results.
All of the studies listed in Table 5-3 have used a log-linear
functional form to specify a nonlinear relationship between air pollution
5-28
-------
and property values. A significant negative relationship between air
pollution and property values has been found in all of these studies.
Unlike the marginal implicit price of air pollution in a linear specifica-
tion, the marginal implicit price in these specifications varies depending
on the ratio of the property value to the pollution level.* Of more
interest, however, is the rate of change of the marginal implicit price
schedule. It is positive for the log-linear specification, implying that
the marginal implicit price (a negative) is an increasing function of the
level of pollution. Since pollution is something to be avoided (i.e., a
disamenity), this means that the marginal willingness to pay to avoid
pollution becomes less negative as pollution increases; in other words, the
marginal willingness to pay to avoid pollution is lower as the level of
pollution rises. Intuitively, one would expect that the higher the level
of pollution, the greater the marginal willingness to pay for an improve-
ment in air quality.
Since the hedonic technique examines the equilibrium relationship
between property values and air pollution, it is not clear whether the
positive slope of the marginal implicit price curve results from the fact
that people living in relatively clean air environments may tend to have
larger incomes, and consequently a higher equilibrium marginal willingness
to pay for additional air quality improvements, than poorer people who may
tend to live in relatively dirty air environments. Not enough empirical
research has been done in this area, however, to justify the hypothesis
that the marginal implicit price curve is negatively sloped.
As mentioned in the Methodology subsection, the hedonic price
technique yields only the equilibrium willingness to pay for marginal
improvements in air quality. The changes in particulate matter considered
in this analysis, however, will involve non-marginal changes. In order to
accurately estimate the benefits of these changes in air quality, the
demand price function of consumers for air quality must be known. This
* The first derivative of the log-linear specification, (Property Value)
a(Pollution)b, is b(Property Value/Pollution).
5-29
-------
function can be calculated using the implicit prices estimated by the
hedonic technique, and information on air quality levels, consumer income
and characteristics. Nelson (8) is the only researcher to specifically
estimate a demand price function for TSP. However, Palmquist (10) has
shown that without some rather limiting assumptions, the inverse demand
function for TSP cannot be identified based on data from one city. Conse-
quently, the benefits estimated in this section will be based solely on the
information yielded by the first stage of the hedonic price technique.
As previously mentioned, the benefits of achieving alternative parti-
culate matter standards will be approximated using the results of the
studies listed in Table 5-3. Table 5-4 provides a comparison of the
elasticities evaluated at the mean TSP level projected under baseline
conditions (i.e., without implementation of particulate matter standards)
and the mean 1980 residential property value for those counties that will
be used to calculate the benefits of particulate matter reductions. As can
be seen in the table, the elasticities range from -0.048 in Nelson's study
to 0.5 in Crocker's analysis. As just mentioned, Nelson used measures of
the monthly mean of particulates from February to July in order to estimate
Table 5-4
COMPARISON OF TSP ELASTICITIES CALCULATED FROM
RESIDENTIAL PROPERTY VALUE STUDIES
Study
Range
Crocker (20)
Peckham (31)
Anderson and Crocker (14)
Polinsky and Rubinfeld (29)
Nelson (8)
-0.2 to -0.5
-0.116
-0.06 to -0.12
-0.132
-0.048 to -0.116
5-30
-------
the relationship between residential property values and TSP. Since this
averaging period may not represent the annual average of TSP, Nelson's
estimates will not be used in calculating the benefits of particulate
matter reductions. Except for Crocker, none of the remaining studies being
considered have elasticities in excess of -0.20. For the purposes of this
study, -0.20 will be used 'as the maximum elasticity. The minimum
elasticity that will be used in the calculation of benefits is based on the
Anderson and Crocker study, and is equal to -0.06. Since three studies
report elasticities in the range of -0.116 to -0.132, -0.12 will be used as
the point elasticity in calculating benefits.
As previously mentioned, the range of elasticities that will be used
to calculate the benefits of particulate matter reductions is based on
studies that estimate a log-linear property value equation. The marginal
implicit price curve for air quality for the log-linear specification is
shown in Figure 5-3. MPV'(P) is a plot of the derivative of the housing
equation with respect to air pollution (i.e., TSP) and also traces out the
equilibrium willingness to pay to avoid air pollution. D(P) is the true
demand curve for air pollution. Since air pollution is a "bad", a large
negative price implies that the household is willing to pay large amounts
to avoid air pollution. Note that the marginal implicit price curve
resulting from the log-linear specification implies that the marginal
implicit price of air pollution increases (becomes less negative) as pollu-
tion increases.
The benefits of an improvement in air quality can be estimated by the
area under the demand curve over the range of improvement. For the
improvement in air quality from PQ to P^ that is shown in Figure 5-3, the
benefits are estimated by the area APQP^C. The demand curve for air
quality has not been estimated in these studies, however, and it is neces-
sary to rely on information contained in the hedonic price equations to
approximate the benefits of an improvement in air quality. The benefits of
an improvement in air quality from PQ to P^ can be approximated by the area
under the marginal implicit price curve, APgP^B. It must be kept in mind
that this approximation implies that all households' marginal willingness-
5-31
-------
Pollution
MPV'(P)
D(P)
-$
Figure 5-3,
Alternative benefit estimates for a
given change in air quality.
5-32
-------
to-pay functions are identical and increasing in pollution abatement.*
Clearly, this is an overestimate of the true benefits of improvement.
Freeman (35) has suggested two alternative ways of approximating the
benefits of non-marginal changes in air quality when the demand curve for
air quality is not known. By assuming that the household's marginal
willingness to pay for air quality is constant over the entire range of air
quality, household benefits can be approximated by the area APgP^E. The
benefits represented by this area can be estimated by:
T, j-• ^. ^Property Value /.„ ,, ^. v /-,,.*
Benefits = * •,-,<.• (APollution) (5.15)
SPollution
For the log-linear specification, this is equivalent to:
_ _.. . Property Value /.„,,,..> fe -,*\
Benefits = b —„ ,, •_, (APollution) (5.16)
Pollution
where b = the estimated coefficient of the pollution variable
in the hedonic equation.
If the true demand curve for air pollution is D(P), this approximation
technique will result in an overestimate if the marginal implicit price
function is increasing in pollution abatement. However, this technique
clearly will result in a closer approximation of true benefits estimated by
the area AP0P^B.
The other alternative is consistent with the .a priori expectation that
the marginal willingness to pay for air pollution abatement' declines as air
quality improves. One point on the household's demand curve for air
* The marginal implicit price function will necessarily be increasing in
pollution abatement only for log-linear specifications. The linear
specification of the property value equation will result in a constant
marginal implicit price. Exponential, semi-log exponential, quadratic,
and the Box-Cox transformation may yield marginal implicit price curves
that are decreasing in pollution abatement.
5-33
-------
quality is known from the hedonic price equation. By assuming that the
•
household's marginal willingness to pay for air pollution abatement
declines linearly from that point to a zero marginal willingness to pay
when air pollution has been completely abated, benefits can be approximated
for a given improvement in air quality. For the reduction in air pollution
from PO to P, shown in Figure 5-3, household benefits can be approximated
by the area ^PQP^D. The benefits represented by this area can be easily
calculated as the difference in triangle OPQA and OPj^D. This area can be
approximated by:
Benefits = \ [(PQA • OPQ) - (P^ • OP1>] (5.17)
For the log-linear specification, this is equivalent to:
(Pollution^2
Benefits = -r Property Value
1 -
(PollutionQ)
2
(5.18)
where PollutionQ = initial pollution level
Pollution^ = pollution level after an air quality change.
Depending on the shape of the actual demand curve for air quality, the
approximation of benefits under the assumption of a linearly declining
marginal willingness to pay curve can result in either an underestimate or
overestimate of true benefits.
Obviously, either of these alternatives will result in closer
estimates of the true benefits of a given air quality improvement than the
benefits estimated by the area under the marginal implicit price curve of
the log-linear hedonic property value specification. The linearly
declining marginal willingness-to-pay alternative, however, is consistent
with the .a priori assumption that the household's marginal willingness to
pay for air pollution abatement declines as air quality improves. Since it
seems reasonable to expect that the marginal willingness to pay for air
quality improvements will be decreasing, the benefits of alternative
5-34
-------
particulate matter standards will be calculated using the linear to the
origin technique.
Household benefits in a particular county will be calculated for the
single-family owner-occupied household with median property value that is
exposed to the average level of pollution within the county. For the
purposes of this analysis, the benefits accruing to this household will be
taken as representative of the benefits accruing to households residing in
rental and multiple-dwelling units. This may lead to an overestimate or
underestimate of benefits if the willingness to pay for air quality
improvements tends to be different for the households residing in these
types of structures.
All of property value studies being considered in this section have
estimated the particulate matter exposure of census tracts within a city
through the interpolation or dispersion modeling of city data. Since these
measures of exposure are more likely to represent the average of the
ambient level of particulate matter monitored throughout the city than the
worst incidence of pollution throughout the city, county benefits will be
calculated using the average of the annual arithmetic means of particulate
matter from all of the monitors within a county.
A list of the sources of data used in the calculation of benefits is
given in Appendix 5A.
The alternative standards being considered in this analysis are listed
in Table 5-5. As the table shows, the standards are stated in terms of
both the annual and 24-hour averages. For each county, the averaging time
that is most stringent is considered to controlling averaging time. Since
the studies being used in the calculation of benefits are stated in terms
of annual averages, the 24-hour average is converted to an equivalent
annual average if the 24—hour average is the controlling averaging time.
5-35
-------
Table 5-5
ALTERNATIVE PARTICULATE MATTER STANDARDS
(in jig/m )
Standard
1
2
3
4
5
6
Measure of
Particulate
Matter
PM10
PM10
PM10
PM10
TSP
TSP
Annual
Mean*
70
55
55
55
75
—
24-Hour
Expected
Value
250
—
250
150
260**
150**
Attainment
Date
1989
1989
1989
1989
1987
1987
* PM10 standards are ^stated in terms of the annual arithmetic mean while
TSP standards are stated in terms of the annual geometric mean.
** Maximum value not to be exceeded more than once a year.
The benefits of achieving alternative particulate matter standards are
listed in Tables 5-6 through 5-11. These benefits represent the benefits
that would be achieved when all counties included in the analysis are in
compliance with the standard for all years under consideration.*
Table 5-6 reports the discounted present value of the benefits
resulting from a PM10 standard of 70 (ig/m3 annual arithmetic mean (AAM) and
a 250 (ig/m3 24-hour expected value (EV). Under this standard, benefits are
estimated to be in the range of $3.4 to $11.4 billion and include a point
estimate of $6.9 billion. The South Pacific region is the region with the
* In the language of Section 9, these benefits represent "Type B" scenario
benefits.
5-36
-------
Table 5-6
ESTIMATED BENEFITS FOR: RESIDENTIAL PROPERTY VALUE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
0.0
6.5
113.3
135.4
1025.2
337.4
35.3
121.4
1481.7
171.1
3427.2
Point
Estimate
0.0
13.0
226.5
270.8
2050.3
674.8
70.5
242.9
2963.3
342.2
Maximum
0.0
21.6
377.5
451.3
3417.2
1124.6
117.5
404.8
4938.9
570.4
6854.3 11423.9
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
1371.8
2743.6
4572.7
5-37
-------
Table 5-7
ESTIMATED BENEFITS FOR: RESIDENTIAL PROPERTY VALUE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
59.4
69.2
205.6
303.4
1392.2
525.0
103.5
253.8
2809.9
236.3
5958.3
118.8
138.5
411.2
606.9
2784.4
1050.0
207.0
507.6
5619.8
472.6
197.9
230.8
685.3
1011.4
4640.6
1750.0
345.0
846.1
9366.3
787.6
11916.6 19861.1
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
2385.0
4770.0
7949.9
5-38
-------
Table 5-8
ESTIMATED BENEFITS FOR: RESIDENTIAL PROPERTY VALUE STUDIES
.Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
59.4
69.2
205.6
303.4
1397.1
535.4
104.3
253.8
2810.0
259.8
118.8
138.5
- 411.2
606.9
2794.1
1070.8
208.5
507.6
5620.0
519.7
197.9
230.8
685.3
1011.4
4656.9
1784.6
347.5
846.1
9366.6
866.1
Total U.S.
5998.0 11996.0 19993.3
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
2400.9
4801.7
8002.9
5-39
-------
Table 5-9
ESTIMATED BENEFITS FOR: RESIDENTIAL PROPERTY VALUE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.I.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
226.2
146.2
389.3
397.4
1637.5
644.6
152.2
387.9
3135.8
515.0
7632.2
452.4
292.3
778.7
794.9
3274.9
1289.3
304.3
775.9
6271.6
1030.0
754.0
487.2
1297 . 8
1324.8
5458.2
2148.8
507.2
1293.1
10452.7
1716.7
15264.4 25440.6
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
3055.0
6110.0 10183.3
5-40
-------
Table 5-10
ESTIMATED BENEFITS FOR: RESIDENTIAL PROPERTY VALUE STUDIES
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AGM/260 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
I
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
185.1
156.0
569.9
630.7
2616.9
891.4
278.5
493.3
4757.7
599.9
11179.4
Point
Estimate
370.3
311.9
1139.7
1261.3
5233.8
1782.9
557.0
986.6
9515.5
1199.9
Maximum
617.1
519.9
1899.5
2102.2
8723.0
2971.4
928.4
1644.3
15859.1
1999.8
22358.8 37264.7
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
3126.3
6252.6 10421.1
5-41
-------
Table 5-11
ESTIMATED BENEFITS FOR: RESIDENTIAL PROPERTY VALUE STUDIES
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
736,
357,
996,
Total U.S.
839.0
3260.1
1126.4
528.8
699.6
5332.3
992.8
14869.3
Point
Estimate
1473.5
714.0
1993.1
1677.9
6520.3
2252.7
1057.7
1399.3
10664.6
1985.5
Maximum
2455.8
1190.0
3321.8
2796.5
10867.2
3754.5
1762.8
2332.1
17774.3
3309.2
29738.5 49564.2
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
4158.2
8316.4 13860.6
5-42
-------
largest benefits. Approximately 43 percent of the total benefits
associated with, the point estimate of this standard accrue to this region.
Another 30 percent of total benefits accrue to the East North Central
region. With the exception of these two regions, none of the remaining
regions receive benefits in excess of $1 billion. In fact, the New England
region receives no benefits under this standard since all of the region is
estimated to be in compliance with this standard.
A more stringent PM10 standard of 55 (ig/m AAM that is reported in
Table 5-7 results in benefits estimates ranging from $6.0 to $19.9 billion.
This range includes a point estimate of $11.9 billion. Again, the majority
of benefits accrue to the South Pacific (47 percent) and East North Central
(23 percent) regions. With the exception of the South Central and South
Atlantic regions which receive about 9 percent and 5 percent, respectively,
of the total benefits, each of the remaining regions receive less than 5
percent of national benefits.
Table 5-8 reports the benefits estimated to accrue under the PM10
standard of 55 (ig/m3 AAM and 250 (ig/nr 24-hour EV. As shown in the table,
benefits range from $6.0 to $20.0 billion and include $12.0 billion as a
point estimate. The regional ranking of benefits remains relatively
unchanged from the benefits reported in Table 5-7.
3 3
Benefits under the PM10 standard of 55 jig/m AAM and 150 jig/m 24-hour
EV reported in Table 5-9 range from $7.6 to $25.4 billion. The point
estimate for this range is $15.3 billion. Based on the point estimate, the
South Pacific, East North Central, South Central, and North Pacific regions
each receive benefits in excess of $1 billion under this standard.
The benefits under a TSP standard of 75 fig/m3 AGM and 260 ng/m3 24-
hour maximum value not to be exceeded more than once a year are reported in
Table 5-10. With a point estimate of $22.4 billion, these benefits range
from $11.2 to $37.3 billion. Under this standard, point estimates of
benefits in excess of $1 billion accrue to each of the following regions:
5-43
-------
South Pacific, East North Central, South Central, South Atlantic, North
Pacific, and Middle Atlantic.
The last TSP standard considered in this report reflects a 150 jig/m
24-hour maximum value. As indicated by Table 5-11, the pdint estimate of
benefits under this standard is $29.7 billion. The minimum and maximum of
the range surrounding this point estimate are $14.9 and $49.6 billion,
respectively. Under this standard, the point estimate of benefits for each
of the regions except the New York-New Jersey region is greater than $1
billion.
Finally, Table 5-12 shows the benefits that accrue under a 70/250 PM10
primary standard when all counties are not in attainment with the standard
throughout the 1989-1995 time horizon.* This can occur because available
control options are exhausted prior to standard attainment. This table can
be compared to Table 5-6 where all counties were assumed to be in compli-
ance with the same 70/250 PM10 standard. As expected, the benefits
estimates in Table 5-6 exceed those shown in Table 5-12.
CONCLUSION
In this analysis, the discounted present value of the benefits
associated with reducing particulate matter levels to comply with alterna-
tive particulate matter standards has been estimated using the results of
past property value differential studies. Table 5-13 provides a summary of
these estimates in discounted present value terms. These estimates range
from $3.4 to $11.4 billion under the most lenient PH10 standard of 70 |ig/m3
AAM and 250 |ig/m3 to $7.6 to $25.4 billion under the strictest PM10
standard of 55 (ig/m3 AAM and 150 (ig/m3 24-hour EV. The estimated benefits
under the TSP standard of 75 (ig/m3 AGM and 260 (ig/m3 24-hour maximum range
from $11.2 to $37.3 billion while the benefits under the TSP standard of
150 ng/m 24-hour maximum range from $14.9 to $49.6 billion. These
* In the language of Section 9, these are "Type A" benefits.
5-44
-------
Table 5-12
ESTIMATED BENEFITS FOR: RESIDENTIAL PROPERTY VALUE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type A.PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
0.0
6.5
109.4
113.8
738.1
232.8
31.5
116.0
859.3
62.1
2269.5
Point
Estimate
0.0
13.0
218.7
227 ..6
1476.3
465.6
63.1
231.9
1718.5
124.2
4538.9
Maximum
0.0
21.6
364.5
379.3
2460.4
776.0
105,
386.
2864.2
207.1
.1
,5
7564.9
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
908.4
1816.8
3028.1
5-45
-------
Table 5-13
BENEFITS OF ATTAINING ALTERNATIVE PARTICDLATE MATTER STANDARDS*
(Billions of 1980 Dollars)
Standard
PM10 - 70 (ig/m3
AAM and 250 (ig/m3
24-hour EV
PM10 - 55 fig/m3
AAM
PM10 - 55 jig/m3
AAM and 250 (ig/m3
24-hour EV
PM10 - 55 jig/iii3
AAM and 150 ng/m3
24-hour EV
TSP - 75 jig/m3
AGM and 260 fig/m3
24-hour MAX
TSP - 150 (ig/m3
24-hour MAX
Minimum
$ 3.4
6.0
6.0
7.6
11.2
14.9
Point Estimate
$ 6.9
11.9
12.0
15.3
22.4
29.7
Maximum
$11.4
19.9
20.0
25.4
37.3
*
49.6
* Benefits are stated in terms of the discounted present value in 1982 and
assume a 10 percent discount rate. Benefits for the PM10 standards are
accumulated from 1989 to 1995 while the benefits for the TSP standards
are accumulated from 1987 to 1995.
estimates are approximations of the benefits of meeting these standards for
basically four reasons:
1) The benefits reported in this section represent the
benefits associated with perceived differences in air
quality. The unperceived benefits associated with air
quality improvements are not reflected in this section.
Consequently, these benefits are underestimates of the
total benefits of reductions in the ambient level of parti-
culate matter.
5-46
-------
2) The results of studies on spec ific cities in the early
1960's and 1970's are used to estimate the benefits of
pollution reductions occurring from 1987 to 1995 for the
counties included in this analysis. The effect of this
assumption is unknown.
3) The assumptions of the hedonic technique are assumed to
hold in the property value market. The effect of this
assumption is unknown.
4) The benefits of the air quality improvements are approxi-
mated without knowledge of the true demand curve for air
quality. The effect of this assumption is unknown.
5) The marginal willingness to pay for air quality improve-
ments of households residing in single-family units is
assumed to be representative of the marginal willingness to
pay of all households. The effect of this assumption is
unknown.
6) The average of the air pollution readings in a county is
taken as representative of the level of exposure of all
households in the county. The effect of this assumption is
unknown.
These estimates are useful, however, because they provide some idea of the
magnitude of the willingness to pay for improvements in residential air
quality associated with implementing alternative particulate matter
standards.
1. Cropper, M. L. et al. Methods Development for Assessing Air Pollution
Control Benefits: Vol. 4 - Studies on Partial Equilibrium Approaches
to Valuation of Environmental Amenities. Prepared for the U.S.
Environmental Protection Agency. University of California, Riverside,
California, September 1, 1978.
2. Court, L. H. Entrepreneurial and Consumer Demand Theories for
Commodity Spectra. Econometrica, 9(1):135-162, April 1941; 9(2):241-
297, July-October 1941.
3. Griliches. Z. and S. Adelman. On an Index of Quality Change. Journal
of the American Statistical Association, 56(296):535-548. September
1961.
4. Griliches, Z. (ed). Price Indexes and Quality Change. Harvard
University Press, Cambridge, Massachusetts, 1971.
5-47
-------
5. Ohta, M. and Z. Griliches. Makes and Depreciation in the U.S.
Passenger Car Market. Mimeographed, Harvard University, 1972.
6. Karn, J. F. and J. M. Quigley. Measuring the Value of Housing
Quality. Journal of the American Statistical Association, 65:532-548.
May 1970.
7. Harrison, D. and D. L. Rubinfeld. Hedonic Housing Prices and the
Demand for Clean Air. Journal of Environmental Economics and Manage-
ment, 5(1):81-102. March 1978.
8. Nelson, J. P. Residential Choice, Hedonic Prices, and the Demand for
Urban Air Quality. Journal of Urban Economics, 5(3):357-369. July
1978.
9. Polinsky, A. M. and S. Shavell. Amenities and Property Values in a
Model of An Urban Area. Journal of Public Economics, 5(1-2): 119-129.
January—February 1976.
10. Palmquist, R. B. The Demand for Housing Characteristics: Reconciling
Theory and Estimation. Unpublished paper, North Carolina State
University, December 1981.
11. Rosen, S. Hedonic Prices and Implicit Markets: Product Differ-
entiation in Perfect Competition. Journal of Political Economy,
82(l):34-55. January/February 1974.
12. Ridker, R. 6. and J. A. Henning. The Determinants of Residential
Property Values with Special Reference to Air Pollution. Review of
Economics and Statistics, 49(2):246-257. May 1967.
13. Freeman, A. M. III. Air Pollution and Property Values: A Methodo-
logical Comment. Review of Economics and Statistics, 53(4):415-416.
November 1971.
14. Anderson, R. J., Jr. and T. D. Crocker. Air Pollution and Property
Values: A Reply. Review of Economics and Statistics, 54(4):470-473.
November 1972.
15. Freeman, A. M. III. Air Pollution and Property Values: A Further
Comment. Review of Economics and Statistics, 56(4):554-556. November
1974.
16. Polinsky, A. M. and D. L. Rubinfeld. The Air Pollution and Property
Value Debate. Review of Economics and Statistics, 57(1):106-110.
February 1975.
17. Small, K. A. Air Pollution and Property Values: Further Comment.
Review of Economics and Statistics, 57(1):111-113. February 1975.
5-48
-------
18. Harrison, D. and 0. L. Rubinfeld. Hedonic Housing Prices and the
Demand for Clean Air. Journal of Environmental Economics and Manage-
ment, 5(1):81-102. March 1978.
19. Zerbe, R., Jr. The Economics of Air Pollution: A Cost Benefit
Approach. Toronto, Ontario Dept. of Public Health, 1969.
20. Crocker, T. D. Urban Air Pollution Damage Functions: Theory and
Measurement. Prepared for U.S. Environmental Protection Agency,
Office of Air Programs. University of California, Riverside,
California, June 15, 1971.
21. Steele, W. The Effect of Air Pollution on the Value of Single-Family
Owner-Occupied Residential Property in Charleston, South Carolina.
Masters Thesis, Clemson University, 1972.
22. Freeman, A. M. III. The Benefits of Environmental Improvement:
Theory and Practice. Johns Hopkins University Press, Baltimore,
Maryland, 1979.
23. Waddell, T. E. The Economic Damages of Air Pollution. U.S. Environ-
mental Protection Agency, Office of Research and Development, Research
Triangle Park, North Carolina, May 1974.
24. Appel, D. Estimating the Benefits of Air Quality Improvement: An
Hedonic Price Index Approach Applied to the New York Metropolitan
• Area. Unpublished Ph.D. dissertation, Rutgers University, 1980.
25. Anderson, R. J., Jr. and T. D. Crocker. Air Pollution and Residential
Property Values. Urban Studies, 8(3):171-180. October 1971.
26. Wieand, K. F. Air Pollution and Property Values: A Study of the St.
Louis Area. Journal of Regional Science, 13(l):91-95. April 1973.
27. Deyak, T. A. and V. E. Smith. Residential Property Values and Air
Pollution: Some New Evidence. Quarterly Review of Economics and
Business, 14:93-100. Winter 1974.
28. Smith, V. K. and T. A. Deyak. Measuring the Impact of Air Pollution
on Property Values. Journal of Regional Science, 15(3):277-288.
December 1975.
29. Polinsky, A. M. and D. L. Rubinfeld. Property Values and the Benefits
of Environmental Improvements: Theory and Measurement. In: Public
Economics and the Quality of Life, Lowdon Wingo and Alan Evans (eds),
Johns Hopkins University Press for Resources for the Future and the
Centre for Environmental Studies, Baltimore, Maryland, 1977.
5-49
-------
30. Brookshire, D. S. e_t al. Methods Development for Assessing Tradeoffs
in Environmental Management. Vol. 2: Experiments in Valuing Non-
Market Goods: A Case Study of Alternative Benefit Measures of Air
Pollution Control in the South Coast Air Basin of Southern California.
Prepared for the U.S. Environmental Protection Agency. University of
Wyoming, Laramie, Wyoming, September 1, 1978.
31. Peckham, B. Air Pollution and Residential Property Values in
Philadelphia. (Mimeo) 1970.
32. Spore, E. Property Value Differentials as a Measure of the Economic
Costs of Air Pollution. Pennsylvania State University, Center for Air
Environment Studies, University Park, Pennsylvania, 1972.
33. Freeman, A. M. III. Hedonic Prices, Property Values and Measuring
Environmental Benefits: A Survey of the Issues. Scandanavian Journal
of Economics, 1979. pp. 154-173.
34. Nelson, J. P. Economic Analysis of Transportation Noise Abatement.
Cambridge, Massachusetts, 1978.
35. Freeman, A. M. III. Estimating Air Pollution Control Benefits from
Land Value Studies. Journal of Environmental Economics and Manage-
ment, l(l):74-83. May 1974.
36. Bureau of the Census. Population and Households by States and
Counties: 1980. PC80-S1-2.
37. Bureau of the Census. 1980 Census of Housing: Selected Housing
Characteristics by States and Counties. October 1981. HC80-S1-1.
5-50
-------
APPENDIX 5A
SOURCES OF DATA
Households
Number of households in a county in 1980.
Source: Bureau of the Census (36).
Property Value
Property value of single—family owner—occupied houses in a county in
1980.
Source: Bureau of the Census (37).
5-51
-------
SECTION 6
HEDONIC WAGE STUDIES
-------
SECTION 6
HEDONIC WAGE STUDIES
SUMMARY OF RESULTS
In this section, estimates of individual willingness-to-pay for air
quality derived from hedonic wage studies are used to calculate benefits of
alternative standards for particulate matter. Hedonic wage studies relate
observed wage differentials to various explanatory factors, including
individual-specific characteristics (such as education or prior experi-
ence), job-specific characteristics (such as risk of injury or death), and
site-specific characteristics (such as climate or air quality). The wage
differential which is attributable to variations in air quality can be
measured and used to approximate willingness-to-pay for air quality.
Estimated coefficients from selected hedonic wage studies are used to
calculate a range of benefits resulting under alternative particulate
matter standards; a point estimate of benefits is also generated for each
standard. These results are summarized in Tables 6-1 and 6-2, which
present benefits associated with selected primary and secondary standards,
respectively.* These estimates represent the present discounted value in
1982 of the stream of benefits which accrue from the attainment date (1989
for PM10 standards; 1987 for the TSP standards) through 1995, using 10
percent as the social discount rate and expressed in 1980 dollars.
Point benefits estimates range from $19.8 billion for the most lax
PM10 primary standard to $58.9 billion under the current TSP primary
* See Section 9 for a more complete description of the particulate matter
standards under consideration.
6-1
-------
Table 6-1
SUMMARY OF ESTIMATED BENEFITS* FOR HEDONIC WAGE STUDIES —
PRIMARY STANDARDS
Standard**
PM10 70 AAM/250 24-Hour
PM10 55 AAM/250 24-Hour
PM10 55 AAM/150 24-Hour
TSP 75 AGM/260 24-Hour
Minimum
9,810
16,611
20,263
29,166
Point
19,815
33,554
40,929
58,914
Maximum
37,208
61,453
72,598
107,138
Table 6-2
SUMMARY OF ESTIMATED BENEFITS* FOR HEDONIC WAGE STUDIES —
SECONDARY STANDARDS"1"
Standard**
PM10 55 AAM
TSP 150 24-Hour
Minimum
16,534
36,376
Point
33,398
73,477
Maximum
61,239
126,840
* Discounted present value in millions of 1980 dollars in 1982.
** The PM10 standards will be attained in 1989, while the TSP standards
will be attained in 1987. Benefits for all standards are accumulated
from the attainment year through 1995.
+ These estimates include benefits which accrue as current (baseline)
particulate matter levels are reduced to a secondary standard.
6-2
-------
standard. If the secondary standards are also included, then point
benefits estimates will range from $33.4 billion to $73.5 billion.
Under most standards, better than 70 percent of all benefits accrue to
the East North Central, South Central, and South Pacific regions. As the
particulate matter standard becomes more stringent, there is some redistri-
bution of benefits towards the New England, Middle Atlantic, New York-New
Jersey, South Atlantic, Midwest and Mountain regions.
These benefits estimates should be interpreted carefully, since
several assumptions are required in order to adapt both the hedonic wage
models and the data to the procedures used in calculating these benefits.
In particular, the following points should be noted:
The procedure used in extrapolating benefits from hedonic
wage models assumes that the marginal willingness to pay
for air quality declines linearly as air pollution is
abated from the baseline observation to zero. To the
extent that the actual relationship between marginal
willingness to pay and particulate matter diverges from
this linear approximation, the true benefits of abating
ambient particulate matter will diverge from the estimates
presented above. The direction of any divergence cannot be
predicted.
The hedonic wage models from which these benefits were
estimated did not include any measures of air pollution
other than TSP as explanatory varibles. Thus, the
estimated coefficient on the TSP measure may also proxy the
effects of omitted air pollutants. To the extent that this
occurs, the benefits presented above may overstate the true
benefits.
These and other caveats are discussed more fully in the conclusion to
this section.
INTRODUCTION
In the previous section, differentials in residential property values
were used to approximate individual willingness-to-pay for air quality.
6-3
-------
This approach may not accurately reflect the benefits which accrue from
improvements in air quality for the following reasons:*
• Residential property-value differentials may not reflect
the value placed on clean air in non-residential areas
(e.g.. the work place).
• Property value estimates of the benefits of clean air may
be unreliable since air quality may not vary significantly
within a 'city.
As an alternative, wage differentials may reflect willingness-to-pay
for air quality, on the premise that, all other things equal, an individual
will require more compensation for working in more highly polluted areas.
Thus, wage differentials attributable to air pollution can reflect the
economic value placed on clean air in the work place.** In addition, since
»
it is assumed that labor is mobile and will relocate in response to
disequilibrium differentials, an inter—city estimate of willingness-to-pay
for air quality is appropriate. This approach has the advantage of
allowing more variation in air pollution levels and thus produces more
stable estimates of willingness-to-pay for air quality.
It should be noted that wage differentials will reflect only the
perceived health and welfare effects of improved air quality. Any
unperceived health effects, for example, which are attributable to declines
in ambient air pollution levels will not be accounted for in hedonic wage
models. Also, real wage differentials may reflect regional differentials
in both nominal wages and prices of local goods (e.g., property values)
which are linked to variations in air quality. Real wage differentials
may, in this event, also measure at least some of the value placed on
improved air quality in residential areas.
* See Cropper and Arriaga-Salinas (1).
** Wage differentials estimated with a properly-specified model of urban
location may also reflect the value placed on clean air in residential
areas. See Cropper and Arriaga-Salinas (1).
6-4
-------
There are two distinct procedures which can be used to measure
willingness-to-pay for air quality from wage differentials:
• A labor supply curve can be estimated which includes air
quality as an argument in the individual worker's utility
function [Cropper (2); Cropper and Arriaga-Salinas (1)].
• A hedonic-wage model can be specified, relating observed
wage levels to various hypothesized determinants of wage
behavior, including air quality. [See, for example, Rosen
(4), Smith (8)].
In each case, willingness-to-pay for air pollution is derived from the
estimated model.
In this chapter, hedonic wage models are used to calculate the
benefits which result from the implementation of alternative control
strategies designed to meet specified primary and secondary national
ambient air quality standards for particulate matter (TSP).* Following a
preliminary discussion of the underlying theory of hedonic wage models,**
empirical attempts to estimate hedonic wage models are briefly reviewed.
This discussion will focus on those models which include air pollution as
an explanation of wage differentials. Coefficients from selected studies
are then used to calculate the benefits which result from the implementa-
tion of various primary standards governing ambient concentrations of
particulate matter. These benefits are subject to a number of caveats
which are summarized at the conclusion of this section.
* The discussion in this chapter centers on the use of compensating wage
differentials to measure the value which an individual places on air
quality. This differs from the wage compensation studies discussed in
the Appendix to Volume II, which use wage differentials associated with
job risk to assess the statistical value of life.
** This discussion is limited to hedonic wage models because the only
attempts to assess the effect of air pollution on wages within a labor-
supply framework do not use TSP as a measure of air pollution [Cropper
(2); Cropper and Arriaga-Salinas (1)]. Thus, the results of these
studies cannot be used to generate benefits estimates for changes in the
levels of particulates.
6-5
-------
HEDONIC WAGE MODELS: THEORETICAL CONSTRUCT
Hedonic wage models are a subset of the hedonic price models discussed
in the previous section. The reader is referred to that section for a more
complete discussion of the hedonic price technique and its implications. A
brief exposition of the hedonic wage model, adapted primarily from the
models discussed in Rosen (3,4), Thaler and Rosen (5), and Lucas (6), is
offered below.
Hedonic price techniques relate the price of a good to the character-
istics which comprise that good. The price of each good thus embodies a
series of implicit prices placed on these characteristics. Hedonic wage
models belong to this general class of functions; however, in addition to
job characteristics, hedonic wage functions also include individual-
specific characteristics (e.g., prior experience, education, etc.) as pre-
dictors of wage differentials. According to Lucas (6), individual-specific
characteristics must be included because of the basic difference between
the labor market and the consumer goods market. Unlike the sale of
consumer goods, it cannot be assumed that entrepreneurs are indifferent to
the identity of the workers to whom they "sell" j.obs.
Hedonic wage models thus relate observed wage differentials to three
factors:
• Differences in individual-specific characteristics.
• Differences in job-specific characteristics.
• Differences in site-specific characteristics.
The marginal implicit price of air quality can be derived from properly-
specified hedonic wage models; this is illustrated below within a simple
theoretical construct of the labor market.
In making their labor-supply decisions, workers are assumed to
maximize a utility function which contains the wage rate, tastes and
6-6
-------
preferences,* consumer goods and services, and job characteristics, which
can include factors relating specifically to the type of job (e.g., risk of
personal injury or death) as well as site-specific characteristics (e.g.,
air quality, climate, etc.). This utility function for worker a is written
as follows:
Ua = U(wJ, Z.., Xa, Oc, P) (6.1)
where Ua is utility of worker a.
w? is the wage accepted by worker a for the i job.
Z. is the vector of characteristics specific to the i job.
Xa is a vector of personal characteristics associated with worker
a.
QC is a vector of goods and services available for consumption.
P represents a measure of air pollution, an inverse measure of air
quality.
If pollution is a disamenity and an increase in air pollution decreases
each worker's utility, then, ceteris paribus. a trade-off exists between
the acceptance wage (which has a positive effect on utility) and air pollu-
tion.** This trade-off is illustrated in Figure 6-1 by a set of indiffer-
ence curves for two different workers; 9 and 9 . Each curve slopes
upward, indicating that, for each worker, higher wages must accompany
higher pollution levels in order to maintain a constant level of utility.
For each worker, a higher indifference curve is associated with a higher
* These can be represented by a vector of individual-specific charac-
teristics [Lucas (6)J.
** It is not necessary to assume here that the worker has detailed
knowledge of the technical relationship between air pollution and
personal health or property damage. It is assumed that he perceives
that his well-being is diminished by the presence of air pollution.
Thus, it is the perceived health and welfare effects of air pollution
which are considered here. Effects of air pollution which are not
perceived by the worker have no impact on this tradeoff between
acceptance wage and air pollution.
6-7
-------
w
W(P)
Figure 6-1. Indifference maps and equilibria
for two workers
6-8
-------
level of utility. The convex shape of each, indifference curve reflects a
diminishing marginal rate of substitution between wages and pollution.
Tastes for clean air may vary among workers; in this example, worker 1 has
a higher preference for clean air than worker 2 (i.e., worker 1 requires a
larger wage increase than worker 2 to compensate for a given change in air
pollution levels).
Each worker is faced with the task of choosing that combination of
wage rate and pollution which maximizes his utility function [represented
by Equation (6.1)], subject to the opportunities offered by the market.
Market-determined equilibrium wage rates associated with varying levels of
pollution are represented by W(P). Each worker's utility is maximized
where his marginal rate of substitution between pollution and the wage rate
(i.e., his marginal willingness-to-pay for air quality)** equals the rate
at which the market compensates for higher pollution levels [i.e., the
marginal implicit price of clean air, or 3W(P)/3P], Thus, in equilibrium,
worker 1 will accept level P., of pollution in return for wage rate W,.
Worker 2 will tolerate higher levels of pollution in return for wage rate
2 *
It is also possible that air pollution may affect labor productivity,
and this has not been considered in the discussion so far. However, the
* Pollution is a disamenity, and individuals must be paid in order to
accept higher levels of air pollution. Thus, ceteris paribus. wages
should be higher when air pollution levels are higher. Conversely, when
pollution levels decline, (or when air quality improves), wages will also
decline. This decline in wages which must accompany an improvement in
air quality in order to maintain a constant level of utility indicates
the worker's willingness-to-pay for air quality.
6-9
-------
productivity effects of air pollution and other site-specific characteris-
tics have received little theoretical discussion to date.*
In order to complete the hedonic wage model, demand factors must be
introduced into the wage-determination process. Below, a simple model of
labor demand, based on the procedure used by Thaler and Rosen (5) in
modeling job safety and labor demand, is used to illustrate the effect of
air pollution on the wage offered by the firm. This discussion abstracts
from the complications introduced when the influence of air quality on the
firm's location decision is considered. Furthermore, for purposes of
clarity in exposition, it is assumed that each individual firm can, through
some variation in its production process, alter the ambient air quality in
its immediate vicinity.
The firm combines a variety of inputs, including labor (L), to produce
marketable output (Q). Air pollution (P) may occur as a by-product of
this production process. Ignoring all other inputs except labor, the joint
production function for this firm is written as follows:
F(Q,P,L) = 0 (6.2)
The inverse of this production function is specified as:
Q = g(P,D (6.3)
It is assumed that this production function exhibits the following
properties:
* One very recent contribution in this area is presented by Roback (7) who
develops a general equilibrium model which includes both the amenity
effects and the productivity effects of site-specific characteristics.
Her general result indicates that wages will rise as air quality
deteriorates if air quality has no impact on labor productivity.
However, if air pollution adversely affects labor productivity, then the
wage change is ambiguous. Roback notes that the direction and strength
of both the amenity and productivity effects of site-specific charac-
teristics can only be determined empirically.
6-10
-------
The marginal product of labor is positive and diminishing.*
The marginal product of labor varies inversely with the
level of pollution.**
The level of output varies directly with the level of
pollution, up to some very large (and technically-
determined) level of pollution (P).+ This implies that, in
order to lower pollution levels, the firm must divert some
resources away from the production of marketable output.
The transformation locus between output Q and air quality
(the inverse of the level of pollution) is negative and
concave.
The cost of improved air quality is expressed as a function of the
level of air pollution, or G(P). These costs are positive and increasing
with the level of air quality.
The firm's profit function is written as follows:
n = g(P,L) - w(P)L - G(P) (6.4)
where w(P) is the competitive wage which must be paid at alternate air
pollution levels.
The firm maximizes profit with respect to labor and pollution; the
resulting first-order conditions are written as follows:
S? = *(p) (6-5)
3Q 3G
* 3Q/3L > 0, 32Q/3L2 < 0.
** 3Q/3L3P < 0.
+ 3Q/3P > 0 for 0 1 P < P, 32Q/3P2 < 0.
++ 3G/3P < 0, 32G/3P2 > 0.
6-11
-------
Equation (6.5) is the familiar result that labor is hired until the
marginal product of labor equals the competitive wage rate. Equation (6.6)
states that air pollution levels are adjusted so that the marginal cost of
air pollution to the firm (i.e., the additional cost of hiring labor to
work in a polluted environment) equals the marginal benefits of air pollu-
tion (i.e., additional market output and cost savings which accrue because
anti-pollution devices are not installed).
An offer function can be specified which indicates the wage paid to
the optimal number of workers at alternative pollution levels in order to
maintain a constant profit level. This offer function, 4(P,n), is
expressed as:
6 = [g(P,D - G(P) - n]/L (6.7)
' - IE <«••>
This offer function defines a set of iso-profit curves for each firm.
A set of iso-profit curves for two different firms is illustrated in Figure
6-2. These curves are positively-sloped, indicating that the cost savings
and increase in revenues from marketable output which occur as air pollu-
tion levels increase are matched by the additional wage costs which the
firm incurs in order to maintain its optimal labor stock.* These iso-
profit curves differ among firms as a result of differences in optimal
technologies. W(p), as the market-generated locus of wage rates which
correspond to each pollution level, also forms an envelope of these iso-
profit curves. At the tangency points between firms' iso-profit curves and
W(P), the internal trade-off between wages and air pollution within the
* Mathematically, 3 0. Thus, the marginal demand
price for pollution, or the iso-profit curves in Figure 6-2, is posi-
tively sloped. However, the second-order conditions do not produce an
unambiguous result. Thus, 32«f/3P2 -^ 0 [see Thaler and Rosen (5)]. The
iso-profit curves in Figure 6-2 are based on the assumption that 3 ^/3P^
< 0.
6-12
-------
w
t
(P)
P2 P
Figure 6-2. Iso-profit lines- and equilibria for
two firms
6-13
-------
firm (i.e., the slope of the iso-profit curve) equals the marginal implicit
price of air quality determined by the market.
Equilibrium in the labor market is established where the worker's
marginal willingness-to-pay for air quality equals the firm's ability to
substitute wage payments for air quality while maintaining an optimal
(i.e., profit-maximizing) input stock. W(P) thus defines a series of
tangencies between workers' indifference curves and firms' iso-profit
curves (see Figure 6-3). Workers with high preferences for clean air
(e.g., worker 1) are employed by firms which produce lower levels of air
pollution. Workers who do not place as high a value on air quality are
employed by firms which produce higher levels of pollution (and are paid
higher wages).
Hedonic wage models relate observed equilibrium wage rates, generated
by the interaction of labor supply and demand decisions, to the underlying
factors which influence these decisions. Thus, estimates of hedonic wage
models trace out the market wage locus [i.e., W(P)]. The implicit marginal
price of each underlying characteristic is expressed by the derivative of
W(P) with respect to that characteristic. The implicit marginal price
schedule of air quality, 3W(P)/3P, is illustrated in Figure 6-4. If the
labor market is in equilibrium, then this implicit price schedule also
represents the marginal willingness-to-pay for air quality of different
•
sets of workers exhibiting different preferences for air quality.
Figure 6-4 also indicates the compensated supply price functions for
pollution derived from the indifference curves for workers 1 and 2 in
Figure 6-3. These supply price functions are defined as 39/9P. They show
the increase in wage rate which is required to compensate each worker for
an increase in air pollution levels while maintaining a constant level of
well-being (or his marginal willingness—to-pay for air quality at each
level of pollution). There is only one point on each worker's compensated
supply price function which lies in the implicit marginal price schedule;
this point corresponds to the equilibrium represented by the tangency
between that worker's highest indifference curve and the market wage locus,
6-14
-------
w
W(P)
P2 p
Figure 6-3. Illustration of labor-
market equilibrium
6-15
-------
3W(P)
3P
Figure 6-4.
Implicit marginal price schedule for air
quality and compensated supply functions
for two workers
6-16
-------
W(P). Thus, the marginal implicit price schedule for air quality indicates
the equilibrium marginal willingness-to-pay for air quality at each level
of air pollution.
EMPUUCAL ESTIMATES OF HEDONIC WAGE MODELS
Hedonic wage models relate wage differentials to variations in three
types of factors:
Individual-specific characteristics, defined as those
characteristics which distinguish job-holders or job-
seeking individuals. These characteristics include such
factors as education, prior experience, race, sez, veteran
status, union-membership, etc. as possible variables in
explaining wage differentials.
Job—specific characteristics, i.e., distinct characteris-
tics exhibited by various types of jobs. Some examples of
job-specific characteristics are risk of injury or death,
work of a repetitive nature, or work within a supervisory
capacity.
Site-specific characteristics, or characteristics asso-
ciated with job location. Since job-location decisions may
be tied to residential-location decisions,* site-specific
characteristics generally reflect the levels of amenities
which influence residence choice. Wage differentials may
compensate for variations in the levels of such amenities
as pollution, climate, crowding, crime, cultural opportuni-
ties, etc.
Thus, hedonic wage models are specified in general functional form follows:
W = F(Zlf Z2, Z3)
where W is the observed wage rate.**
Z-. is a vector of individual-specific characteristics.
* That is, an individual first selects an area of residence; this decision
constrains his job search to a specific geographic area.
** Smith (8) argues that the real wage rate, and not the nominal wage rate,
is the appropriate dependent variable.
6-17
-------
Z~ is a vector of job-specific characteristics.
Z, is a vector of site-specific characteristics.
To date, most estimates of hedonic wage models generally include
individual-specific characteristics (i.e., Z^) and either job-specific
characteristics (i.e., Z2) or site-specific characteristics (i.e., Zj) as
explanatory variables [Smith (8)]. Attempts to link wage differentials to
job-specific characteristics include studies conducted by Lucas (6), Thaler
and Rosen (5), Viscusi (9), Hamermesh (10), and Olson (11). These models
have produced conflicting results regarding the effect of job-specific
characteristics on wage rates. Brown (12) failed to confirm the hypothesis
that these inconsistent estimates result from a misspecified model which
omits important job characteristics.
Other investigators have used the hedonic wage model to impute the
values attached to urban amenities (i.e., the site—specific characteristics
defined above), and thus measure the quality of urban life [see Rosen (4);
Meyer and Leone (13); Getz and Huang (14); Izraeli (15); Hoch (16)].
However, there are many site-specific characteristics which might
conceiveably affect wage rates, and several distinct methods of quantifying
some of these variables.*
A general comparison of hedonic wage model estimates indicates that,
while individual-specific characteristics exert a consistent and predict-
able effect on wages, the impact of both job- and site-specific character-
istics is less consistent. Comparisons among these models are difficult,
since they vary in their data sources, sample designs, model specification,
and measurement of both wage rates and explanatory variables. However,
even within models which include all three general characteristics as
* For example, the climate can be measured by rainfall, number of days of
sunshine, high, low, and/or average temperature, humidity, etc. Each
measures a somewhat different climatic effect; however, the inclusion of
all variables may not be feasible where degrees of freedom are limited.
In addition, interrelationships among the different measures may cause
collinearity and thus blur the effect of each factor on wages.
6-18
-------
explanatory variables, coefficient estimates for site-specific character-
istics are sensitive to the choice of the particular site-specific charac-
teristics included in the model [Rosen (4); Smith (8)].
It is not the intent of this review to evaluate all attempts to
estimate hedonic-wage models.* Instead, models which include the effect of
air pollution (specifically, ambient TSP levels) on wage rates are
evaluated for use in the calculation of benefits obtained under alternative
particulate matter (PH) standards. The selection of studies utilized in
these benefits calculations is based on the criteria discussed in Section
1. The studies used in this section meet the following specific criteria:
They each use TSP as an explanatory variable. This is an
obvious prerequisite for use in benefits calculations.
They each express the dependent variable as the real wage
rate or earnings. A regional cost-of-living deflator will
adjust earnings for differences in the prices of local
goods- (e.g., property values) which may also reflect
regional variations in air quality.
4
Each model is properly specified in that it includes,
either explicitly or implicitly, explanatory variables
measuring individual-, job-, and site-specific charac-
teristics.
Estimated coefficients produced by each model are plausible
in terms of the underlying theoretical construct.
Despite an abundance of empirical work relating wage differentials to
site—specific characteristics within a hedonic—wage framework, only nine
studies have been identified which analyze the effect of ambient levels of
particulate matter on wages [Getz and Huang (14), Izraeli (15,18), Meyer
and Leone (13), National Academy of Science (19), Hathtech (20), Roback
(7), Rosen (4), and Smith (8)J. Of these, only two studies are suitable
for use in benefits calculations [i.e., Rosen (4) and Smith (8)].
* The reader is referred to Smith (17) for a review and critique of hedonic
wage models.
6-19
-------
In their study of consumer preferences for site amenities, Getz and
Huang (14) analyze the relationship between median annual earnings of white
males employed full-time and both individual-specific variables and site-
specific variables, including air quality. A principal component* of the
ambient levels of two pollutants (sulfur dioxide and TSP) is used as an
inverse measure of air quality; this relationship is estimated for each of
nine urban occupational groups, as well as over all occupations. Getz and
Huang obtained mixed results; air pollution generally exerted an insignifi-
cant effect on median earnings except for professionals, whose earnings are
significantly higher in the presence of higher levels of air pollution.
Other explanatory variables used in this analysis include median worker age
and the proportion of adults in each SMSA who are high school graduates as
human capital variables, and a variety of site-specific characteristics,
such as the FBI violent crime index, the number of days of freezing
temperatures, the proportion of teachers with graduate degrees, and a
principal component measure of the per capita number of hospital beds and
physicians. Since the coefficients estimated by Getz and Huang do not
differentiate between TSP and sulfur dioxide in measuring the effect of air
pollution on earnings, they cannot be used in the current benefits calcula-
tions.
Meyer and Leone (13), NAS (19) and Izraeli (15,18) focus on the role
which urban amenities (or site-specific characteristics) play in explaining
wage or income differentials. Using aggregate data for 39 SMSAs, Meyer and
Leone examine the impact of both TSP and sulfur dioxide on several specifi-
cations of the dependent variable (i.e., real median family income and the
real wage rate for skilled workers, for unskilled workers, and for computer
analysts) within a log-linear functional form. In most instances TSP
levels exerted an insignificant negative effect on earnings or wages;
however, real wages for computer analysts were significantly higher in the
presence of higher TSP levels. Sulfur dioxide levels had a mixed but
always insignificant effect on the various dependent variables. The
* Principal components analysis is a statistical device which allows the
investigator to express two or more collinear explanatory variables as a
linear combination.
6-20
-------
inclusion of two highly correlated air pollution measures together in the
same model (i.e., sulfur dioxide and TSP) may explain the resultant
insignificant coefficient estimates for these variables. The NAS study
(19) is a slightly expanded version of the models estimated by Meyer and
Leone. Additional site amenities, including NOj, are included, and both
linear and log-linear specifications are estimated over various subsets of
explanatory variables. Once again, the estimated coefficients for the air
quality variables are generally not significant, and are sometimes
negative. TSP is particularly problematic in this respect: negative and
insignificant coefficients are often observed. The very small coefficient
values and the lack of significance exhibited by the air quality variables
may be due to collinearity among these variables, or between these
variables and other explanatory factors.
The performance of other explanatory variables is mixed. Some consis-
tently exhibit the "wrong" sign, and many are insignificant. Given the
2
relatively high R s which characterize some of the specifications, the
evidence indicates a severe problem with collinearity among the explanatory
variables.
A similar model was estimated by Izraeli (15), using median real wage
rates as a dependent variable from 67 SMSAs between 1964 and 1967. Using
three air pollution measures (i.e., TSP, sulfur dioxide, and benzene
soluble organic matter), Izraeli obtained a positive and significant
coefficient for sulfur dioxide; the coefficient in TSP was negative but
insignificant. When TSP was used as the only pollution variable, the
estimated coefficient was positive but not significantly different from
zero. In a recent study, Izraeli (18) estimates a more refined model using
a greatly expanded sample which includes 237 SMSAs- in the year 1970.
Median annual earnings are related to variables measuring the median age
and schooling of the population, the number of weeks worked, the racial
composition of the population and a series of environmental variables,
including air quality, the crime rate, climate, and population size and
growth. Air quality is measured by suspended particulate matter; no
additional air pollutants are included. Separate equations are estimated
6-21
-------
for males and females. In each case, TSP exerts a positive but insignifi-
cant effect on median earnings.
Given the aggregate nature of the data used in the above-mentioned
studies, no job-specific characteristics can be explicitly included as
explanatory variables. Similarly, individual-specific characteristics,
such as prior experience, education, etc., are difficult to include in
their models. However, median schooling is used as an explanatory variable
in both Meyer and Leone (13) and NAS (19). Some other explanatory
variables, such as the median age or racial composition of the population,
may also reflect the education or experience levels of the population.
Since these models do not account for the impact of both individual- and
job-specific characteristics on wage or earnings differentials, they cannot
be regarded as properly—specified hedonic wage models. Thus, benefits
calculations should not be based on their results.
The remaining studies [Mathtech (20), Roback (7), Rosen (4), Smith
(8)] use micro-data to relate wage rates or earnings reported by surveyed
individuals to individual-specific characteristics, job-specific
characteristics, and site-specific characteristics. The Hathtech study
(20) used observations drawn from the University, of Michigan Panel Study
of Income Dynamics for the 1971 interview year for 699 counties. The log
of the real wage,* is related to three measures of pollution (i.e., TSP,
sulfur dioxide, and nitrogen dioxide), entered in both linear and quadratic
terms. Since pollution data is available only over 247 counties, two
procedures were suggested to replace missing observations:
• The use of the means of the observed values.
• A method adapted from a technique developed by Dagenais
(21), which involves regressing each pollution variable in
all remaining (non-pollution) explanatory variables and on
relevant auxiliary variables. Values for the missing
* Defined as hourly money income from labor received by the head of house-
hold deflated by the BLS indicator of comparative living costs for i
family of four at the lowest living standard.
6-22
-------
pollution variables are predicted using these regression
results.
Regression results are reported using a restricted subsample where all
missing observations are deleted, as well as for the full sample (i.e.,
1,395 observations),* using the two procedures outlined above to replace
missing pollution variables. Neither sulfur dioxide nor nitrogen dioxide
significantly affect real wages. The coefficient on the linear TSP
variable is positive and significant only when the mean values of the
pollution variables are used to approximate missing observations; the
coefficient on the quadratic TSP term is negative and significant in this
instance. Thus, real wages will rise at a declining rate as TSP concentra-
tions increase. However, the estimated elasticity of real wage with
respect to the TSP level is -0.16, measured at the mean value of TSP;**
this contradicts a. priori expectations regarding the relationship between
the real wage and TSP levels. These implausible results leads to the
rejection of these coefficients as a basis for benefits calculations.+
Rosen (4), Roback (7), and Smith (8) each use micro-data from the
Current Population Survey (CPS) in their respective analyses. Each
includes TSP levels (measured at the annual geometric mean) as an explana-
tory variable in conjunction with several other individual-specific, job-
* This sample is restricted to households which received no transfer
income and whose head worked 400 hours or more per year.
** This elasticity indicates the percentage change in the real wage rate
which will occur for a given percentage change in TSP levels. It thus
measures the responsiveness of the real wage to changes in TSP levels.
In this instance, a 10 percent increase in TSP levels will lead to a 1.6
percent decline in the real wage.
+ The hedonic wage model specified in the Mathtech study was also
estimated over various portions of the data set based on sex, age, and
race. Coefficients estimated for air pollution variables were seldom
significantly different from zero in these equations.
++ The Current Population Survey data allows the investigator to link
specific individual- and job-specific attributes to each surveyed
individual; this provides better control over these explanatory
variables than the more aggregated data used in many previous studies.
6-23
-------
specific, and site-specific factors, and each produces plausible estimates
for the TSP coefficients. Roback's model uses nominal wage rates as the
dependent variable, while both Rosen and Smith specify the dependent
variable as the real wage rate.
Variations in nominal wages or earnings may be explained in part by
regional differences in the cost of living. Cost-of-living differentials
may reflect the following:
• Regional differences in the prices of goods which are
exchanged between regions due, for example, to transport
costs.
• Price differentials for nontraded goods (e.g., housing)
which are attributable to variations in environmental
conditions across regions.
The effect of these regional variations in cost of living should be
accounted for in assessing the relationship between wage rates and air
quality. Roback (7) does not account for this factor; thus, this study is
excluded from further consideration.
Both Rosen (4) and Smith (8) use an appropriate data base to estimate
well-specified hedonic models which meet the following criteria:
• Each includes TSP as an explanatory variable.
• Each specifies real wage or earnings as the dependent
variable.
• Each includes, either explicitly or implicitly, explanatory
variables measuring individual-, job- and site-specific
characteristics.
• Each produces a plausible relationship between TSP and the
real wage rate.
Thus, both studies meet the general criteria for use in benefits calcula-
tions. A more detailed review of each study is offered below; this discus-
sion is summarized in Table 6-3.
6-24
-------
Table 6-3
SUMMARY OF SELECTED HEDONIC WAGE MODELS
ro
in
Study
Rosen
(4)
smith
(8)
Sample
Males out of school
& reporting earn-
ings in 1969 over
19 major SMS As.
Obtained from 1970
CPS.
All surveyed indi-
viduals residing in
44 major SMSAs who
reported earnings
in 1977. Obtained
from 1978 CPS.
Dependent Variable;
Pollutant
Total real earnings
(wages & salaries +
self -employment
income); Annual
geometric mean for
particulates
(ug/m3).
Real wage rate;
Annual geometric
mean for particu-
lates (ug/m3).
Functional
Form
Semi -log
Semi-log
Estimated Coefficients;
Elasticities of
Pollution Variables
(At Means)
Estimated coefficients
range from 0.000553 to
0.0015; Elasticities
range from 0.0627 to
0.1702.
Estimated coefficients
range from 0.000615 to
0.00112; Elasticities
range from 0.0413 to
0.075.
Estimated
Benefits;
Base Year
A reduction in
levels by 1 ug/m3
leads to a change in
mean annual earnings
ranging from $4.92
to $13.35; 1970.
Reducing TSP levels
by 1 ug/m will
decrease mean annual
earnings by $6.05 to
$16.50; 1978.
-------
In his analysis of wage rate differentials and urban amenities, Rosen
used data from the 1970 CPS over 19 major SHSAs for males who were out of
school and who reported earnings in 1969. He uses a semi-log specification
where the dependent variable is the log of real annual earnings* and the
explanatory variables are entered in linear form. In addition to TSP
levels, air pollution is also measured by sulfur dioxide levels and by the
number of inversion days. Individual-specific variables include schooling
and previous experience (entered in both linear and quadratic form), as
well as dummy variables for race, sex, marital status, veteran status, and
head-of-household. Dichotomous variables are also entered for self-
employed individuals, government employees, full-time employees, and
individuals who worked for 35 hours per week or less, as well as for
individuals who were not employed at any time during the year and indivi-
duals who were unemployed once during the year. Job-specific characteris-
tics are not explicitly recognized. However, Rosen uses dummy variables to
distinguish across one—digit occupational and industry classes; this may
provide a crude measure of differences in characteristics common to various
jobs across the sample. Finally, measures of site-specific characteristics
include climate (i.e., the number of sunny days, the number of rainy days,
and the number of days when the temperature exceeded 90°F), crime (measured
by the total crime rate), crowding (i.e., population density, population
size, and whether the surveyed individual lives in the central business
area), and labor market conditions (approximated by the unemployment rate
and the rate of population growth). In addition to the air pollution
variables cited above, a measure of water pollution is also included.
Rosen estimated his model using all individual- and job-specific
characteristics and various combinations of the site-specific attributes
discussed above.** Estimated coefficients for individual-specific attri-
butes are generally significant and of the same sign and order of magnitude
* Real annual earnings are defined here as nominal earnings (wages and
salaries, plus self-employed income) reported in the CPS divided by a
cost-of-living index for either low-, medium-, or high-expenditure
families, depending on individual circumstances.
** Regression results reported by Rosen are listed in Appendix 6A.
6-26
-------
as those observed in other studies using less restricted samples. Results
•
for the site-specific variables are less robust. Measures of climate,
crime, and market conditions generally affect wage rates in the expected
fashion; however, the size of the effect varies with the particular set of
site-specific attributes included in the model. Significance levels are
also inconsistent across specifications; this probably reflects varying
degrees of multicollinearity among these variables. TSP had a consis-
tently positive and generally significant impact on real earnings.*
Reported significant estimates for the TSP coefficient range from 0.00055
to 0.00150; this translates to elasticity estimates ranging from 0.062 to
0.170.** The measured impact of changes in TSP levels on real earnings is
thus very slight: at the mean TSP level, an increase in TSP concentrations
by 10 percent will increase real earnings by less than 2 percent.
Smith (8) explicitly examines the effect of both site-specific and
job-specific characteristics on real wage in his hedonic wage model, using
data obtained from the 1978 CPS to define the dependent variable and
several explanatory variables over 44 SMSAs. Like Rosen, Smith uses a
semi-log functional form which specifies the log of the real wage rate as
a function of several explanatory variables. Individual-specific attri-
butes include years of schooling and years of job experience (both entered
in linear and quadratic form), as well as a series of dummy variables
reflecting socio-economic characteristics. Job-specific characteristics
* The number of inversion days also had a positive and significant impact
on real wage. The results for sulfur dioxide are less robust; signifi-
cant negative coefficients appear in several instances.
** For the semi-log functional form, the elasticity of the real wage with
respect to TSP at a particular level of TSP is computed as the product
of the estimated coefficient on the TSP variable and the level of TSP.
Thus, the responsiveness of the real wage to changes in TSP concentra-
tions varies with the level of TSP. Here we calculate elasticities at
the mean TSP level observed in the sample.
+ Defined as the nominal hourly wage divided by the 1977 Bureau of Labor
Statistics' cost-of-living index for families at an intermediate living
standard. Since this index is only available for 27 SMSAs, Smith uses a
procedure suggested by Cebula (22) and Cebula and Smith (23) to estimate
this index for the remaining SMSAs.
6-27
-------
include injury rates, an index of exposure to carcinogens, an indicator of
worker knowledge of job hazards,* and occupational dummy variables.
Smith's model also allows for an interaction between experience and the
effect of product market uncertainty on the firm's willingness to invest in
on-the-job training (OJT). Another set of risk-interaction terms for race,
head of household, and union membership are also included in some regres-
sions. In his preliminary analysis. Smith experimented with a large number
of site-specific characteristics, including various measures of climate,
cultural amenities (i.e., dummy variables indicating presence of symphony
or live theatre, the number of major newspapers, art museums, and profes-
sional sporting teams), the crime rate, the unemployment rate, and the
number of hospitals. Three measures of air pollution, TSP, sulfur dioxide
and ozone, were also considered. Here, Smith found that TSP provided the
most consistent results. The model was estimated over all individuals in
the sample and separately for males and for females; Smith's best results
using the most robust set of site-specific, attributes are presented in
Table 6A-3 in Appendix 6A.
The site-specific attributes which are included in Smith's final
analysis are TSP levels, the unemployment rate, the crime rate, and the
mean annual percentage of possible sunshine (i.e., the number of hours of
actual sunshine/the number of hours between sunrise and sunset).
Generally, the estimated coefficients for these and other explanatory
variables are statistically significant and their signs conform to both a.
priori expectations and the results of previous studies.**
Real wage rates are significantly higher for higher levels of TSP;
coefficient estimates over the entire sample range from 0.00083 to 0.00087.
* Measured as the relative number of workers in each industry covered by
collective bargaining agreements with provisions relating to health and
safety conditions.
** Smith notes that his estimated coefficient for the unemployment rate
differs in sign from estimates reported by Rosen (4). He suggests that
differences in the industrial composition of the two samples may explain
this inconsistency.
6-28
-------
The elasticity of the real wage rate with respect to TSP; calculated at the
mean observed TSP levels, varies from 0.056 to 0.058. Smith's results for
men indicate that elasticities range from 0.0726 to 0.0750; this confirms
Rosen's finding that TSP levels have an extremely small, albeit statisti-
cally significant, impact on real earnings.
Generally, the hedonic wage studies cited above produce consistent
coefficient estimates for individual-specific attributes. However, there
is less agreement among these models with respect to coefficient estimates
for both job-specific and site-specific attributes. Smith (8) suggests
that the constraints imposed by data availability result in non-random
samples which, in turn, produce diverse coefficient estimates. He tests
the sensitivity of his site-specific and job-specific coefficient estimates
to sample composition by deleting selected SMSAs and industry classes from
the sample and re— estimating his model. He concludes that coefficients
estimated for site-specific attributes are quite sensitive to sample compo-
sition, while job-specific effects are less sensitive in this regard.
*
Before proceeding with benefits calculations based on these results
derived by Rosen and by Smith, some general remarks concerning certain
unresolved empirical issues which arise within hedonic wage models are in
order.
Estimates of hedonic wage models may be subject to bias from at least
two sources:
Bias resulting from the omission of important explanatory
variables.
Sample selection bias.
Investigators who attempt to estimate hedonic wage models are faced with
the task of choosing among a large number of potential explanatory
variables. The omission of important variables from the model introduces a
bias of unknown magnitude into the estimates of the model's parameters.
6-29
-------
Hedonic wage models may also be biased as a result of sample trunca-
tion. Many of the empirical models cited above are estimated over somewhat
restricted samples.* Even when arbitrary restrictions are not placed on
the data set, the sample is necessarily restricted to those individuals who
report wages for the sample period [Smith (8)]. In either case, a sample
selection bias is introduced into the estimated coefficients.** Addi-
tional assumptions are required in order to use these coefficients to make
some inference about the behavior of those individuals not included in the
sample.
BENEFITS
Benefits calculations for the alternative primary and secondary parti-
culate matter standards described in Section 9 will be based on coefficient
estimates obtained by Rosen (4) and Smith (8) for their respective hedonic
wage models. These estimates will be used to calculate individual
willingness-to-pay for air quality and benefits will be extrapolated over
the labor force.
In order to use the coefficients estimated by Rosen and Smith to
obtain an assessment of benefits over the entire labor force, the following
assumptions are required:
The estimated coefficient on the TSP variable in each study
accurately captures the effect of changes in TSP levels on
real earnings.
Benefits based on these coefficients reflect the
willingness-to-pay for air quality of those individuals in
the labor force who are not included in the samples from
which these coefficients were estimated.
* For example, Rosen (4) restricted his sample to males, and the Mathtech
study (20) deleted all households which reported transfer income or
whose head worked less than 400 hours per year.
** The reader is referred to He c km an (24,25) and Gronau (26) for a discus-
sion of sample selectivity. Wales and Woodland (27) survey various
methods of estimating labor supply functions using truncated samples.
6-30
-------
Measures of TSP may be correlated with, other possible explanatory
variables, including other types of air pollution, which are not considered
in the model. In this event, the estimated coefficient for the TSP
variable may also capture the effect of these omitted variables on real
earnings. It is assumed here that correlation between TSP and excluded
explanatory variables does not substantially affect the coefficient
estimates obtained in the selected studies. This assumption is certainly
tenuous, particularly in view of the sensitivity of Rosen's reported
results for TSP as different combinations of site-specific characteristics
are introduced within his model. There is also strong evidence that TSP
concentrations tend to be highly correlated with other air pollutants.
Unfortunately, very few hedonic wage models report results for more than
one measure of air pollution. Those which include the effect of more than
one air pollutant in real wages or earnings are deficient in other
respects.* It is possible that the estimated coefficient on TSP as the
sole pollutant in this type of model may also proxy the effects of other
pollutants in real wages or earnings. However, it is difficult to assess
the extent to which this might occur because neither of the studies with
well-specified models (Rosen and Smith) report equations containing more
than one air pollutant.
The possibility of biased coefficient estimates due to sample selec-
tion or truncation must also be .considered. Smith attempted to measure the
bias introduced by sample truncation by applying adjustment indices
developed by Olsen (28) to his OLS estimates.** The resultant adjusted
coefficients increased in absolute value by one to two percent, while the
standard errors of these coefficients rose by one percent. Smith concluded
that sample truncation bias did not pose a serious problem in his analysis.
* See, for example, Meyer and Leone (13), MAS (19), or Izraeli (15).
** Olsen (28) developed a simple approximation to the maximum likelihood
estimators for the truncated regression model. A table of conversion
factors is applied to OLS estimators in order to produce these approxi-
mate maximum likelihood estimates.
6-31
-------
The hedonic-wage models developed by Smith (8) and Rosen (4) were
estimated over different samples,* using somewhat different model specifi-
cations. Despite this, the results obtained within both models for TSP
corroborate each other. The estimated coefficients appear to be quite
robust with respect to both the sex and the residence of surveyed indivi-
duals;** these results support the use of these coefficients in calculating
willingness-to-pay for air quality for every worker in the labor force.
Both Rosen and Smith use a semi-log specification of the hedonic
wage function, expressed as follows:
ln(W) - o + 0 TSP + XX (6.9)
where W is the real wage rate (Smith) or real annual earnings (Rosen).
TSP is the annual geometric mean level of total suspended particu-
lates.
X is a vector of other explanatory variables.
a, p and X are estimated parameters.
Estimates of this model reveal the equilibrium value of air quality.
From this, the decline in real wage rate or earnings which will accompany a
marginal improvement in air quality; (i.e., the marginal willingness-to-pay
* Rosen used a sample of males over 19 major SHSAs drawn from the 1970
CPS, while Smith used a sample of all surveyed individuals over 44 SMSAs
drawn from the 1978 CPS.
** Smith reports that the difference between coefficients estimated
separately for males and for females is not statistically signifi-
cant. His coefficient estimates for the total sample over 44 SHSAs
lie within the range reported by Rosen for 19 SHSAs.
+ Here, it is also assumed that the coefficients estimated by Smith (8)
and Rosen (4) also apply to those individuals in the labor force who
were not represented in these studies because they did not report
earnings for the sample periods. In essence, it is assumed that these
benefits calculations are not affected by sample selection bias. As
discussed above, sample selection bias does not appear to influence the
coefficient estimates generated by Smith (8).
6-32
-------
for air quality) can be estimated. This marginal willingness-to-pay for
air quality is calculated as the first derivative of real wages (or
earnings) with respect to air pollution. For the semi-log function given
in Equation (6.9), the marginal willingness-to-pay for air quality at a
given level of real earnings is defined as:
If the estimated coefficient ({J) is positive, then individuals must receive
higher wages in order to compensate for deteriorations in air quality.
Their marginal willingness-to-pay for air quality, or the decrease in the
wage rate which they would accept in order to enjoy improved air quality,
is positive and varies in direct proportion with their wage level. Indivi-
duals with higher real earnings will relinquish more of their earnings in
response to a marginal improvement in air quality as compared to those with
lower real earnings. The second derivative of Equation (6.9) is expressed
as follows:
W > 0 (6.11)
Marginal willingness-to-pay for improved air quality will increase at an
increasing rate as air quality deteriorates. The magnitude of this effect
will also vary directly with the wage rate.
Table 6-4 lists calculations of individual marginal willingness-to-pay
for air quality on an annual basis using coefficients estimated by Smith
(8) and Rosen (4).
The implementation of the proposed particulate matter standards will
probably result in a non—marginal change in ambient TSP concentrations. In
order to calculate the precise effect of these non-marginal improvements in
air quality on real wages (and thus on benefits), some knowledge of the
supply price function for TSP for each individual is required. This
6-33
-------
Table 6-4
ESTIMATED BENEFITS FOR MARGINAL CHANGES IN AIR QUALITY*
BASED ON RESULTS REPORTED BY ROSEN (4) AND SMITH (8)
Study
Rosen (4)
Smith (7)+
Total sample
Males
Females
Change in average annual real earnings**
Minimum
$ 10.44
13.22
20.17
7.70
Maximum
$ 28.33
13.88
20.84
8.44
* A change in TSP levels of 1 jun/g is assumed.
** Calculated at the average wage or earnings levels reported for each
sample and expressed in 1980 dollars.
+ The reported average nominal wage rates are divided by the cost-of-
living index for a 4-person family at an intermediate expenditure level
in order to calculate the real wage. The change in the real wage as a
result of the change in TSP levels is then multiplied by 2,080 hours in
order to approximate annual real earnings.
function is not estimated within the hedonic wage model. However, these
benefits can be approximated using a simple technique suggested by Freeman
(29).
Figure 6-5 illustrates the equilibrium marginal implicit wage that the
sampled individuals will accept in return for working under varying levels
of ambient TSP levels. This marginal implicit wage, W (TSP), is the first
derivative with respect to TSP of the hedonic wage model specified above as
calculated in Equation 6.10. Each point in W (TSP) corresponds to a
single point in each individual's supply price function for TSP. S(TSP)i
is the hypothesized supply price function of individual 1, indicating that
this worker must be paid higher real wages as compensation for higher TSP
levels in his work environment.
6-34
-------
Pollution
Figure 6-5,
Illustration of benefits calculation for
non-marginal changes in TSP levels
6-35
-------
Benefits calculations for non-marginal changes in TSP levels involve
measuring the area under S(TSP)j_ between the two pollution levels (i.e.,
area AP^B for a decline in TSP levels from Pj to P2). Since SdSPJj is
not directly observable, this area is approximated by assuming that the
supply price functions for each individual for TSP decrease linearly from
the observation point (i.e., point A) to zero as TSP is continuously
abated. Thus, benefits can be calculated as the difference between the
areas of two triangles AP^O and CPjO, or
Benefits = | [ (APX ' OP],) - (CP2 ' OP2>] (6.12)
For the semi-log specification of the hedonic wage model, benefits are
calculated as follows:
Benefits = *~- P, - =- (6.13)
where P^ is the initial TSP level.
?2 is the TSP level after implementation of the appropriate
control strategy.
and all other variables are as defined above.
Clearly, the accuracy of this approach to benefits calculations
depends on how closely the linear segment CA approximates the curvilinear
segment BA. If segment CA falls below segment BA (as drawn in Figure 6-5),
then Equation (6.13) will underestimate the true benefits obtained from the
given non-marginal change in TSP concentrations. Conversely, if segment CA
lies above segment BA, the result is an overestimate of these benefits.
Since segment BA is not observed, the direction and extent of the bias in
benefits calculations resulting from this approach is unknown.
The average real wage expressed in 1980 dollars is used to calculate
individual willingness-to-pay for air quality; these benefits are multi-
plied by average annual work hours (i.e., 2,080 hours per year) to
6-36
-------
approximate average annual benefits per individual at the attainment date.
Individual benefits are then multiplied by the projected labor force at the
attainment date in order to extrapolate benefits over the population.
These benefits are extended over the relevant time horizon under the
assumption of continuous growth in the labor force, and then discounted
back to 1982.
Both Rosen and Smith use data aggregated to the SHSA level in
estimating their models; however, information on ambient TSP concentrations
is compiled by county. In this study, benefits from hedonic wage models
are calculated at the county level; this facilitates comparisons with the
benefits estimates from health effects studies and property value studies
presented elsewhere in this report.
The following information is required for benefits calculations from
hedonic wage models:
• Real wage rates, expressed in 1980 dollars.*
• Projected changes in ambient TSP concentrations for a given
attainment date resulting from the attainment of a speci-
fied standard.
• Projections of labor force growth into the future.
Data sources and the procedures used in transforming the raw data into a
more suitable form are summarized in Table 6-5.
Benefits estimates are calculated for six alternative particulate
matter standards. For each standard, minimum and maximum estimates provide
* Technically, the real wage at the attainment date should be used. In
these benefits calculations, the average real wage in each county in 1978
is used. The average nominal wage rate in 1978 for each county is
deflated by the BLS cost-of-living index for intermediate expenditure
families in order to approximate this average real wage. The consumer
price index is then used to inflate this real wage to 1980 dollars. The
real wage is then extrapolated to the attainment date and beyond using
income projections provided by the Bureau of Economic Analysis,
Department of Commerce (30).
6-37
-------
Table 6-5
DATA SOURCES AND TRANSFORMATIONS
Variable
Source
Comments
Nominal
wage
rate
Payroll and employment information
for non-government* and Federal
government employees obtained from
County Business Patterns 1978 pub-
lished by the Bureau of Census (31)
The combined Federal government and non-
government payroll is divided by the total
number of employees reported by each sector
in order to estimate average annual earnings
per individual. This estimate is divided by
2,080 (i.e., no. of hrs. worked/year) as an
approximation to the average hourly wage.
Cost-of
living
deflator
Index of comparative costs based on
an intermediate budget for a 4-
person family (Autumn 1977),
published by the Bureau of labor
Statistics.
This index is calculated for SMSAs and for
non-metropolitan areas within each census
region. Generally the SMSA index associated.
with each county is used to deflate that
county's nominal wage rate. When a defla-
tion factor is not available for an SMSA the
U.S. metropolitan index is used. The
regional non-metropolitan index applies for
counties which are not part of an SMSA.
Consumer
Price
Index
Bureau of Labor Statistics
Consumer price indices are published on a
regional basis for the 4 census divisions.
These regional price indices are used to
convert benefits estimates to 1980 dollars.
Income
growth
projection
The real wage rate at the attain-
ment date is extrapolated from the
real wage rate in 1980 using income
growth projections to year 2000.
Continuous growth is assumed.
Growth rates are calculated from
projections of personal income to
year 2000 obtained from the U.S.
Bureau of Economic Analysis (30)
Income projections are available by state.
TSP
concentra-
tions
Provided by EPA
PM10 concentrations are converted to equiva-
lent TSP levels expressed as the annual geo-
metric mean. These TSP readings are then
transformed to an annual geometric mean to
conform with data used by Rosen (4) and
Smith (8). This data in its original form
represented the worst incidence of air
pollution within each area. Data was trans-
formed in order to derive a measure of
average exposure throughout the county. The
procedure used in this transformation is
discussed in Section 9.
Labor
force
Labor force as of attainment date
is extrapolated from the labor
force in 1978 under the assumption
of a continuous growth rate. Labor
force by county in 1978 is obtained
from unpublished data provided by
the Bureau of Labor Statistics.
The employment & population projec-
tions to 2000 used to calculate the
growth factor are provided by the
Bureau of Economic Analysis (32,33)
Employment projections are available on an
SMSA level and are used to calculate labor
force growth for counties within these
SMSAs. Population projections at the state
level are used to approximate labor force
growth in rural counties.
* Excluding self-employed individuals, railroad employees, farm workers, and domestic-service
workers, and government employees at the state and local levels.
6-38
-------
a likely range of benefits, while a single point estimate indicates the
best defensible benefits estimate.* The choice of coefficient for the
maximum benefits estimate is based on a review of a series of specifica-
tions estimated by Rosen (4). The coefficient was selected from that
specification which included a maximum number of explanatory variables,
most of which had a statistically significant effect on real earnings.
This minimizes, to the extent possible, the impact of omitted explanatory
variables on the estimated TSP coefficient.
The point benefits estimate is derived from results reported by Smith
(8). In order to incorporate the uncertainty which is associated with this
estimate due to the stochastic nature of the estimation procedure, the
lower bound of the confidence interval around this point estimate is used
to define the lower bound of the reported range of benefits.**
The coefficients on the TSP measure derived by both Smith and Rosen
were estimated from air pollution data at the SMSA level. Table 6-6
summarizes the extent of ambient TSP exposure which each model assumed, and
the adjustments made to the design values used in the current benefits
estimates in order to replicate these conditions.
Table 6-7 reports calculations of individual willingness-to-pay for
air quality, based on the coefficients and the air pollution data which are
used in benefits estimates. Estimates of willingness to pay for marginal
changes in concentrations of particulate matter range from $3.32 per year
to $10.98 per year, depending on the sample. These estimates appear to be
quite reasonable, and reflect a highly inelastic relationship between real
wages and particulate concentrations.
* The coefficients used in these calculations are given as follows:
Minimum estimate 0.000431; Point estimate 0.000871; Maximum estimate
0.000921.
** This procedure is not used to define the maximum benefits estimates
because sufficient information is not available to calculate a statisti-
cally-meaningful confidence interval around the coefficients estimated
by Rosen (4).
6-39
-------
Table 6-6
AIR QUALITY DATA: ORIGINAL MODELS AND CURRENT BENEFITS ANALYSIS
Study
Geographic Area
Monitor(s) Used in
Original Study
Pollution Measure Used in
Benefits Analysis*
Rosen
19 SMSAs
Monitor representing
worst incidence
within SMSA
Design value monitor in
county
Smith
44 SMSAs
Average over all
monitors within SMSA
Average over all monitors
in county
* See Section 9 for a complete description of the air quality data used in this
benefits analysis.
-------
Table 6-7
ESTIMATED BENEFITS FOR MARGINAL CHANGES IN AIR QUALITY*
Standard
PM10 70 AAM/250 24-hr
(93 counties)
PM10 55 AAM
(161 counties)
PM10 55 AAM/250 24-hr
(163 counties)
PM10 55 AAM/ 150 24-hr
(297 counties)
TSP 75 AGM/260 24-hr
(282 counties)
TSP 150 24-hour
(499 counties)
Change in Annual Real Earnings**
Minimum
$3.32
3.52
3.52
3.65
3.62
3.71
Point"*"
$6.71
7.11
7.12
7.38
7.31
7.49
Maximum
$11.36
11.36
11.36
11.21
11.15
10.98
* A change in ambient TSP levels of 1 ug/m3 (annual geometric mean) at the
design value monitor is assumed. This is roughly equivalent to a change
in ambient PM10 levels of 0.55 ug/m3.
** Calculated at the average real wage over all affected counties for each
standard, expressed in 1980 dollars and multiplied by 2,080 hours in
order to approximate annual real earnings. Thus, these figures repre-
sent a change in average annual earnings. Similar calculations by Smith
(8) suggest that the range of estimated benefits could be somewhat wider
if these benefits were cumulated over observed individual earnings.
+ The coefficients used here reflect the relation between real wage rates
and average exposure to ambient TSP. Thus, for these calculations, a 1
|ig/m3 change in ambient TSP at the design value monitor is translated
into the equivalent change in average exposure using the ratio of the
mean average exposure to mean design value exposure for each standard.
6-41
-------
Benefits estimates for three alternative PM10 standards and one
associated PM10 secondary standard, as well as for the current TSP primary
and secondary standards, are presented in Tables 6-8 through 6-13. The
standards are arranged in the following order:
• PM10 Primary Standard ~ 70 AAM/250 24-hour expected value
(Table 6-8).
• PM10 Secondary Standard — 55 AAM (Table 6-9).
• PM10 Primary Standard — 55 AAM/250 24-hour expected value
(Table 6-10).
• PM10 Primary Standard — 55 AAM/150 24-hour expected value
(Table 6-11).
• TSP Primary Standard — 75 AGM/260 24-hour second-high
(Table 6-12).
• TSP Secondary Standard —• 150 24-hour second-high (Table 6-
13).
In each case, a range of benefits, accompanied by a point estimate, is
«
presented by air pollution control region and then aggregated over the
entire nation. The stream of benefits is estimated from the attainment
date (1989 for PM standards; 1987 for TSP standards) to 1995 and then
discounted back to 1982, using a 10 percent social discount rate. The
present discounted value of benefits in 1982 is expressed in millions of
1980 dollars.
Table 6-8 presents estimates of the benefits which will accrue when
the most lax standard (PM10 70 AAM/250 24-hour expected value) is imposed.
Total benefits estimates range from almost $10 billion to about $37
billion, with a point estimate of about $19.8 billion. About one-third of
these benefits accrue to the South Pacific air pollution control region.
Another one-third will accrue to the East North Central region, while the
South Central region will receive about 12 percent of the total benefits.
The remaining benefits are divided among the Middle Atlantic region (5
percent), the South Atlantic region (5 percent), the Midwest region (2
percent), the Mountain region (4 percent), and the North Pacific region (5
6-42
-------
Table 6-8
ESTIMATED BENEFITS FOR: HEDONIC WAGE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
10.0
436.3
533,
2842.
1230.8
150.8
382.8
3697.8
525.2
,6
.2
Point
Estimate
0.0
20.3
881.4
1077.9
5741.1
2486.1
304.6
773.3
7469.3
1060.8
Maximum
0.0
40.1
1404.7
1850.5
14659.9
4875.1
547.0
1259.8
10848.6
1721.8
Total U.S.
9809.6
19814.9 37207.5
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
3926.6
7931.4 14893.3
6-43
-------
Table 6-9
ESTIMATED BENEFITS FOR: HEDONIC WAGE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
117.6
121.9
772.
1183
3994.
1918.2
426.2
772.
6557.
.5
.2
.1
.5
.1
Total U.S.
670.7
16534.1
Point
Estimate
237.6
246.3
1560.3
2390.0
8067.9
3874.7
860.8
1560.5
13245.1
1354.8
Maximum
362.3
541.1
2494.5
3889.2
21219.5
7294.4
1394.4
2287-7
19474.0
2281.6
33398.0 61238.5
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
6618.2
13368.4 24512.4
6-44
-------
Table 6-10
ESTIMATED BENEFITS FOR: HEDONIC WAGE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
117.6
121.9
772.5
1183.2
4004.2
1940.5
428.0
772.5
6557.7
713.0
16611.1
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
237.6
246.3
1560.3
2390.0
8088.2
3919.7
864.6
1560.5
13246.1
Maximum
362.3
541.1
2494.5
3889.2
21252.2
7359.9
1399.8
2287.7
19475.5
1440.3
2390.5
Total U.S.
33553.6 61452.7
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
6649.1
13430.7 24598.1
6-45
-------
Table 6-11
ESTIMATED BENEFITS FOR: HEDONIC WAGE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South. Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
381.4
239.0
1121.6
1509.4
4545.9
2251.1
597.4
1159.4
7239.5
1217.8
20262.5
Point
Estimate
770.5
482.8
2265.5
3048.9
9182.4
4547.0
1206.8
2341.9
14623.3
2459.9
Maximum
1233.9
944.6
3533.7
4825.4
22940.6
8350.8
1898.4
3244.4
21613.7
4012.6
40929.1 72598.1
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
8110.6
16383.0 29059.3
6-46
-------
Table 6-12
ESTIMATED BENEFITS FOR: HEDONIC WAGE STUDIES
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AGM/260 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
369.7
303.9
1734.6
2245.3
6931.2
2999.6
1018.
1393,
10816.
.1
.1
.1
Total U.S.
1354.6
29166.1
Point
Estimate
746.7
613.8
3503.7
4535.4
14000.7
6058.9
2056.5
2814.0
21848.0
2736.2
Maximum
1145.9
1190.8
5471.1
7156.6
36390.0
11855.5
3170.1
3947.9
32258.7
4551.2
58913.9 107137.7
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
8156.3
16475.3 29961.0
6-47
-------
Table 6-13
ESTIMATED BENEFITS FOR: HEDONIC WAGE STUDIES
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
1105.3
600.8
2588.2
2807.4
8262.0
3536.0
1672.8
1867.8
11858.1
2232.7
1213.5
5228.0
5670.8
16638.8
7142.5
3379.0
3772.9
23952.7
3475.5
2158.6
7933.7
8832.1
39904.8
12477.3
5015.8
5198.6
35013.4
Total U.S.
2077.2
36375.5
4195.7
6829.8
73476.5 126839.6
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
10172.4
20547.7 35470.7
6-48
-------
percent). Tie New York-New Jersey region accounts for a miniscule propor-
tion of total national benefits (0.1 percent). Tie entire New England
region is projected to be in compliance with tie standard by 1989; tins, no
benefits will accrne to tiis region as a result of tie proposed standard.
Tie remaining tables indicate tie benefits wiici are associated witi
increasingly stringent PH10 standards, as well as witi tie current TSP
standard (wiici is more stringent tian tie strictest PM10 standard). As
more stringent standards are applied, benefits increase accordingly. Point
benefits estimates rise from $19.8 billion for tie most lax PH primary
standard to $73.5 billion under tie current TSP secondary standard. Tie
regional distribution of tiese benefits also varies as stricter standards
are imposed. Tie proportion of benefits received by tie East Norti
Central, Souti Central, and Souti Pacific regions declines in favor of tie
remaining areas.
Tie very large benefits reported in Tables 6-8 tirougi 6-13 are
consistent witi tie small marginal effects observed in Table 6-7. While
tie responsiveness of eaci individual's real earnings to cianges in tie
ambient levels of particulate matter is iigily inelastic, tie effect on
real earnings of non-marginal cianges in particulate matter concentrations
cumulated over tie entire labor force is considerable. For example, assume
for tie moment tiat a decrease in TSP levels by 1 ug/m will cause tie real
wage rate to fall by $0.005 per iour, or by $10.40 per year. A very large
decline in ambient levels of particulate matter — say about 100 jig/m —
could lead to a decline in real annual earnings of rougily $1,000. If a
regional labor force of 10,000 individuals is assumed, tien tie total
benefits at tie attainment date alone will amount to some $10 million.
Tiis figure will, of course, grow as tie labor force expands and as tie
effects of improved air quality are felt in tie future.
Tie benefits estimates presented in Tables 6-8 tirougi 6-13 are based
on tie assumption tiat all affected counties attain tie standard in tie
implementation year and maintain it tirougi 1995. As noted in Section 9,
tiese are referred to as "B" scenarios. An alternative "A" scenario was
6-49
-------
also considered. In this scenario, the appropriate control strategies used
to implement the standards may not bring all counties into attainment.
This can occur because available control options are exhausted prior to
standard attainment. In order to test the sensitivity of these benefits
estimates to the attainment assumption, an alternate set of benefits
estimates for the -most lax PM10 standard (PH10 70 AAM/250 24-hour expected
value) are presented in Table 6-14. The assumption that all counties
comply with the standard is relaxed, i.e., residual nonattaimnent counties
are not forced into attainment.
Table 6-14 should be compared with Table 6-8, where all counties are
assumed to be in compliance with the same PM10 standard. As expected, the
benefits estimates in Table 6-8 exceed those shown in Table 6-14.
Benefits estimates for the remaining particulate matter standards when
the assumption of full compliance is relaxed are presented in Section 11.
CONCLUDING
In this section, the results reported in two hedonic wage models
[Rosen (4) and Smith (8)] are used to estimate the benefits which result
from the attainment of three alternative primary standards for ambient
levels of particulate matter. Point estimates of these benefits* range
from $73.5 billion under the most stringent standard (the current TSP
secondary standard) to about $19.8 billion for the most lax PH10 primary
standard.
The marginal willingness- to-pay for air quality calculated from the
data is quite reasonable, ranging from about $3.30 per year to over $11 per
year. The very large benefits reported in Tables 6-8 through 6-13 reflect
two factors:
* Calculated as the discounted present value of a stream of benefits over a
7-year (9-year) horizon from an attainment date of 1989 (1987), expressed
in 1980 dollars in 1982.
6-50
-------
Table 6-14
ESTIMATED BENEFITS FOR: HEDONIC WAGE STUDIES
Benefits Occurring Between 1989 and 1995
Scenario: Type A PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
0.0
10.0
422.0
458.2
2267.8
829.2
136.4
364.8
2073.4
178.6
6740.4
Point
Estimate
0.0
20.3
852.3
925.6
4580.8
1675.0
275.6
736.9
4188.1
360.8
Maximum
0.0
40.1
1359.5
1553.4
10830.1
3278.4
485.1
1088.4
6064.2
600.0
13615.3 25299.1
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
2698.0
5449.9 10126.7
6-51
-------
The effect of larger (i.e., non-marginal) improvements in
air quality on real annual earnings.
The extrapolation of individual willingness-to-pay over the
entire labor force in the affected areas.
The benefits estimates cited above are derived from the application of
the estimated relationship between real earnings and particnlate matter
levels reported in Smith (8) and Rosen (4) to a set of data obtained from
other sources. The assumptions which are required in order to adapt these
models and data to the procedure used in the benefits calculations may bias
these calculations. Some cautionary remarks relating to possible sources
of bias are summarized in Table 6-15.
REFERENCES
1. Cropper, H. L. and A. S. Arriaga-Sal inas. Inter-City Wage
Differentials and the Value of Air Quality. Journal of Urban
Economics, 8:236-254. 1980.
2. . Methods Development for Assessing Air Pollution Control
Benefits. Vol. IV, EPA 600/5-79-001/d, February 1979.
3. Rosen, Sherwin. Hedonic Prices and Implicit Markets. Journal of
Political Economy, 82(l):34-35. January/February 1974.
4. . Wage-Based Indices of Urban Quality of Life. In: Current
Issues in Urban Economics, P. Mieszkowski and M. Straszheim (eds.).
Baltimore: Johns Hopkins Press, 1979.
5. Thaler, R. and S. Rosen. The Value of Saving a Life: Evidence from
the Labor Market. In: Household Production and Consumption. N. E.
Terleckyj, ed.. New York: NBER, 1975.
6. Lucas, R. Hedonic Wage Equations and Psychic Wages in the Returns to
Schooling. American Economic Review, 67(3):549-558. September 1977.
7- Roback, J. The Value of Local Urban Amenities: Theory and
Measurement. Unpublished Doctoral Dissertation, University of
Rochester, 1980.
8. Smith, V. K. The Role of Site and .Job Characteristics in Hedonic Wage
Models. Forthcoming in Journal of Urban Economics.
9. Viscusi, W. Wealth Effects and Earnings Premiums for Job Hazards.
Review of Economics and Statistics, 60(3):408-416. August 1978.
6-52
-------
Table 6-15
SUMMARY OF BIASES IN BENEFITS CALCULATIONS: HEDONIC WAGE MODELS
Source of Bias
Direction of Bias
Hedonic wage models can only measure
perceived health, and welfare effects
of improved air quality.
Benefits underestimated
since unperceived health or
welfare effects omitted.
Formula for benefits calculations based
on procedure suggested by Freeman (28)
that assumes marginal willingness to
pay for air quality declines linearly
from initial concentration of ambient
PM to zero as PM is abated.
Cannot be predicted since
the true supply price
function for particulate
matter is not observed.
TSP may be highly correlated with other
site characteristics (including other
pollutants) which are omitted from
model. Est. coefficient on TSP may
also include some of the effect of the
omitted variables on real earnings.
Cannot be predicted in
the absence of additional
information on the
omitted site characteris-
tics and their correla-
tion with TSP.
Benefits are extrapolated over entire
labor force while the coefficients on
TSP were estimated over specific sub-
samples of the labor force.* It is
assumed that unemployed individuals
place the same value on improved air
quality as employed individuals.
To the extent that
unemployed individuals
place a lower (higher)
value on air quality,
benefits are over-
estimated (under-
estimated) .
Certain occupations not included in
calculation of county real wage rate.**
Unknown.
* Rosen (4) used data defined over employed males while Smith (8)
estimated his model over all employed individuals (both male and
female). There is some evidence that the response of real wage to
changes in air quality does not vary by sex (8). However, neither study
can measure willingness-to-pay for air quality among the unemployed,
although Smith shows that this sample truncation does not seriously bias
his coefficient estimates.
** These data exclude self-employed individuals, railroad, farm and
domestic-service workers, and state and municipal employees.
6-53
-------
10. Hamermesh, D. Economic Aspects of Job Satisfaction. In: Essays in
Labor Market and Population Analysis, Ashenfelter and W. Gates, eds.
New York: Wiley, 1977.
11. Olson, Craig A. An Analysis of Wage Differentials Received by Workers
on Dangerous Jobs. Journal of Human Resources, 16(2): 167-185. Spring
1981.
12. Brown, C. Equalizing Differences in the Labor Market. Quarterly
Journal of Economics, 94(374) :113-134. February 1980.
13. Meyer, J. and R, Leone. The Urban Disamenity Revisited. In: Public
Economics and the Quality of Life, L. Wingo and A. Evans, eds.
Baltimore: Johns Hopkins Press, 1977.
14. Getz, M. and Y. Huang. Consumer Revealed Preference for Environmental
Goods. Review of Economics and Statistics, 60(3) :449-458. August
1978.
15. Israeli, 0. Differentials in Nominal Wages and Prices. Unpublished
doctoral dissertation. University of Chicago, 1973.
16. Hoch, Irving. Wages, Climate and the Quality of Life. Journal of
Environmental Economics and Management, l(4):268-295. December 1974.
17. Smith, Robert A. Compensating Wage Differentials and Public Policy:
A Review. Industrial and Labor Relations Review, 32:339-352. April
1979.
18. Izraeli, 0. The Effect of Environmental Goods and City Size on
Earning Levels and Housing Values Across SMSAs: Empirical Evidence.
Oakland University, Rochester, Michigan (presented at the Western
Economics Association Meeting, San Francisco, California, July 1981).
25 pp.
19. National Academy of Science. Air Quality and Automobile Emission
Control: Volume 4 - The Costs and Benefits of Automobile Emission
Control. Prepared for the Committee on Public Works, United States
Senate, Washington, D.C., U.S. Government Printing Office, September
1974. 470 pp.
20. Mathtech, Inc. Benefits Analysis of Alternative Secondary National
Ambient Air Quality Standards for Sulfur Dioxide and Total Suspended
Particulates. Prepared for the U.S. Environmental Protection Agency
(July 1981). Draft Final Report.
21. Dagenais, M. The Use of Incomplete Observations in Multiple
Regression Analysis. Journal of Econometrics, 1:317-328. 1973.
22. Cebula, Richard J. Determinants of Geographic Living-Cost
Differentials in the United States: An Empirical Note. Land
Economics, 56:477-481. November 1980.
6-54
-------
23. Cebula, Richard J. and Lisa Karen Smith. An Exploratory Empirical
Note on Determinants of Inter-Regional Living-Cost Differentials in
the United States, 1970 and 1975. Regional Science and Urban
Economics, 11:81-85. 1981.
24. Heckman, J. The Common Structure of Statistical Models of Truncation,
Sample Selection and Limited Dependent Variables and a Simple
Estimator for Such Models. Annals of Economic and Social Measurement,
5:475-492. Fall 1976.
25. . Sample Selection Bias as a Specification Error. Econometrica,
47(1):153-162. January 1979.
26. Gronau, R. Wage Comparisons — a Selectivity Bias. Journal of
Political Economy, 82(6):1119-1143. November/December 1974.
27. Wales, T. and A. Woodland. Sample Selectivity and the Estimation of
Labor Supply Functions. International Economic Review, 21:437-468.
June 1980.
28. Olsen, R. J. Approximating a Truncated Normal Regression with the
Method of Moments. Econometrica, 48(5): 1099-1106. July 1980.
29. Freeman, A. Myrick. The Benefits of Environmental Improvement.
Baltimore: Johns Hopkins University Press, 1979.
30. Bureau of Economic Analysis. U.S. Department of Commerce News. BEA
80-74,'December 9, 1980.
31. U.S. Bureau of the Censns. County Business Patterns, 1978.
32. U.S. Bureau of Economic Analysis. OBER-BEA Regional Projections, Vol.
Ill, SMSAs. July 1981.
33. U.S. Bureau of Economic Analysis. Projections of Population 1976-
2000. March 23, 1981.
6-55
-------
APPENDIX
-------
Table 6A-1
HEDONIC WAGE MODEL ESTIMATED BY ROSEN (4):
EFFECTS OF INDIVIDUAL-SPECIFIC CHARACTERISTICS
Explanatory
Variable
Race
(white = 1)
Head of Household
(yes - 1)
Married
(yes = 1)
Employed Full Time
(yes = 1)
Work 35 Hours/Week
or Less
(yes = 1)
Self- Employed
(yes = 1)
Government Employee
(yes = 1)
Never unemployed
During the Year
(yes = 1)
Unemployed Once
During the Year
(yes = 1)
Veteran
(yes = 1)
Sales
(yes = 1)
Crafts
(yes = 1)
Coefficient
0.150*
0.160*
0.085*
0.597*
-0.142*
0.160*
0.125*
0.050
-0.068
0.050*
-0.198*
-0.157*
Explanatory
Variable
Operatives
(yes = 1)
Laborer
(yes = 1)
Service
(yes = 1)
Durable Goods
(yes = 1)
Nondurables
(yes = 1)
Transport
(yes = 1)
Trade
(yes = 1)
Other Service
(yes = 1)
Public Administra-
tion
(yes = 1)
Education
Experience
t\
(Experience)
Log Weeks Worked
R2
Coefficient
-0.263*
-0.270*
-0.302*
-0.086*
-0.105*
-0.074*
-0.165*
-0.186*
-0.039
0.048*
0.015*
-0.0002*
0.835*
0.337
* Coefficient has a t-statistic >, 2.
6-57
-------
Table 6A-2
HEDONIC WAGE MODELS ESTIMATED BY ROSEN (4):
EFFECTS OF SITE-SPECIFIC CHARACTERISTICS
Explanatory Variables
TSp(a)
Inversion Days
Water Pollution*3*
Rainy Days(a) (No.)
Sunny Days(a) (No.)
Temperature 90°F+'a'
(Number)
Crime Rate(b)
Unemployment Rate
Population Growth Rate
Population Density*13'
Population Level *c^
Live in Center City
(yes = 1)
Selected Regression Results: Estimated Coefficients
0.1060*
0.1220*
0.1070
-0.0943*
0.0681*
0.2860*
-0.0521
0.0457*
-0.0908*
0.0897*
0.0102*
-0.0868
-0.9620*
0.2390*
0.3160*
-0.0823*
0.1500*
0.0406
I
0.0270
0.4790*
-0.0910*
0.1100*
0.2250*
-0.1700
0.0539*
0.5330*
-0.0863*
0.0692*
0.2300*
0.1200
-0.1100*
0.0966*
-0.0960*
0.0602*
0.2720*
0.0614
0.0149
-0.0998*
0.0899*
-0.0950*
0.0553*
0.2950*
NA
0.0926
0.0095
-0.1120*
0.0967*
-0.9950*
0.0921*
0.1740*
0.0492
0.1950*
-0.0899*
0.1040*
-0.0864*
01
00
a Coefficient scaled up by 100.
* Coefficient has a t-statistic
NA = Not reported.
2.
Coefficient scaled up by 10,000.
Coefficient scaled up by 10,000,000.
-------
I
Ul
Table 6A-3
HEOONIC WAGE MODELS ESTIMATED BY SMITH (8)*
Explanatory Variable
Intercept
Education
(no. of years)
(Education)2
Experience
(age - education - 6)
(Experience)2***
Race
(white = 1)
Sex
(male = 1)
Veteran
(yes = 1)
Unemployment Rate
Full
Sample
0.3411
(6.15)
0.0244
(3.88)
0.0013
(5.05)
0.0261
(32.44)
-0.0455
(-26.62)
0.0560
(5.85)
0.1663
(17.59)
0.0749
(7.80)
-0.0138
(-5.59)
Hale
0.6512
(8.98)
0.0308
(4.06)
0.0010
(3.30)
0.0309
(25.67)
-0.0532
(-22.26)
0.1118
(8.66)
—
0.0359
(3.60)
-0.0208
(-6.47)
Female
0.2021
(2.30)
0.0283
(2.60)
0.0009
(2.13)
0.0181
(15.94)
-0.0301
(-11.91)
-0.0261
(-1.86)
—
—
-0.0051
(-1.38)
Full
Sample
0.4038
(6.93)
0.0234
(3.72)
0.0013
(5.22)
0.0263
(32.60)
-0.0458
(-26.76)
0.0026
(0.13)
0.1662
(17.58)
0.0748
(7.80)
-0.0138
(-5.59)
Hale
0.6443
(8.24)
0.0304
(4.02)
0.0010
(3.30)
0.0310
(25.78)
-0.0535
(-22.41)
0.0734
(2.55)
—
0.0363
(3.65)
-0.0206
(-6.40)
Female
0.2328
(2.53)
0.0254
(2.34)
0.0011
(2.40)
0.0182
(16.05)
-0.0303
(-12.00)
-0.0329
(-1.10)
—
—
-0.0050
(-1.35)
(continued)
-------
Table 6A-3 (Continued)
i
o^
o
Explanatory Variable
Professional
(yes = 1)
Manager
Sales
Clerical
Craftsman
Operative
Transport Equipment
Operator
Labor (Non-farm)
Service
Full
Sample
0.3471
(16.67)
0.3740
(17.32)
0.1491
(6.52)
0.2001
(10.15)
0.2646
(12.26)
0.0780
(3.62)
0.1225
(4.68)
0.0776
(3.25)
-0.0098
(-0.48)
Male
0.0871
(2.79)
0.1411
(4.45)
-0.0019
(-0.05)
-0.1009
(-3.06)
0.0171
(0.54)
-0.1465
(-4.40)
-0.1180
(-3.35)
-0.1288
(-3.81)
-0.2533
(-7.77)
Female
0.5650
(19.21)
0.5228
(16.19)
0.2010
(6.36)
0.3924
(15.45)
0.4495
(8.76)
0.2375
(8.18)
0.3694
(5.38)
0.2004
(3.99)
0.1702
(6.40)
Full
Sample
0.3499
(16.80)
0.3741
(17.33)
0.1479
(6.47)
0.1994
(10.11)
0.2625
(12.14)
0.0741
(3.44)
0.1236
(4.72)
0.0767
(3.21)
-0.0100
(-0.49)
Male
0.0903
(2.89)
0.1426
(4.50)
-0.0011
(-0.03)
-0.0975
(-2.96)
0.0184
(0.58)
-0.1496
(-4.50)
-0.1134
(-3.22)
-0.1293
(-3.82)
-0.2488
(-7.64)
Female
0.5678
(19.32)
0.5184
(16.05)
0.1981
(6.27)
0.3911
(15.39)
0.4432
(8.65)
0.2244
(7.69)
0.3587
(5.22)
0.1872
(3.72)
0.1683
(6.33)
(continued)
-------
Table 6A-3 (Continued)
Explanatory Variable
Injury Rate (BLS)
Cancer Exposure Index
TSp(a)
(ug/m3)
Household Head
Union
OJT ' Experience
Crime Rate(b)
Sun (mean annual % of
possible sunshine)
Dual Job-Holder
Full
Sample
0.0114
(12.87)
0.0219
(2.76)
0.0871
(3.88)
0.1570
(18.08)
0.1832
(22.32)
-0.0012
(-0.98)
0.0943
(4.60)
-0.0015
(-2.62)
-0.0439
(-2.28)
Hale
0.0113
(10.65)
0.0283
(2.77)
0.1120
(3.85)
0.2318
(16.96)
0.1730
(17.09)
-0.0022
(-1.60)
0.0782
(2.94)
-0.0021
(-2.79)
-0.0408
(-1.71)
Female
0.0117
(7.66)
0.0089
(0.72)
0.0675
(1.97)
0.0694
(6.02)
0.1857
(13.47)
-0.0010
(-0.40)
0.1007
(3.24)
-0.0002
(-0.25)
-0.0256
(-0.82)
Full
Sample
0.0037
(1.47)
0.0231
(2.91)
0.0830
(3.70)
0.1641
(10.90)
0.1174
(7.01)
-0.0017
(-1.38)
0.0954
(4.66)
-0.0014
(-2.47)
-0.0426
(-2.21)
Hale
0.0122
(3.39)
0.0290
(2.84)
0.1084
(3.73)
0.3035
(11.69)
0.1035
(4.70)
-0.0022
(-1.65)
0.0797
(3.00)
-0.0020
(-2.71)
-0.0400
(-1.67)
Female
0.0100
(2.36)
0.0100
(0.81)
0.0615
(1.80)
0.1106
(4.74)
0.0971
(3.25)
-0.0021
(-0.82)
0.1007
(3.25)
-0.0001
(-0.13)
-0.0228
(-0.73)
(continued)
-------
Table 6A-3 (Continued)
Explanatory Variable
Workers' Knowledge of
Cancer Hazard
Injury Rate * Race
Injury Rate ' Head of
Household
Injury Rate * Union
R2
S
N
Full
Sample
4.3032
(6.01)
0.460
0.171
16.199
Hale
3.8789
(4.70)
0.462
0.160
9,105
Female
5.7079
(4.22)
0.322
0.173
7.094
Full
Sample
4.1518
(5.79)
0.0070
(3.01)
-0.0001
(-0.67)
0.0078
(4.53)
0.461
0.171
16.199
Hale
3.8538
(4.66)
0.0044
(1.49)
-0.0084
(-3.24)
0.0073
(3.55)
0.464
0.160
9.105
Female
5.52
(4.07)
0.0012
(0.30)
-0.0064
(-2.06)
0.0133
(3.76)
0.323
0.173
7.094
* t-statistics are given in parentheses.
a Coefficients are scaled up by 100.
Coefficients are scaled up by 100,000.
-------
SECTION 7
ECONOMIC BENEFITS OF REDUCED SOILING
-------
SECTION 7
ECONOMIC BENEFITS OF KEDUCED SOILING
INTRODUCTION
Overview
Soiling is the accumulation of particulate matter on the surface of an
exposed material. This accumulation leads to changes in the quality of
reflectance or transparency for materials such as painted surfaces and
glass so that these materials "look dirty" [Beloin and Haynie (1)]. In
addition, textiles can weaken and become faded with exposure to abrasive
particles, sunlight, and other pollutants [Criteria Document (2)]. The
discoloration of fabrics is also considered to be a soiling effect.
With the implementation of PH control programs, particle deposition
will decrease, and households and firms may find that'they can maintain
desired levels of cleanliness with fewer resources (e.g., labor and
materials) than previously required. As a consequence, less time and money
would have to be devoted to the variety of cleaning tasks usually under-
taken. Appropriate identification of these cost savings provides a measure
of the benefits of reduced soiling.
Two major questions are addressed in this section:
• Can a consistent method be identified for calculating the
economic benefits of increased cleanliness due to reduced
ambient levels of particulate matter (PM)?
e What are the monetary benefits of reduced soiling, given
the alternative PM10 (i.e., PM <. 10 urn) and TSP scenarios
described in Section 9?
7-1
-------
Much of this section focuses on the first of these questions. Over
the past two decades, a number of studies have been completed which attempt
to assess the costs associated with PM-related soiling damages. Our objec-
tive is to review the basic approaches utilized in these studies and to
assess their suitability for the calculation of benefits. No new research
is attempted as part of this effort. The evaluation of alternative studies
is based on how well the studies conform to certain criteria which reflect
desirable traits of a benefits analysis. These criteria include issues
such as the methodological soundness of the study, consideration of all
relevant variables, and the legitimacy of the empirical analysis.
Most of the studies which examine the benefits of reduced soiling
focus on the household sector. While none of the household studies
reviewed here completely satisfies all the criteria we identify as being
relevant, a class of studies, the behavioral models, are sufficiently well-
specified to warrant the calculation of benefits. In our judgment, the
necessary analytical tools have been developed to the point where reason-
able estimates of benefits from reduced soiling in the household sector can
be derived.
While other sectors of the economy have received less emphasis, there
have been several attempts to identify benefits from reduced soiling for
specific industries in the manufacturing sector and for commercial opera-
tions associated with cleaning services. Of these studies, only the
industry analysis, which examines the impact of changes in PM on production
costs, provides a basis for the calculation of benefits. The results of
the non-household sector analyses will also be reviewed in this section.
Scope of A*1*lysis
Evaluation Criteria —
The principal objective of this study is to review the analyses which
have attempted to measure the benefits of reduced soiling and to assess the
soundness of the estimates. To provide a consistent method for evaluating
7-2
-------
the alternative studies, several criteria, which, are important for the
successful completion of any benefits analysis, are identified. These
criteria help identify where individual analyses may be particularly weak
or strong, and are useful in comparing the plausibility of benefit
estimates generated from different studies. Thus, the criteria form a set
of conditions by which the range of studies to be considered can be
narrowed.
Five general criteria are used to evaluate critically the various
studies which examine the benefits from reduced soiling. These criteria
include:
• Theoretical basis for the study — Is the study's technical
approach consistent with the requirements for calculation
of economic benefits as prescribed in theoretical welfare
economics?
• Consideration of relevant variables — Has the study
accounted for all relevant variables (e.g., economic,
social, pollution, climate) that may influence cleaning
decisions?
• Quality of input data — Does the study use data consistent
with the variables of the theoretical model and are these
data accurately measured?
• Legitimacy of the empirical analysis — Were the empirical
techniques used in the study appropriate and properly
applied, and were the results correctly interpreted?
• Transferability of models and/or results — Can the study
be used as a basis for estimating benefits of reduced
soiling given the implementation of various PM control
strategies?
Although these criteria are fairly stringent, they represent the types of
conditions that must be satisfied if defensible, quantitative estimates of
benefits are to be derived. In the next major subsection, these criteria
are used to assess strengths and weaknesses of the various studies.
7-3
-------
Coverage of Studies —
An important issue that arises in the development of benefit estimates
is the degree of coverage associated with the alternative studies. As
mentioned above, the literature concerned with the benefits of reduced
soiling has focused primarily on household responses. Clearly, many other
sectors of the economy may also be affected. Commercial establishments,
industrial plants, and government facilities are all likely to experience
less soiling as a result of reduced levels of ambient PH. In this section,
benefits are calculated only for the household sector and two industries in
the manufacturing sector — SIC 344 (Fabricated Structural Metal Products)
and SIC 354 (Metalworking Machinery). Although a limited number of studies
have been designed and undertaken for other sectors of the economy, to date
these efforts have not found a statistically significant association
between PM and cleaning-related cost data. Consequently, no benefits are
calculated for these other sectors. If these other sectors of the economy
do derive benefits from reduced soiling, then the estimates of benefits
developed in this section should be considered conservative estimates of
total benefits from reduced soiling.
Pollutant Measures —
The development of physical and economic damage functions for use in
benefits analysis requires that specific assumptions be made about measures
of air quality. At least three questions arise in the course of designing
a study, estimating damage relationships, and calculating benefits. These
questions are:
• What is (are) the most appropriate pollution variable(s)
for the model?
• What variable(s) is (are) actually used by the study?
• Will the non-marginal improvements in ambient air quality
that are implied by poll.ution controls change the
composition of PM and consequently alter physical effects
and/or behavioral responses?
7-4
-------
These questions involve issues of both a physical and statistical
nature. The issues related to the physical dimension include particle
size, deposition rates, and composition. These three attributes of PM help
define the soiling potential of the pollutant. The statistical issues
include choice of averaging time and definitions of spatial and temporal
indices.
The questions posed above are directly addressed in a later
subsection. Here, we note that particle size plays a prominent role in the
analysis since alternative standards are stated in terms of TSP and PM10.
However, with either type of standard, data are available for both particle
size divisions.* Consequently, benefit calculations for studies which rely
on TSP as a measure of air pollution can be conveniently carried out in the
units of the original study, even when the standard is stated in terms of
PM10. In this case, the estimates represent the benefits of the TSP
reduction that results from PM10 controls.
Based on the review in this section, national benefits for reduced
soiling are estimated for the household sector and a part of the manu-
facturing sector. Benefit estimates for the household sector are developed
from three studies: Cummings et al. (5), Watson and Jaksch (6), and
Mathtech (7). Benefit estimates for the manufacturing sector are derived
from a single study, Hathtech (8).
Tables 7-1 and 7-2 summarize the benefit estimates of reduced soiling
for six alternative ambient air quality standards examined in the report.
Table 7-1 reports the benefits associated with three PM10 primary standards
and the current TSP primary standard. Table 7-2 shows the benefits
associated with two alternative secondary standards. The values in Table
7-2 represent the total benefits of moving from pre-control to post-control
* These data are generated under the assumption that the base year
relationship of ambient PM10 to TSP is fixed at 0.55 (4).
7-5
-------
Table 7-1
SUMMARY OF ESTIMATED BENEFITS FOR ALTERNATIVE PRIMARY STANDARDS*
Standard
PM10 70/250
PM10 55/250
PM10 55/150
TSP 75/260
Sector
Household
Manufacturing
Household
Manufacturing
Household
Manufacturing
Household
Manufacturing
Benefits
Minimum
0.73
0.73
1.30
1.32
1.62
1.48
2.41
2.43
Point
3.14
1.30
5.68
2.41
7.16
2.86
10.74
4.80
Maximum
13.85
9.45
25. 12
20.51
32.76
26.24
49.90
40.75
* Discounted present values in 1982 in billions of 1980 dollars.
Table 7-2
SUMMARY OF ESTIMATED BENEFITS FOR ALTERNATIVE SECONDARY STANDARDS*
Standard
PM10 55
TSP 150
Sector
Household
Manuf actur ing
Household
Manufacturing
Benefits
Minimum
1.29
1.32
3.15
2.81
Point
5.64
2.41
14.21
5.81
Maximum
25.53
20.25
66.36
55.13
* Discounted present values in 1982 in billions of 1980 dollars.
7-6
-------
concentration levels. To ascertain the incremental benefits of the
secondary standards, given that the primary standards are attained, it is
necessary to net out the benefits associated with the relevant primary
standard. In Table 7-2, the 55 PM10 secondary standard should be compared
with the 70/250 PM10 primary standard, while the 150 TSP secondary standard
should be compared with the 75/260 TSP primary standard.
The benefit numbers reported here represent total U.S. benefits. The
estimates are reported as discounted present values in 1982 in billions of
1980 dollars. The real discount rate is assumed to be equal to 10 percent,
and the estimates are developed over a horizon beginning at the start of
1987 (TSP) or 1989 (PM10) and continuing through to the end of 1995. For
the estimates reported here, it is assumed that all counties are in attain-
ment with the relevant standard over the entire horizon. In the body of
the report, estimates are also derived when not all counties are assumed to
be in compliance throughout the time period. This latter scenario is
possible if available means of controlling emissions do not allow for a
control level sufficient for standard attainment throughout the time
period.*
Total benefits from reduced soiling in the household sector and a
subset of the manufacturing sector are estimated to range from $1.46 to
$23.30 billion for the most lenient primary standard (70/250 PM10), and
from $4.84 to $90.65 billion for the most stringent primary standard (TSP
75/260). An important point to note is that even though the primary
standards are typically associated only with health effects, substantial
welfare benefits are also derived with attainment of the primary standards.
For the two ranges cited above, the resultant point estimate of the
benefits associated with moving to a secondary standard, given that the
corresponding primary standard is attained, are $3.61 and $4.48 billion for
PM10 and TSP, respectively.
* This latter scenario is equivalent to the Scenario A described in Section
9.
7-7
-------
Although the numbers reported in Table 7-1 and 7-2 are large in magni-
tude, on a per-household or per-plant basis, they appear intuitively
q
reasonable. For example, the per-household benefit of a 1 jig/m reduction
in TSP is estimated to range from about $0.40 to $11.50 per year. Similar
calculations can be developed for the manufacturing sector. The benefit
associated with a 1 jig/m^ reduction in TSP for SIC 354 is estimated to be
approximately $260 per plant per year. This amounts to about 0.019 percent
of the total production cost for an average plant in the industry. For SIC
344, these numbers are $610 per plant per year, which is about 0.031
percent of the total cost for an average plant in the industry.
The estimates reported in Tables 7-1 and 7-2 are subject to certain
qualifications. The major factors that limit the generality of the results
reported above include:
All possible benefit categories (sectors) are not included
in the analysis. (Leads to an underestimate of benefits.)
The benefit estimates are sensitive to a variety of model
and extrapolation biases. These are discussed in detail in
the report. (Benefits affected in an unknown direction.)
The air quality data used in the benefits analysis relies
primarily on design value monitor readings. This implies
that a correction factor is required in some of the studies
to better characterize population and facility exposure.
(Impacts benefits in an unknown manner.)
The physical effects and behavioral responses observed for
a given composition of PH do not change when pollution
controls are applied. (Affects benefits in an unknown
direction.)
Air quality is assumed to improve only in the subset of
counties included in the air quality data file. (Leads to
an underestimate of benefits.)
These limitations, as well as others that are specific to particular
studies, are discussed in the body of the report. While it is difficult to
assess whether the various limitations lead to an under- or overestimate of
benefits, there has been a conscious effort to adopt a more conservative
assumption when warranted. As used here, a conservative assumption is one
7-8
-------
that leads to benefit estimates that are lower in. magnitude relative to
estimates associated with a competing assumption. On the basis of our
review and critique of studies, our evaluation of biases, and the develop-
ment of ranges of benefits to reflect uncertainty, we believe the point
estimates reported in this section represent conservative estimates of the
total benefits from reduced 'soiling.
MODELS OF BENEFITS FROM REDUCED SOILING
Introduct ion
The object of an analytically proper analysis of benefits from reduced
soiling is to provide an estimate of the income-equivalent change in
welfare for a specified improvement in some measure of PM. This subsection
describes the general approaches that have been used to estimate this value
and assesses the success individual studies have had in obtaining benefits
estimates.
As noted in the Criteria Document (2), there are several relationships
that must be formalized if an acceptable benefits analysis is to be con-
ducted. These relationships start with the transformation function between
emissions and ambient concentrations, and conclude with the specification
of the relationship between economic damage and the benefits associated
with reduced soiling. In this study, the factors that translate emissions
into ambient concentrations are taken as given. Therefore, the driving
force of the analysis is the expected change in ambient concentration
levels associated with the implementation of alternative air quality
standards.
Figure 7-1 shows the linkages between ambient levels of PM and the
realization of economic benefits. As ambient levels of PM decrease, it is
expected that there will be reductions in observed physical effects (e.g.,
dirtiness of windows), which ultimately translate into reduced effort
required to maintain a given level of cleanliness. In turn, this reduced
effort means a cost savings to households and firms. Under certain
7-9
-------
Ambient levels of PM
i
Physical damage
Physical damage functions
i
Economic damage functions
Economic benefits of
reduced soiling
Figure 7-1. Processes leading to economic benefits
conditions, the cost savings can be interpreted as the amount that house-
holds or firms are willing to pay to have the improved air quality rather
than the air quality that existed prior to the PM reduction. In this case,
the cost savings are an estimate of economic benefits.
This subsection examines the data and methods that have been used in
completing the steps identified in Figure 7-1. The discussion focuses
initially on the air quality measures that are most relevant for an
analysis of soiling damage. Following this, we begin our evaluation of the
various models which have been developed to measure economic benefits of
reduced soiling. In general, there are two distinct classes of models:
7-10
-------
the physical damage function models and the behavioral models. Each of
these model types is evaluated in terms of the criteria outlined earlier,
and individual studies within each class are identified. A detailed review
of selected individual studies is also presented to identify those studies
best suited for the calculation of benefits.
PM Measures and Soiling
As mentioned earlier, several questions relating to appropriate
measures of air quality arise in the course of designing and completing an
air quality benefits analysis. The questions reflect concern about both
physical and statistical characteristics of pollutants.
With respect to the physical characteristics, the Staff Paper for PM
(3) asserts that the soiling of textiles and vertical surfaces is generally
associated with fine particle deposition. On the other hand, horizontal
surfaces are considered to be more susceptible to deposition of large
particles. However, it is also pointed out in Reference (3) that the
direct relation between increased soiling potential and increased particle
size may be mitigated by lighter color of coarse mode particles, smaller
transport distances, and lower penetration rates of larger particles to
indoor surfaces. Because of these offsetting factors, no concensus has
been reached on the most appropriate particle size division to use in an
analysis of horizontal surface soiling.
With respect to the statistical issues, it is a commonly held belief
that long-term measures of PH concentrations such as annual means are most
representative of soiling damage. However, this appears"to be based only
on qualitative evidence. The real test of this issue involves an analysis
of behavioral responses to long-term as well as episodic PM measures. In
fact, one of the studies examined later in this section finds that a
measure of the 24-hour second highest reading is statistically the most
robust measure in the economic function that is estimated.
7-11
-------
The other statistical issues that arise in defining air pollution
measures involve definitions of indices. Since air quality varies in time
and space, it is necessary to construct a summary index that provides a
single-valued representation of air quality in a given area. The
procedures used to develop such indices in the present study are described
in Section 9.
The above discussion indicates that no concensus has been reached on
the most appropriate pollution measure to use in a soiling study. Despite
this conclusion, it is the case that physical characteristics of the
pollutant will affect the degree of physical impact and the'extent of
economic response. Although qualitative observations link ambient particle
loadings and soiling, attempts to provide quantitative measures of
association have not directly addressed the issues of particle size,
deposition rates, and composition. In fact, the studies reviewed in this
section all use a statistical measure of ambient TSP. At issue then is
what biases may be introduced by using TSP in a statistical relationship
when other measures of particulate matter may be the factors that are
appropriate for the model posited.
This is a classic errors-in-variables problem. To determine what
biases may arise in such a situation, it is necessary to know the
underlying relationship between TSP and the "true" measure. If, for
example, TSP is always exactly two times greater than the true measure,
then the coefficient of TSP will be biased in inverse proportion to the
factor. In this case, the coefficient will be one-half the value of the
coefficient in the true model. However, it is not necessarily the case
that benefits will be understated as well. In fact, benefits will be
unaffected if the proposed model is linear and all observations exhibit the
same proportional relationship. With nonlinear models and/or more
complicated relationships between TSP and the true variable, it becomes
more difficult to define precisely the extent of bias and the ultimate
impact on benefits.
7-12
-------
Even if the estimation phase of the analysis is complicated using the
appropriate measure of pollution, problems can still arise once air quality
is allowed to change. With non-marginal changes in ambient PM, one would
expect the composition of PM to differ after the change. If this occurs
and behavioral responses are sensitive to the composition, then benefits
estiated from a model with observations for the pre-change composition may
be misstated. Since the composition of PH is not considered in the studies
reviewed here, the implicit assumption is that any changes in pollutant
composition will not alter the marginal responses.
Phsical
To date, many of the studies that have attempted to identify the
benefits of reduced soiling have relied on an analysis plan like that
portrayed in Figure 7-1. When each of the five processes shown in the
figure is a direct part of the analysis, the methodology for determining
benefits is typically classified as the damage function approach. In this
approach, the first step involves a determination of exposure for specific
materials at places with different levels of ambient particulate concentra-
tions. Exposure to PM is then related to an objective physical measure
appropriate to the material under review. For example, with glass, the
measure might be the loss of transparency. The relationship between
exposure and physical impact provides an estimate of the physical damage
function.
The foremost example of estimating physical damage functions for
soiling effects is Beloin and Haynie (1). In their study, damage functions
are developed from data collected for six types of building materials:
painted cedar siding, brick, limestone, concrete block, asphalt shingles,
and window glass. Data on PM concentrations and soiling effects were
collected over time for a cross-section of five controlled sites around
Birmingham, Alabama.
In the best specifications, they found that soiling, as measured by
reduced reflectance, could be stated as a function of PM levels and
7-13
-------
duration of exposure. Physical damage functions for acrylic emulsion paint
and white asphalt shingles are provided in Reference (1), and these speci-
fications are reproduced in Table 7-3. Note that the constant term in the
left-hand side variable of each regression equation represents an initial
reflectance value prior to exposure. Thus, the regressions can be inter-
preted as showing the expected change in reflectance for a given level of,
and exposure to, TSP.
Naturally, in the development of physical damage functions, care must
be taken to control for factors other than PM that may contribute to a
change in the physical status of the material. In soiling studies, clima-
tological parameters would seem to be especially relevant explanatory
factors. However, because the sites in Beloin and Haynie (1) are located
near one another, the expectation is that the microclimate is similar from
site to site. Nevertheless, Beloin and Haynie monitored temperature,
Table 7-3
REGRESSION RESULTS FROM BELOIN AND HAYNIE
Material
Acrylic Emulsion
Paint
White Asphalt
Shingles
Equation*
ln(92.5-y) = -0.311 + 0.345
+ 0.612
ln(41.8-y) = -4.881 +• 1.007
+ 0.595
In(TSP)
ln(t)
In(TSP)
ln(t)
y = Measured percent reflectance.
TSP = Annual geometric mean of total suspended particulates (jig/m3).
t = Exposure time in months.
* The R for the equations is 0.896 and 0.608, respectively. The number of
observations is 640 and 40, respectively.
Source: Table IV, Beloin and Haynie (1).
7-14
-------
rainfall, dew duration, and relative humidity to test for significant
differences in these data across sites. Their statistical analysis
indicated that the only difference among the variables was a higher level
of humidity at one of the sites. In turn, the higher humidity led to an
increase in mildew at the site. Despite this finding, none of the recom-
mended specifications in Reference (1) includes a measure of humidity.
Consequently, as noted in Reference (1), the increased mildew formation is
confounded with soiling by TSP.
Given relationships like those shown in Table 7-3, the next step in
the damage function approach links the PM—related physical impacts to a
measure of economic damage. Ideally, an assessment of economic damage
would be able to account for the alternative preventative or ameliorative
actions that may be taken by individuals in response to perceived physical
damage. In terms of household soiling, responses would likely take the
form of more frequent and/or more intense cleaning. However, preventative
measures such as filtered air conditioners may also be employed. As noted
in Reference (1), if an accurate assessment of soiling damage is to be
developed, it is important to consider how such decisions are made by
individuals. Unfortunately, the mechanism of choice is usually not con-
sidered directly in the physical damage function studies.
In addition to information on the types of responses that may be
observed, an analysis of soiling benefits via the damage function approach
must also consider the range of tasks and materials affected as well as the
unit value of the added cleaning costs.* Not only are these information
requirements burdensome, but the process of developing benefits estimates
with the physical damage function models implies some strong underlying
assumptions and limitations. For example, Waddell [(10), p. 24] mentions
that extrapolation of controlled study results to the real world ignores
the possibility of nonconstant marginal products, nonlinearities, and
problems of aggregation and substitution. Freeman (11) stresses the lack
* See Geomet (9) for an example of what is required to develop an inventory
of materials in a sulfates materials damage study. A similar effort
would also be necessary for a PM study.
7-15
-------
of consideration of consumer adjustments, and the difficulty in assessing
the inventory of exposed area for the multitude of surfaces that may become
soiled.
Evaluation of Physical Duwge Function Studies —
Many of these limitations can be related to the evaluation criteria
listed earlier.
Theoretical basis — Physical damage function studies
typically do not consider the adjustment opportunities that
are available to economic decision-makers. Estimates of
cost-savings reflect only changes in expenditures required
to return a material to a "clean" state and do not provide
accurate estimates of willingness to pay.
Consideration of relevant variables — Oftentimes physical
damage functions are determined from data collected in a
controlled laboratory setting. These conditions may not
adequately reflect the conditions in the real world. In
particular, climatic variables may act synergistically with
PH to increase the potential soiling effect.
Quality of input data — Most data used in these analyses
are micro in nature and are well-suited to the analysis of
specific materials. However, data must be collected for
many very specific tasks and/or materials. Furthermore,
estimates of exposed surface areas are usually very rudi-
mentary.
Legitimacy of the empirical analysis — The estimation of
physical and economic damage functions typically involves
simple linear regression techniques. Little ji priori
evidence is available on the most appropriate specification
for alternative damage functions. Hisspecification of the
damage functions can be especially important to a benefits
analysis when dealing with non—marginal changes in pollu-
tion levels.
Transferabilitv of models — Individual physical damage
functions can be adapted easily to a variety of alternative
settings. The major difficulties that arise are: the
definition of damage functions for each affected material;
and the estimation of the exposed inventory of each
material.
7-16
-------
The negative tone of the above evaluation is reinforced by discussions
in both the Criteria Document (2) and the Staff Paper (3). The Criteria
Document reports damage functions for the Beloin and Haynie (1) study
mentioned earlier, and offers an example of how repainting frequencies can
be obtained from their model for different PM levels. However, the conclu-
sion reached in the Criteria Document is that:
The least reliable of the "significant" damage functions are
those for soiling from particulate matter. The poorly under-
stood deposition rates and poorly characterized chemical and
physical properties make general application of the functions
difficult, if not impossible.
The Criteria Document goes on to say:
The limitations of these and other physical damage functions
hinder accurate estimates of total material damage and soiling.
Coupled with these limitations is the lack of material exposure
estimates. These problems presently preclude complete and
accurate estimates of the costs of damage based on a physical
damage function approach.
In our judgment, we agree with the assessment that currently available data
limit the usefulness of calculating national benefits from reduced soiling
when physical damage functions are required.* However, as will be seen in
the discussion below, we also believe that recent research based on
behavioral models is sufficiently well-defined to warrant the calculation
of benefits.
Although the data requirements of the physical damage function
approach may preclude its widespread use in estimating national benefits of
reduced soiling, the research embodied in the development of specific
damage functions is very important. In general, there is no & priori
economic reason for believing that a particular air quality variable
affects specific economic choices. As a consequence, the behavioral
* Lodge .et al. (12) reach similar conclusions in their review of non-health
effects of PM.
7-17
-------
approach (described below) for estimating the benefits of reduced soiling
must rely on information like that available in physical damage studies to
narrow the range of alternative specifications to be considered.
Behavioral Models of Reduced Soiling Benefits
An alternative approach to the physical damage function studies is the
analysis of benefits based on behavioral models. Referring back to Figure
7-1, it can be seen that a direct link between "ambient levels of PM" and
"economic damage functions" is also indicated. In this case, the steps
dealing with physical soiling effects are bypassed. The rationale behind
this way of characterizing the benefits process is that damages from
soiling may eventually be manifested in economic choices. As soiling
damage effects are perceived by households and firms, they respond by
purchasing goods that help to maintain a desired level of cleanliness. In
deciding how much of their available budget to allocate to cleaning-type .
goods, they indicate their willingness to pay for units of cleanliness.
Thus, willingness to pay for PH reductions may be observed in the economic
choices made by households and firms.
Within the class of behavioral models, there are several approaches
which can be distinguished. These include:
• Property value studies.
• Surveys of frequency and expenditures for cleaning activi-
ties.
• Economic demand and/or supply models.
Property Vmlne Studies —
The hypothesis underlying the property value studies is that
structures in areas with relatively better air quality will have higher
values than similar structures in relatively dirty areas. In effect, much
like a swimming pool or an enclosed garage adds to the value of a home,
good air quality is also a positive attribute. Many studies have been
7-18
-------
completed which employ this approach, yet the Criteria Document (2) and
Staff Paper (3) provide only a limited discussion of the technique. The
principal concerns identified in these reports are that:
• It is difficult to identify separately the benefits of
reduced soiling versus other benefits such as health and
aesthetics.
• It is difficult to extrapolate results from a single city
(study) to obtain national estimates of benefits.
Because of the first concern, no additional discussion of property
value studies is provided in this section. However, further discussion can
be found in Section 5.
Surveys of Cleaning Expenditures —
This type of study has been the most popular for measuring reduced
soiling benefits.* Among the studies included in this group are:
Michelson and Tourin (13). Ridker (14), Narayan and Lancaster (15), Booz,
Allen and Hamilton (16), Liu and Yu (17), Brookshire et al. (18), and
Cummings et al. (5). In many cases, the survey studies use some variant of
a paired cities approach to generate data for their analysis. For example,
Michelson and Tourin (13) compare the frequencies of household maintenance
and cleaning activities in Steubenville, Ohio (a dirty area) with the
frequencies of these activities in Uniontown, Pennsylvania (a clean area).
Under the assumption that all other factors that may influence the
responses of households to PM levels are similar in the two cities, a
relationship between ambient PH and frequency of cleaning activities is
estimated. Then, for a given cost per activity, it is possible to predict
how changes in PM impact on frequency and ultimately on total expenditures
* As noted in the Criteria Document (2), the physical damage function
approach has been the most widely used method in materials damage
studies. These studies usually involve metal corrosion or material
deterioration. For the soiling effects considered in this section, we
believe it is more appropriate to view the survey approaches as part of
the behavioral class of models since no quantitative measure of physical
effect is typically included in these analyses.
7-19
-------
for cleaning tasks. In the Michelson and Tourin study cited above, the
analysis indicated that annual' per capita costs for cleaning activities
were $84 (1967 dollars) higher in Steubenville than in Uniontown. Since
the difference in annual mean particulate concentration was approximately
115 (ig/m3, the value per unit change in ambient PM is estimated to be $0.73
per capita.
The Narayan and Lancaster (15) study also used a paired cities
approach. In this study, the costs of maintaining a house were compared
for two cities in Australia. This study, as well as Michelson and Tourin
(13) are faulted in the Criteria Document (2) for potentially suspect data.
In particular, it is noted that attitudes of the respondents to questions
directly connected to air pollution may have provided an incentive for the
respondents to misstate purposefully their true, historical responses.
This type of bias is, of course, possible in any survey situation where
historical data are collected and respondents have a desire to influence
the study results. A procedure for ameliorating this bias is to collect
data on observed activities or to use data collected for reasons not
directly related to air pollution. One of the economic studies reviewed
below uses this procedure.
Another example of the paired cities approach is Brookshire et al.
(18). In this experiment, a more sophisticated survey instrument was
designed to test for the presence of various biases. In addition, the
questionnaire was better suited to obtain estimates of household
willingness to pay than earlier studies which focused only on task
frequency and fixed unit costs. Unfortunately, there are several problems
with using this "contingent valuation" method in a soiling effects study.
The major drawback is that each respondent implicitly makes a choice of
what a given change in PM level means in terms of actual soiling effects.
In effect, there is the presumption that individuals can translate changes
in PM into the effect on average household cleanliness and in turn calcu-
late the added cost if PM levels were to increase. Since different
individuals may have different implicit "physical damage functions", it is
difficult to isolate the value of a unit change in cleanliness.
7-20
-------
In the Brookshire .e_t aJL. (18) study, the difficulty raised above was
overcome by assessing how changes in visual range affected individuals'
willingness to pay. Studies have shown that scientific measures of visual
range and people's perception of visual range correlate well. However, in
valuing changes in visual range, it is difficult to relate changes in
willingness to pay directly to changes in PM levels alone, since multiple
pollutants contribute to visibility impairment. Furthermore, since the
"bids" elicited in the survey did not distinguish among types of effects,
it is difficult to ascribe a part of total benefits to reduced soiling
damages. For this reason, Brookshire et al. (18) is not considered further
in this section.
In one of the first studies of the economic costs of air pollution,
Ridker (14) developed a series of analysis plans to test for the signifi-
cance of TSP in the cleaning decisions of commercial and industrial firms
as well as households. The analyses relied on cross-sectional data
collected from both interurban and intraurban areas. In the analysis of
interurban data, Ridker looked at economic data from three types of
cleaning services. These included receipts from laundry and cleaning
establishments, the costs of cleaning office and apartment building
interiors, and the performance frequency of contract cleaning firms. In
these studies, the hypothesis tested was that cleaning costs per unit time
increased with higher levels of suspended particulates.
For each of these studies, the null hypothesis of no difference could
not be rejected. This conclusion was based on a scan of scatter diagrams
and the calculation of rank correlation coefficients. In order to control
for factors that may confound the identification of a soiling effect,
Ridker partitioned his data to hold constant such factors as climate, per
capita income and interurban price (wage) differentials. Even with these
factors held constant, no clear association between cleaning costs and
suspended particulates was found.
With the intraurban data, Ridker examined supermarket sales of
cleaning supplies and maintenance procedures of firms with branches in
7-21
-------
different parts of a city. Again, no significant association was evident
from the data.
Ridker offers several possible explanations for the lack of associa-
tion between TSP and cleaning cost data. These include: 1) the levels of
air pollution experienced may not be severe enough to lead to an identifi-
able response; 2) measurement error may be an important factor; 3) the
averaging times and statistical aggregation of the pollution data may not
be appropriate; and 4) there may be confounding variables that prevent the
identification of an independent effect of air pollution. Each of these
factors is plausible and could indeed lead to the results reported by
Ridker.
Despite the finding of no association between cleaning costs and
levels of suspended particulates, Ridker's proposed analysis plans were
very innovative and undoubtedly had a significant impact on the design of
soiling cost studies for the following decade. This is especially evident
with respect to the study of soiling costs for households. In Reference
(14), Ridker summarizes a soiling study conducted in Philadelphia, which is
similar in design to a study later conducted by Booz, Allen and Hamilton
(16). In both instances, a survey instrument was designed to elicit
information on cleaning expenditures, frequencies, and time durations for a
variety of household cleaning activities. The specific attributes of this
type of household study are considered in more detail below.
The remaining three studies in the group of survey studies have an
element in common. Both Liu and Yu (17) and Cummings et al. (5) use data
collected by Booz, Allen and Hamilton (BAH) (16) to estimate the relation-
ship between frequency and ambient PM levels. While the BAH study is
perhaps the best known of the many soiling studies, the data set suffers
from many inconsistencies. This becomes especially evident in Cummings .et
al. (5), where a great deal of effort goes into constructing a sound data
base from the original BAH information.
7-22
-------
Although, the type of data collected by BAH is similar to that
collected for the paired city approaches described earlier, the data are
not drawn from a sample within paired cities. Rather, the data come from
four "pollution zones" in the greater Philadelphia area. Consequently, the
analysis of the data must control for demographic characteristics that vary
across pollution zones and may influence household response to alternative
PM levels.
Given the average frequency and aggregate expenditure per task in each
pollution zone, the analysis of soiling damages in References (16), (17),
and (5) is carried out much in the same fashion as described for the paired
city studies. There are, however, some changes of substance that serve to
alter the flavor of the analysis. For example, Liu and Yu (17) use Monte
Carlo techniques to generate a larger data sample, while Cummings et al.
(5) incorporate the opportunity cost of labor into the analysis. Still,
the basic method of calculating soiling damages remains fundamentally the
same in all the survey studies. For this reason, we have chosen to review
in detail only one of the studies in this group. The study chosen is
Cummings et al. (5). In our judgment, this study is a good example of the
type of analysis that can be carried out with task frequency and expendi-
ture data and it offers some novel additions which are not present in the
other survey studies. In addition, since the Cummings study has only
recently been completed, a summary of this work is not currently available
in the most recent Criteria Document. Thus, a thorough review of the
Cummings analysis provides new information on the status of household
soiling benefits models. Following the discussion of the Cummings study,
the survey-type studies will be compared against the evaluation criteria
set forth earlier.
CuBings et al. (5) — There are essentially four major divisions in
the Cummings report. These can be summarized as:
• Background literature review of previous soiling damage
function studies.
7-Z3
-------
A careful analysis of the Booz, Allen and Hamilton (BAH)
data set with corrections made as appropriate.
An analysis of a modified version of the BAH soiling damage
study.
A contingent valuation study of soiling damage.
The discussion here focuses on the third of these topics. As an
aside, we note that the Cummings report asserts that the fourth topic, the
contingent valuation method, is an infeasible approach for estimating
particulate—related household soiling damages.
The objective of the Cummings soiling study is to determine the
economic damages associated with soiling effects. This is accomplished by
estimating expressions that predict how out-of-pocket expenditures and
labor time are likely to vary as ambient PH changes. As noted earlier, the
data for the analysis are drawn from the BAH study (16), although an
extensive data-cleaning effort was undertaken to eliminate inconsistencies
in the BAH data base.
The BAH survey was completed in 1970, and involved 1,800 households in
the greater Philadelphia area. For analysis purposes, the region was
divided into four pollution zones representing concentration levels of:
a
• Less than 75 ug/m .
• 75 to 100 ug/m3.
• 101 to 125 ug/m3.
a
• Greater than 125 ug/m .
Although BAH collected data for 27 specific maintenance and cleaning
activities, the Cummings study focuses on the 11 particulate-sensitive
tasks in the BAH study. Table 7-4 summarizes these activities. Note that
plausible cleaning tasks such as dusting and vacuuming are not included
since they were not part of the BAH survey. This could lead to an under-
estimate of total benefits.
7-24
-------
Table 7-4
CLEANING ACTIVITIES IN CUMMINGS' ANALYSIS
Inside Activities
1. Replace air-conditioner filter
2. Wash floor surfaces
3. Wax floor surfaces
4. Wash windows (inside)
5. Clean Venetian blinds/shades
Outside Activities
1. Clean/repair screens
2. Wash windows (outside)
3. Clean/repair storm windows
4. Clean outdoor furniture
5. Maintain driveways, walks
6. Clean gutters
Source: Cummings et al. (5), Table 8.
For each task, data were collected on the number of households that
performed the task (both households where all tasks were performed by the
residents as well as households that hired outside help), as well as the
mean annual frequency. These data were summarized by pollution zone. In
order to estimate damages, the following cost equations were assumed:
Cleaning costs incurred by those households that hire out-
side help (HIRE)
Cl ' al ' Nl ' Fl
where a^, is the unit cost of having the task performed; Nj
is the number of HIRE households; and Fj is the mean annual
frequency for the task in HIRE households.
(7.1)
Cleaning costs incurred by do-it-yourself households (DIT)
Cj = N2 ' [V • T + a2 ' F2]
where N2 is the number of DIY households performing the
task; V is the imputed labor cost per unit of time in DIY
households; T is household time spent performing the task;
a2 is the unit out-of-pocket cost of the task in a DIY
home; and F« is the mean annual frequency for DIY
(7.2)
7-25
-------
households. The inclusion of a measure for the value of
labor is a unique feature that distinguishes the Cummings
study from other benefits analyses of reduced soiling.
Total costs are determined as the sum of Cj and Cj. Note that since data
for N. and F. are summarized by pollution zone, total cost estimates are
derived for each zone. In addition, since F^ and possibly V and T will
vary with pollution levels, changes in PH will impact total costs. This
provides the way in which soiling damages can be assessed for changes in
PM.
Given data for N- and F., the next step in the Cummings analysis is to
determine values for a^, aj, V, and T. The unit cost parameters a^ and a~
are determined from the BAH study. Unfortunately, these costs are provided
for only five of the 11 particulate-sensitive tasks included in the
Cummings analysis. Consequently, costs must be attributed to the other
tasks on the basis of similarity with tasks that do have cost data. This
is likely to be an important source of measurement error in the analysis.
However, the direction of bias is unknown.
In order to develop quantitative estimates for T and V, a contingent
valuation survey was developed for the Cummings study. This questionnaire
was used to obtain measures of the opportunity cost of household labor as
well as the time spent in cleaning operations. The sample respondents were
drawn from each of the four pollution zones in the Philadelphia area, with
each zone represented by 30 respondents.
With data identified for T and V by pollution zone, the next step
involved relating F, T, and V to PM. Since the various pollution zones
include households with diverse incomes, this variable should also be
controlled for in the regression analysis. Cummings et al. posit the
following relationships:
F = o0 + oxP + ojl (7.3)
P2T (7.4)
7-26
-------
V =
XQ +
+ XjY
(7.5)
where F, T, and V are as defined previously; P is ambient annual average
TSP concentrations; and Y is household income. When these regressions were
run, approximately half the specifications failed to reject the null
hypothesis that all the estimated coefficients were zero. Three of the
frequency equations and one time equation had negative signs reported for
the pollution coefficient. Thus, estimates of the task-specific F and T
functions are only moderately successful. Furthermore, the assumption of
linearity and the consideration of only two independent variables may be
too restrictive. Alternative specifications may yield very different
results.
Because pollution levels have decreased from the time of the original
BAH study Cummings et al. were forced to make some adjustments in the data
prior to the determination of total costs by pollution zone. Given these
adjustments, total costs for 10 of the 11 particulate—sensitive tasks were
computed using Equations (7.1) and (7.2).* Table 7-5 summarizes the
*
findings.
Table 7-5
TOTAL PER-HOUSEHOLD SOILING COSTS BY POLLUTION ZONE
(1980 dollars)
HIRE Households
DIY Households
Pollution Zone (|ig/m )
40
$1,531
763
81
$2,887
905
102
$2,558
1,067
123
$2,683
1,386
Source: Cummings .e_t jil. (5), Tables 18 and 19.
* The task of maintaining driveways was not considered.
7-27
-------
With, the per-household costs identified, economic damage estimates by
pollution zone can be derived by multiplying cost/task by the number of
households performing the task in a given zone. Note that the term
"damage" as used here is really appropriate only in the case where cleaning
costs are zero when particulate levels are zero. By labelling all cleaning
costs as economic damages, the implicit assumption is that all cleaning
occurs only because of pollution. A better way to describe economic
damages is the added costs incurred because of increases in ambient PM
levels. Thus, it is the valuation of changes in cleaning activities that
is most relevant.
The economic damages derived in the Cummings study are on the order of
$300 million for each of the pollution zones. Although these numbers are
much larger than those found in BAH, this can be traced to the value of
time which is included in Cummings analysis.
The data on per-household cleaning costs by pollution zone can be
related to the corresponding zone pollution level to identify an economic
damage function. Given the Cummings data, a linear specification was
estimated as:
D = 251.43 + 6.63P (7.6)
where D represents damages in dollars per household per year, and P is the
annual average level of TSP. The linear damage function implies that
marginal damages are $6.63 (1980 dollars) per household per year for each
microgram change in TSP levels.* This is the information that would be
required to make extrapolations to national benefits estimates. Benefits
in county i in a given year would be estimated as:
B£ = 6.63 • AP. • H^ (7.7)
* Recall that the marginal value in the Michelson and Tourin (12) study was
estimated to be $0.73 in 1967 dollars, or about $2.50 in 1980 dollars.
7-28
-------
where Bj^ is benefits in the ith county; AP is the change in TSP levels in
the county, and H is the number of households in the county.*
It should be pointed out that the linear damage function in Equation
(7.6) is simplistic and as Cummings et al. note, "must be viewed as little
more than of expository value." While it is true that a nonlinear damage
function would be expected, we also feel that data problems (e.g., unit
cost data), and the several statistically insignificant task frequency
regressions add to the difficulty in accepting the damage function in
Equation (7.6). While the derived marginal damage estimate is not strong
enough to stand alone as an indicator of economic damage, it does seem
reasonable to use Equation (7.7) as part of the information base available
for developing estimates of reduced soiling benefits. In fact, given the
framework of the Cummings study, it may be possible to ascertain whether
this estimate should be an upper- or lower-bound to an estimate derived
from another study. This may help to define a reasonable range for
benefits from reduced soiling.
We have described in some detail the major elements of the Cummings
soiling study. In the report, an effort is also made to clarify some
confusing issues in the BAH work and they close with several recommenda-
tions for the future course of frequency studies. First, they believe that
it is more appropriate to define aggregate tasks rather than specific
tasks. Second, time spent per unit time in an activity is perhaps a more
relevant variable than frequency in the estimation of the relationship
between activities and pollution levels. Cummings et al. believe that a
mixed approach based on the contingent valuation and soiling frequency
techniques would best be able to obtain the data that would incorporate
these considerations.
Evaluation of Surrey Studies — Several of the direct survey studies
have been discussed briefly, and one study reviewed in some detail. This
* This is also the equation used by SRI (19) to estimate benefits from
reduced soiling.
7-29
-------
class of studies is now examined to see how well they conform to the
evaluation criteria.
Theoretical basis — With the exception of Brookshire et
al. (18), the survey studies do not collect data sufficient
to obtain the appropriate measure of economic benefits. In
general, frequency of cleaning data and/or total expendi-
ture data do not allow one to observe the variations in
demand or supply that are required for the proper estima-
tion of benefits. Consideration of variation in total
expenditures alone will yield underestimates of "true"
benefits. See Courant and Porter (20) for additional
treatment of this issue.
Consideration of relevant variables — Ideally, the paired
city approaches are designed so that other potentially
significant explanatory variables in the frequency/PH
regressions would be constant across study areas. In those
studies which use the BAH data, however, factors such as
income should be included in the frequency specifications,
since this attribute may influence cleaning decisions and
will differ across the pollution zones being analyzed.
Quality of input data — Based on comments in the Criteria
Document (2), this would appear to be of major concern.
The principal problem with the paired cities studies is
that the frequency data may contain inherent biases.
Furthermore, the BAH data set also suffers from inconsis-
tencies in many of the variables. Thus, since these
studies typically collect only expenditure data, unit cost
(price) information is not directly available from the
surveys. As a consequence, there is likely to be measure-
ment error in the cost data collected from outside sources.
Legitimacy of the empirical analysis — The frequency
functions that are estimated are usually simple linear
specifications. As with the physical damage functions,
there is little information available on the most appro-
priate functional form for these economic damage functions.
In the Cnmmings study, it was noted that the null hypothe-
sis that all parameters in an equation were zero could not
be rejected in almost half the frequency equations.
Despite this statistical problem, cost estimates were still
derived for each task. More work is required to give a
better theoretical basis to the estimated equations.
Transferability of models — As evidenced by Equations
(7.6) and (7.7), it is relatively easy to identify the
information needed to develop national estimates of
benefits. It is perhaps wise to reiterate that the extra-
polation relies on a marginal damage estimate developed
7-30
-------
from data from a single city. As a consequence, an
analysis of benefits from the Cummings results must assume
that individuals in other areas will react to soiling
damage in a fashion similar to that of Philadelphians.
Overall, the evaluation for the survey studies is not much more
optimistic than that presented for physical damage function studies.
Indeed, the negative comments quoted earlier from the Criteria Document (2)
likely were meant for these studies as well. There are, however, some
distinct advantages to this approach that should be recognized. First, the
approach focuses on people's responses to air pollution rather than an
imputed value of expenditures for returning to some hypothetical clean
state. Second, the data requirements would seem to be less burdensome.
With the survey studies, it is no longer necessary to develop estimates of
surface area exposed or to identify damage functions for a variety of
objects.
In summary, it does not appear that any one of the survey studies
could stand alone and provide a defensible estimate of benefits. However,
when used as part of a general information base, marginal damages derived
from the survey studies may be valuable components of a more general
benefits analysis.
Economic Deaaad and/or Supply Models —
The final class of behavioral models includes those studies that rely
on economic demand or supply functions to estimate benefits. We are aware
of three models that can be classified in this manner.
• Watson and Jaksch (16).
• Mathtech's model of household expenditures (7).
• Mathtech's model of cost and production relationships in
the manufacturing sector (8).
Since these models differ markedly, it is not possible to provide a blanket
description of how benefits estimates are developed. However, the
7-31
-------
underlying rationale on how economic benefits should be defined is the same
in each case. Basically, the value to society of a reduction in ambient PM
levels should be measured by the aggregate amount members of society would
be willing to pay to be in a preferred state (i.e., cleaner air) as opposed
to a less preferred state (i.e., dirtier air). This willingness-to-pay
measure includes two components. The first of these is the level of
reduced out-of—pocket cost that would accompany lowered PH levels. This is
the measure captured by the survey methods described in the previous sub-
section. The second component allows for the possibility that the cost
savings may reduce prices and thus stimulate an increase in the quantity of
goods and services demanded. Since the increased quantity demanded is
direcly attributable to improved air quality, the willingness-to-pay
measure should include this component as part of the benefit estimate as
well.
Measures of willingness to pay can be ascertained from knowledge of
demand and supply curves. In Figure 7-2, the demand curve is given by D
and the supply (marginal cost) curve is designated as MC If these curves
are thought of as the households' demand and supply of cleanliness, then a
decrease in PM levels may imply that each unit of cleanliness could be
produced at lower unit cost than before. That is, less resources would be
required to attain the level of cleanliness that occurred before the air
quality improvement. In this case, the supply curve would shift down to
MC'. The shaded area ABCO represents the change in economic surplus, or
the amount consumers would be willing to pay to have the improved air
quality. Thus, it is the correct measure of economic benefits.
The same general principles for measuring benefits are used in each of
the three studies reviewed in this subsection. As noted above, however,
because the underlying models that produce the relevant demand and supply
functions do differ, each of the studies will be evaluated individually.
Watson and Jaksch (6) — The study by Watson and Jaksch (WJ) is some-
what unique in that it could easily be classified as a physical damage
function study, a survey study, or an economic demand/supply model study.
7-32
-------
Price per
unit
Me
MC1
Q,
Q, Quantity per unit time
Figure 7-2. Example of economic surplus
7-33
-------
Elements of each type of study appear in. the WJ analysis. However, since a
major distinguishing feature of their analysis is the use of demand and
supply curves (rather than cleaning frequency or expenditure), it seems
most appropriate to include the study in this subsection.
The data set used in the WJ analysis is taken from the Booz, Allen and
Hamilton (16) study discussed earlier. Since the WJ study was completed
prior to the Cummings e_t aJL (5) study, WJ did not have the benefit of the
data-cleaning effort undertaken by Cummings et al. Thus, some caution is
warranted since the BAH data in its original form does contain inconsisten-
cies.
The WJ reliance on the BAH data set did provide for an easy charac-
terization of the demand curve. In the original BAH analysis, it was
determined that expenditure remained constant across different levels of PH
for a variety of tasks. If alternative levels of PM reflect different
states of household cleanliness, then this implies that expenditures remain
constant across units of cleanliness. In this case, the demand curve for
cleanliness would have constant unitary negative elasticity (i.e., a
rectangular hyperbola). This makes specification of the demand function
straightforward.
The assumption of unit elasticity is an important part of the WJ
analysis. However, several recent commenters have questioned whether this
assumption can legitimately be drawn from the BAH data. Cummings et al.
note that many of the tasks that were identified as not being sensitive to
particulate matter involved materials damage-type impacts. In addition,
Cummings .e_t .§_!. point out that activites identified in BAH as having low
unit values (and thus providing only small additions to total cost) are
actually performed many times over the course of a year so that total
annual costs may be high. Rowe (21) also provides an example which shows
how increasing pollution coupled with a constant cleaning frequency can
imply expenditure increases. In this case, outlays on cleaning are
positively related to pollution levels and an inelastic demand curve would
be implied.
7-34
-------
These comments indicate that the constant elasticity demand curve used
by WJ may impose constraints on the model that are not supported by avail-
able data. If the demand curve is suspected to be inelastic, then as Rowe
(21) points out, it is likely that the WJ benefits estimates are over-
estimates for improvements in air quality on this point.
Similar problems are possible with the WJ supply (marginal cost)
curve. The assumption made in the WJ case 1 (with which we deal
exclusively) is that the marginal cost schedule rotates in proportion to
the ratio of the final and base pollution levels.
MC = a • (Pf/Pb)a ' Q (7.8)
where HC is the marginal cost per unit of cleanliness; Pf is the final
level of pollution; P^ is the base level of pollution; Q is units of
cleanliness; and a and a are parameters to be determined.
Several comments can be made about Equation (7.8). First, WJ argue
that the HC curve will be increasing in cleanliness because cleaning costs
and the opportunity cost of labor will increase as pollution increases.
This would imply that "a" is positive. However, it is not explicitly
recognized by WJ that there may be economies of scale related to certain
cleaning activities (e.g., buying larger boxes of soap) which could alter
the slope of the MC curve. Indeed, even the linear representation of the
MC schedule may be at odds with * priori expectations. One would instead
expect the MC curve to increase at an increasing rate.*
In addition to these observations on the shape of the MC curve, it is
also pertinent to examine the impact of pollution changes on MC. As
written. Equation (7.8) implies a multiplicative rotation. The factor that
determines the degree of shift, a. is derived from several damage function
studies. In particular, the painted cedar siding equation from Beloin and
* WJ recognize this point during their discussion of the MC curve,
treat it as linear for ease of presentation.
but
7-35
-------
Haynie (1) yields estimates of o of 0.56 and 1.0. The study by Esmen (22)
was used to obtain an upper-bound estimate for a of 2.0. These factors
were used to rotate the MC schedules for each of the eight cleaning and
maintenance tasks examined by WJ. These tasks represent a subset of the
BAH activities. They include: painting indoor and outdoor walls, washing
indoor and outdoor windows, washing indoor walls, wallpapering, painting
tria, and cleaning Venetian blinds. As with the Cummings study, plausible
cleaning activities such as dusting and vacuuming are not considered
because they were not included in the BAH survey. As the earlier discus-
sion of physical damage functions noted, the absence of damage functions
for a variety of materials makes it difficult to measure benefits. WJ rely
on functions developed for only two materials to derive their estimates of
a. Thus, the shape and rotation of the MC schedule may also contribute to
a bias in the benefits estimates. With so little empirical evidence
available for a, it is difficult to assess whether the bias will lead to an
overestimate or an underestimate of benefits.
Figure 7-3 shows the WJ demand and supply curves for cleanliness. The
shaded area OBF represents the consumers loss from an increase in pollution
to ?2 from P^. Given the assumptions on the demand elasticity, original
outlays OAB equal final outlays OEF. The important point to observe is
that economic benefits are generated even though no additional outlays are
made. Thus, the WJ analysis picks up a component of benefits that could
not be observed in the survey methods reviewed earlier.
With specific algebraic expressions for the demand and supply curves,
it is possible to solve for the value of economic surplus as a function of
the base and final pollution levels, the value of a, and total expendi-
tures, by pollution zone. The derived expression is:
-ACW.. = a In
(7.9)
where ACW^. is the change in welfare for task i in pollution zone j.
?2i is the final level of pollution in zone j.
7-36
-------
Price per
unit
MC(P2)
•MC(P1)
A Q units of cleanliness
Figure 7-3.- Demand and supply curves in the WJ analysis
7-37
-------
Pj^ is the initial level of pollution in zone j.
a is the factor estimated from the damage function studies.
It's value is 0.56, 1.0, or 2.0.
X.. is the outlay for task i in zone j.
Equation (7.9) is the equation needed to estimate national benefits.
Because of the required extrapolations, we believe that separate calcula-
tions by task and pollution zones are not warranted. Thus, in any benefits
calculations, pollution values would be county averages and total expendi-
tures by county would be developed by multiplying the implied per-household
total cost (expenditures) in Philadelphia times the number of households in
each specific county. The necessary data on household expenditures in
Philadelphia is developed in WJ.
In terms of the evaluation criteria, the following comments can be
made:
Theoretical basis — The manner in which economic benefits
are calculated is appropriate for the single activity con-
sidered in the study. Questions do arise on the validity
of the assumptions implicit in the shapes of the demand and
supply curves.
Consideration of relevant variables — The assumptions
underlying the demand curves imply that specific household
classes do not have to be identified when estimating
benefits. This occurs because of the constant unit
elasticity.
Quality of input data — With the BAH data set used to
estimate benefits, there is some concern on the reliability
of the data. As noted earlier, Cummings .e_t .§_!. (5) devote
some time to cleaning up inconsistencies in the BAH data.
If possible, it would be appropriate to use the corrected
BAH data in any calculations performed with Equation (7.9).-
This may be limited by differences in tasks between the
studies.
Legitimacy of the empirical analysis — Despite the fact
that WJ use demand and supply curves, no new estimation is
performed in this study. The crucial aspect is whether the
assumptions made by WJ in specifying the demand and supply
7-38
-------
relations are reasonable. The only numerical calculations
involve evaluation of Equation (7.9).
Transferabilitv of Model — This can be readily accom-
plished using Equation (7.9). Data on expenditures are
available from WJ. As with Cummings et al.. extrapolations
to national benefits estimates rely on data collected for
only one city.
In general, this evaluation coincides with comments made in the
Criteria Document (2) and the Staff Paper (3). Both of these reports are
cautious in accepting the estimated benefits of WJ because of the limited
number of soiling functions which are available to represent a variety of
cleaning tasks. In addition, we reiterate comments made earlier that
additional empirical work is required to validate the assumptions under-
lying the forms of the demand and supply schedules. Though there are areas
where improvements are possible, the WJ study is an innovative approach to
estimating soiling benefits. While it is difficult to rank WJ in relation
to Cummings et al. (5), it seems prudent to view WJ as providing additional
information for defining the probable range of soiling benefits.
Mathtech Household Model (7) — In the WJ analysis, household
decisions are examined for only one of the many commodities (services) that
influence economic behavior, the production and consumption of cleanliness.
In addition to the services provided by cleanliness, individuals also
require shelter, nutrition, clothing, and other services to maintain a
desired quality of life. In the Mathtech household model (MTH), the inter-
dependency of economic decisions is recognized, and systems of demand
equations are estimated. That is, household demands for a variety of
services are considered simultaneously. This permits consideration of some
of the adjustment opportunities available to individuals as air quality
improves.
The basic structure of the MTH model involves a two-stage decision
process. Figure 7-4 outlines the major Components of this process. The
initial decision facing the household is the allocation of a fixed budget
among the many market goods that may be purchased. To determine these
7-39
-------
Environmental
Conditions
Relative
Prices
Socioeconomic
Characteristics
Demand for
Market Goods
Index of Produced
Service Flows
Fignre 7-4. Household decision process
"demands", various factors that are beyond the control of the household
must be taken into account. In particular, the relative prices of the
goods, income, and various demographic factors all help shape the pattern
of household demand. Furthermore, environmental variables such as ambient
concentrations may also influence the demand for certain goods. For
example, the demand for detergents or other cleaning products may be sensi-
tive to the level of PM concentrations.
Once the allocation decision for market goods has been made, condi-
tional on the factors mentioned above, the decision-making role of the
household is essentially complete. However, there is a natural extension
to the process that is important for benefits analysis. This additional
step represents the second stage mentioned above.
The idea behind the extension is that items like detergents are not
demanded for their own sake, but rather for the services they provide.
7-40
-------
Thus, it may be instructive to view the demand for detergents as a derived
demand based on a more fundamental consumer demand for cleanliness. Viewed
in this way, the allocation decision made by households with respect to
demands for detergents and other cleaning items is consistent with the
attainment of a particular level of cleanliness. Furthermore, the role of
air pollution is clear. Air pollution increases the cost of cleanliness by
increasing the quantity of detergents and other cleaning items required to
obtain a unit of cleanliness. Conversely, a reduction in ambient air
pollution can lead to a reduction in the unit cost of cleanliness.
In terms of Figure 7-4, this discussion implies that a link must be
established between an index of produced services such as cleanliness and
market goods. This link is formalized in the MTH model with the adoption
of "separability" assumptions that permit the many market goods that may be
purchased to be grouped naturally into a series of categories. These
categories represent the various "produced" goods described above. The
implication of the separability assumption is that the various items that
contribute to the production of nutrition (e.g., meats, dairy products,
fruits, vegetables, etc.) can be analyzed independently of the market goods
which contribute to the production of cleanliness (e.g., laundry products,
floor wax, etc.). Thus, the separability assumption narrows the dimen-
sionality of the system estimation effort. Furthermore, the assumption
plays an important part in the definition of quantitative indices for the
unobserved produced services.
The model described above is structured to deal directly with house-
hold adjustments to air quality changes. The focus is on the value house-
holds place on certain activities rather than specific pollution-induced
damage. A physical damage function between pollution and objects that may
be damaged is not included in the analysis. Instead, pollution enters the
model as a proxy for damage. It is not imperative that the type or extent
of physical damage be identified explicitly. The knowledge that is
required is the value household decision makers attach to activities or
services that may be sensitive to air quality changes.
7-41
-------
A factor that can limit the usefulness of this approach is that price
and quantity data are not available for the various produced services.
Thus, it is not possible to estimate directly demand relations for these
non-market services. Fortunately, the assumptions that permit the grouping
of market goods are also instrumental in establishing consistent price and
quantity indices for the produced services. As would be expected, these
indices, for any particular service, depend on the relative prices
(demands) for the market goods in the service category. This is an
important observation, since it implies that the aggregate indices may be
indirectly affected by levels of air quality through the demand equations
of the market goods.
Thus, with information on price and quantity indices for the produced
services, it is possible to estimate a system of demand equations for these
services. Furthermore, since air quality indirectly affects the indices,
it is possible to ascertain the impact of changes in air quality on the
service indices. This permits the identification of economic benefits
associated with the postulated air quality change.
The data used to estimate the HTH model are drawn primarily from the
1972-73 Consumer Expenditure Survey (23). Expenditure data for over 100
current consumption items are available for 24 large Standard Metropolitan
Statistical Areas (SHSA). Price data for items defined in Reference (23)
are taken primarily from Reference (24). The SHSAs included in the HTH
analysis are shown in Table 7-6, while the goods eventually included in the
analysis are shown in Table 7-7. Note that the 20 basic goods are grouped
into seven commodity aggregates. This is in accord with the structure of
the decision process outlined above.
The goods included in the HTH analysis account for approximately 40
percent of total annual consumption expenditures. Two categories of
commodity demand that may be affected by changes in PM levels are omitted
from the analysis. Data and conceptual problems did not permit the
analysis of recreational (leisure) services nor services generated by
property. Since labor-leisure tradeoffs and location adjustments are two
7-42
-------
Table 7-6
SMSAs INCLUDED IN MTH ANALYSIS
REGION I: NORTHEAST
1. Boston
2. Buffalo
3. New York City
4. Philadelphia
5. Pittsburgh
REGION III: SOUTH
1. Atlanta
2. Baltimore
3. Dallas
4. Houston
5. Washington, D.C.
REGION II: NORTH CENTRAL
1. Chicago
2. Cincinnati
3. Cleveland
4. Detroit
5. Kansas City
6. Milwaukee
7. Minneapolis
8. St. Louis
REGION IV: WEST
1. Denver
2. Honolulu
3. Los Angeles
4. San Diego
5. San Francisco
6. Seattle
Source: Mathtech (7), Table 4-1.
7-43
-------
Table 7-7
•
GOODS INCLUDED IN THE MTH ANALYSIS
Food
Cereal and Bakery Products
Meat
Dairy Products
Fruits and Vegetables
Miscellaneous Foods
Shelter
— Home Repair
— Utilities
Home Operations
— Laundry and Cleaning Products
— Other Household Products
Home Furnishings and Equipment
— Textiles
— Furniture
— Major and Minor Appliances
— Housewares
Clothing
— Clothing
— Dry Cleaning
Transportation
— Gas and Fuel
— Other Vehicle Operations
Health and Personal Care
— Personal Care
— Non-Prescription Drugs
— Non-Insured Medical Treatment
Source: Mathtech (7), Table 4-11.
7-44
-------
plausible means of responding to perceived changes in PM levels, failure to
recognize these adjustment possibilities could lead to an underestimate of
the households' true willingness to pay. In addition, lack of data pre-
cluded the use of activity-specific measures of time in the analysis. This
can also lead to an underestimate of benefits.
The estimating forms for the demand specifications in the first stage
of the decision process are derived from a Stone-Geary utility function.
This function has the mathematical properties required for the two-stage
decision process to be valid. The demand system associated with this form
of the utility function is known as the Linear Ezpenditure System.*
In the first stage estimation, the demands for two goods are found to
be sensitive to PM levels. As expected, the demand for laundry and
cleaning products is directly related to PM concentrations. Also, the
demand for utilities (gas and electricity) exhibits a statistically signi-
ficant direct relation with PM. A possible explanation for the latter
finding is that electricity is used in conjunction with other goods to
mitigate the effects of ambient PM. These two equations are shown in Table
7-8. Recall from the discussion above that these equations represent only
a part of the decision process of the household.
The measure of PM included in the soiling studies reviewed previously
was an annual arithmetic or geometric mean measure of TSP. In the MTH
study, the measure of PM that is found to be most robust in the various
specifications is a 24-hour averaging time, second-high TSP value for the
year. The index of TSP for the SMSA is formed by taking the maximum of
second-high concentrations across all sites in the SMSA,
An important feature of the MTH analysis is the variety of checks
undertaken to test the plausibility and sensitivity of the results. These
checks include: consideration of an alternative functional form (the
* A discussion of the properties of the Stone-Geary utility function and
the LES system can be found in Stone (25).
7-45
-------
Table 7-8
DEMAND EQUATIONS WITH TSP
o
Y31 = (1 - Al) • (CO + C
Y21 = (1 - Al) • (CO + Cl • S
1 • TX2HI) • (LAUCPI/MOPS) + Al - Al • ( DO + Dl
Coefficient
Al
CO
Cl
DO
Dl
Value
0.42462
-21.09430
0.01865
-32.13680
-10.13580
St. Er.
0.04504
12.46690
0.00647
16.82930
4.97639
X2HI) • (RPRPI/MSHELTER) + Al - Al • ((DO + Dl
Coefficient
Al
CO
Cl
DO
Dl
Value
0.23301
-32.23830
0.02393
-149.34600
0.12010
St. Er.
0.01562
15.55350
0.00977
47.04200
0.05375
• REGION) • (OHSEPI/MOPS)
• TX2HI) • (UTILPI/MSHELTER))
Definition of Acronyms
Y31 = Expenditure share of home operations budget on laundry and cleaning.
TX2HI = 24-hour averaging time, maximum .second-high concentration of TSP.
LAUCPI = Price index for laundry and cleaning goods.
MOPS = Total expenditure on home operations.
REGION = Dummy variable for region of country.
OHSEPI = Price index for other household service goods.
Y21 = Expenditure share of shelter budget on home repairs.
SX2HI = 24-hour averaging time, maximum second-high concentration of SO-.
RPRPI = Price index for home repairs.
MSHELTER = Total expenditure on shelter.
UTILPI = Price index for utilities.
Source: Tables 4-16 and 4-17 of MTH (7).
-------
homogeneous translog); looking at different ways to enter pollution
variables in the demand specifications; comparisons of price elasticities
with elasticities from other studies; and reasonableness checks for implied
TSP impacts. These tests add support to the plausibility of the basic
results in MTH, and lend additional credence to benefit numbers generated
from the household model.
Benefits estimates are derived in the MTH study by comparing expendi-
ture functions before and after the PM change.* This procedure differs
slightly from the method described earlier, where benefits were defined in
terms of the area to the left of the ordinary demand curve and between the
marginal cost curves (see Figure 7-2). In the MTH analysis, the comparison
of expenditure functions is equivalent to measuring benefits in terms of
the area to the left of the compensated (i.e., utility constant) demand
curve and between the marginal cost curves. For air quality improvements,
this latter way of measuring benefits will lead to a lower estimate than
that obtained via the former method.
The calculation of benefits through a comparison of expenditure
functions can also be identified with the compensating variation measure of
benefits. In particular:
CV - E(P1,U) - E(P2,U) (7.10)
where CV is compensating variation; E(*) is the expenditure function; P is
a vector of price indices in an initial situation; P2 is the vector of
price indices after a change in the indices due to an air. quality improve-
ment; and U represents a constant level of the utility index. The CV
measure is interpreted as the amount a consumer would be willing to pay (or
would have to be paid) in order to be indifferent between an original
situation and a new situation with lower (higher) prices. It is thus a
proper measure of economic benefits.
* For a discussion of expenditure functions, see Diamond and McFadden (26),
7-47
-------
In terms of the evaluation criteria, the following comments can be
•
made:
Theoretical basis — Estimates of economic benefits are
derived in a manner that is consistent with tenents of
welfare economics.
Consideration of relevant variables — The focus of MTfl on
the optimizing behavior of consumers allows one to take
advantage of & priori restrictions available in conven-
tional models of consumer behavior. This helps narrow the
scope of relevant economic variables. However, it may be
the case that socioeconomic factors or other environmental
measures may also contribute to the explanation of varia-
tion in quantity demanded. Because of the complexity of
the system estimation procedures, only four exogenous
factors are analyzed in the various demand specifications.
Quality of input data — The model derived in MTH requires
observations on individual household behavior. Available
data are limited to average, SHSA-level observations for 24
cities in two years. Air pollution data are also averaged
over the entire SMSA. It is not clear what biases may be
introduced by reliance on the aggregate data. A positive
aspect of the economic input data is that it was collected
for purposes other than an air quality benefits analysis,
and so would not be subject to the inherent biases
suspected in the survey studies.
Legitimacy of the empirical analysis — The MTH study uses
advanced econometric techniques which are appropriate to
the model being analyzed. The choice of functional forms
for the demand relations is dictated by the choice of a
particular utility function. The Stone-Geary function
chosen for this study has been used often in economic
analysis, is fairly easy to work with, and has properties
that are required for the two-stage decision problem to be
valid. In addition, a variety of plausibility checks were
undertaken, including the specification of an alternative
functional form.
Transferabilitv of models — Benefits are calculated in the
MTH model by comparing expenditure functions before and
after pollution-induced price changes. Although the
expression used to evaluate benefits is complex, benefits
can be derived in a fairly straightforward manner for the
scenarios to be reviewed in this report. The extrapolation
to national benefits requires assumptions be made on
regional prices and expenditures. These variables are
assigned values based on the data observed for the 24 SMSAs
used in the MTH analysis.
7-48
-------
Despite the recent completion of MTH, the Criteria Document (2)
contains a brief commentary on the overall Mathtech effort. The Document
notes that the Mathtech models allow consideration of substitution possi-
bilities and appear to be well based from a theoretical and empirical
perspective. In addition to the concerns raised above with respect to the
quality of the input data, it is noted in the Criteria Document that
additional analysis would be beneficial in helping to explain more fully
some of the behavioral adjustments implied by the models. The Staff Paper
(3) does not specifically address the theoretical or empirical methods
employed by Mathtech. However, additional detailed comments are available
from an EPA-sponsored public meeting held in July 1981 to assess the
overall Mathtech study.* At this meeting, a panel of leading environmental
economists was asked to review and comment on each of the sector analyses.
With respect to the household model, the general conclusion was that this
was an excellent piece of research. Theoretical and econometric methods
were properly carried out, and sufficient care was given to assessing the
limitations imposed by required assumptions. The major area of concern was
with the input data. The aggregate nature of the data and the limited
number of sample points made for less than an optimal data set.
Despite the data limitations, the general consensus of the review
panel members was that the MTH model would generate the most defensible
estimates of benefits from among currently available studies. Because of
assumptions made in the analysis, it was recognized that benefits numbers
generated from the MTH model would be conservative estimates of willingness
to pay.
This evaluation leads us to believe that reduced soiling benefits
generated from the MTH model should be weighted most heavily in a compari-
son of benefits across studies. Benefit estimates derived from survey
studies like Cummings ejt al. (5) and the demand/supply model of Watson and
Jaksch (6) can serve as plausibility checks to the MTH numbers, and help
define a probable range of benefits from reduced soiling for households.
* See Volume VI of Mathtech (27) for a summary of the public meeting.
7-49
-------
Hath tech. Manufacturing Model (8) — The economic models discussed
earlier all focus on the household sector. Clearly, other sectors in the
economy can also be affected by soiling damage. Commercial establishments,
industrial plants, and government facilities may all experience decreases
in costs associated with soiling damage as PM levels are reduced. The
Mathtech manufacturing analysis (MTM) examines the benefits associated with
reduced PH levels for six manufacturing industries at the 3-digit level of
the Standard Industrial Classification.*
The MTM analysis uses economic techniques to derive a general model of
production costs for manufacturing firms. The hypothesis of the analysis
is that soiling or contamination effects within manufacturing plants add to
the total cost of production. The costs of production may increase because
of 1) increased expenditures for cleaning, maintenance, and repairs; 2)
substitution to costlier materials which are more resistant to damage; or
3) reduced performance from affected equipment or structures.
Comparisons of production costs among similar types of manufacturing
*
firms located in regions with different levels of PM permit the estimation
of PM effects. Naturally, in making such comparisons, it is important to
control for other factors which may cause production costs to vary among
regions. These factors may include variations in: wage rates, capital
costs and/or capital investment in—place, taxes, prices for materials and
services, etc. Many of these factors are controlled for by including them
in the analysis of production cost variations.
Also taken into account in the analysis are variations in climato-
logical conditions which may influence costs directly (e.g., through varia-
tions in heating and cooling costs), and which may also influence the
extent to which ambient PM causes physical and economic damage. The clima-
tological factors considered include ambient temperature and precipitation.
Since SC^ levels may act in concert with PM to cause damage, ambient
measures of SOj are also controlled for in the analysis.
* The MTM study also looks at SC^ effects in the manufacturing sector.
These are not discussed here.
7-50
-------
The approach, described above results in an estimate of the total costs
of production for a firm as a function of input prices, the level of
output, and other exogenous factors. An evaluation of costs before and
after implementation of a PM control strategy provides an estimate of
benefits.
The formulation of the model in terms of production costs has several
advantages relative to the more standard physical damage function
approach.*
Data used are'for production costs actually incurred by
firms as opposed to estimates of what typical maintenance
and repair costs might be.
The model recognizes that firms can make choices in their
response to pollution damage. For example, the model
allows for the possibility that firms may substitute to
costlier but more damage—resistant materials rather than
incur air pollution damages.
Effects for which damage functions have yet to be developed
can still be measured.
Subtle effects such as efficiency losses are captured.
Data employed in the estimation of the total cost functions are taken
from the 1972 Census of Manufactures (28). Because of confidentiality
restrictions, the Census Bureau does not report data on individual plants.
For this reason, the MTM analysis focuses on county data for the 3-digit
SIC level. An assessment of where PH damage might be expected to be
greatest, coupled with data availability constraints, limits the MTM
analysis to six industries. Estimation of the cost relationships indicates
that PM, specifically 24-hour average second-high TSP values, is an
important explanatory variable in two industries, SIC 344 (Fabricated
Structural Metal Products) and SIC 354 (Metalworking Machinery). Our
* Most of the physical damage function studies that are based on industry-
related impacts deal only with corrosion. Models of soiling referenced
in the Criteria Document (2) and Staff Paper (3) are all household
studies.
7-51
-------
evaluation of reduced soiling benefits is limited to these two 3-digit SIC
classifications and the 2-digit SICs (34 and 35) of which they are a part.
Since these industries contribute only a small fraction to the value of
product in the manufacturing sector, estimates of benefits are conservative
estimates of total benefits in the sector.*
An important component of the MTH analysis is a series of validation
checks which are performed to assess the plausibility of the model results.
These checks include an evaluation of various estimated economic para-
meters, tests for sensitivity of results to different methods for including
pollution variables in the models, reasonableness checks for implied pollu-
tion effects, and tests for spurious correlations due to omission of own
pollution control costs. These tests add support to the plausibility of
the basic results and thus lend credence to benefit numbers generated from
the models.
As noted earlier, the Criteria Document (2) briefly reviews the
overall Hathtech effort, with the general conclusion being favorable with
some qualifications. In addition, the July 1981 Public Meeting also con-
sidered a draft version of the MTH study. The general conclusion at the
Public Meeting was that the MTM analysis is a careful and sophisticated
piece of research. The assumptions underlying both the model structure and
the empirical analysis are made explicit and plausibility checks are con-
ducted at many points in the study.**
There were three major areas of concern. First, general comments made
at the meeting by representatives of the American Iron and Steel Institute
(AISI) were that the relationship observed between air pollution and manu-
facturing costs should not be construed as indicative of causation, but only
correlation. The representatives cited several factors which they felt
* Even within a particular SIC category, data were not complete. For
example, in SIC 344, 57 counties were included in the analysis. Plants
in these counties represented 38 percent of the industry output.
** See Reference (27) for a summary of comments made at the Public Meeting.
7-52
-------
were "tme" causes of higher production costs: age of plant, wage rate,
tax rates, and age of labor force. It was asserted by the AISI representa-
tives that these factors need to be included in the cost functions so that
included air pollution variables do not reflect the impact of these
variables on production cost. Mathtech authors acknowledged that omitted
variables can lead to statistical bias in the pollution coefficient, but
felt that the manufacturing sector analysis adequately controlled for most
of the other factors influencing production cost so that the relationship
observed between air pollution and production cost in the study suggested a
cause-and-effect relationship. In addition to the results of their
analysis, the Hathtech representatives noted that physical evidence from
other studies, and supporting anecdotal evidence collected as part of their
study, were both consistent with the kinds of air pollution effects
observed in their analysis. This adds further credence to the results
obtained in the MTM study.
The second major area of concern involved the quality of the data.
Members of the review panel and the AISI representative commented on the
small number of observations available in the analysis, and the possibility
of measurement error associated with the aggregate data and proxy
variables.
Finally, panel members were concerned with the magnitude of the
benefits estimated for SIC 344. The implied effect of a 1 |ig/m3 increase
in TSP is estimated to be a $90 to $340 increase in total production costs
for a firm. This is only 0.01 to 0.03 percent of average total costs.
However, this one industry accounted for a large fraction of the total
benefits estimated in the overall Hathtech study.
At the suggestion of the review panel, additional investigation of SIC
344 was undertaken. One check involved informal interviews with managers
of some plants in SIC 344. The purpose of the interviews was twofold. The
first was to look for evidence of damage and/or behavioral adjustments due
to PM deposition. The second objective was to determine whether plant
7-53
-------
managers perceived ambient PM deposition as affecting either their produc-
tion processes or their production costs.
Most of the managers who were interviewed reported no impacts on
either production processes or costs. However, air pollution effects may
not be readily perceptible because of the predicted small changes in cost.
For example, it was suggested that the effects of temperature and moisture
may be larger than those of PM, and yet it was found that managers did not
respond differently in cities with quite different levels of temperature
and humidity. In addition, there was evidence at some plants of prior
adjustments made to prevent metal corrosion and contamination. These
included the use of coatings on exposed metal surfaces, indoor storage of
metal inventories, and surface cleaning before painting or welding.
However, these activities might be undertaken for a variety of reasons so
that they could not be attributed exclusively or primarily to ambient PH.
The overall results of the interviews were thus inconclusive.
As a second check, additional sensitivity analyses were performed for
SIC 344. In one test, an outlying observation was eliminated from the
sample. When the equations were re-estimated, the implied effect of a
change in TSP increased by about 30 percent, but the estimated coefficients
for TSP were not as statistically significant.* The implication of this is
that the expressions for SIC 344 are sensitive to sample composition.
Thus, some caution is warranted in interpreting benefits estimates
generated from the MTM cost equations.
Because of the finding for SIC 344, a similar outlier analysis was
conducted for the other industry, SIC 354. In this case, elimination of
one outlying data point resulted in a decrease in the predicted effect of
TSP on production costs, and a decline in statistical significance for the
estimated TSP coefficients. However, examination of the data for the
* The re-estimated coefficients were significantly different from zero at
the 10.6 percent level, while the coefficients with the outlier were
significant at the 1 percent level.
7-54
-------
outlying observation suggested that it was valid information, and thus the
original version of the model was retained.
In terms of the evaluation criteria, the following observations can be
made concerning the Mathtech manufacturing sector analysis:
• Theoretical basis — Economic benefits are estimated as
differences in total production cost evaluated at two
levels of ambient PM. As a consequence, the model yields
estimates consistent with the theoretical definition of
benefits. Furthermore, the structure of the model takes
into account behavioral adjustments on the part of affected
firms.
• Consideration of relevant variables — A variety of
economic and non—economic variables are controlled for in
the cost relationships. However, data limitations pre-
vented controlling for some possibly relevant factors such
as local property taxes. Omission of these variables might
bias the coefficients of the air pollution variables if
these excluded variables tend to be correlated with air
pollution.
• Quality of input data — Confidentiality restrictions pre-
clude the use of disaggregate plant data. Consequently,
data for 3-digit SICs at the county level are used. Even
with this level of aggregation, lack of appropriate data
limit the number of industries which can be analyzed. Of
the six industries for which data are collected, only one
has more than SO observations. In terms of specific
variables, proxies are created for several of the desired
variables, including the input price data series. This may
introduce measurement error into the analysis.
• Legitimacy of Empirical Approach — The MTM analysis uses
econometric techniques that are appropriate for the model.
The functional form chosen for the cost function analysis
is the transcendental logarithmic function (translog).
This is a very general function which imposes fewer main-
tained assumptions than other forms (e.g., Cobb-Douglas).
• Transferabilitv of Model — As with the MTH model, the
expression used to evaluate benefits is complex. Much of
the data needed to evaluate benefits for this study's
scenarios must come from the MTM data base.
7-55
-------
Sm«»«rv of Models
In this subsection, models which describe the impact of reduced
soiling on household and manufacturing plant decisions have been reviewed.
To provide a point of reference for judging alternative studies, a set of
five evaluation criteria were identified. These criteria help define the
desired structure of any defensible benefits analysis model.
Another way in which studies can be evaluated is in terms of the
biases implicit in the development and application of the models.* In the
discussion of each of the four studies that have been reviewed in detail,
the implications of specific assumptions have been noted. Some of the
major biases associated with the models are summarized in Table 7-9.
Because each model contains biases that will impact benefits in an unknown
direction, it is not possible to determine conclusively whether any parti-
cular study will provide an over- or underestimate of "true" benefits.
Consequently, an important element in reporting benefits should be a range
of estimates reflecting, in part, this degree of uncertainty. It should be
noted that the types of biases listed in Table 7-9 refer only to biases
inherent in the models. There are also biases possible when the models are
used to provide estimates of national benefits. These biases are discussed
in the next subsection.
In the household sector, three studies were evaluated. Because of
differences in study design and model bias it is difficult to make
meaningful comparisons across studies. For example, the Cummings model
accounts for changes in out-of-pocket expenditures and the opportunity cost
of labor time in cleaning, the Watson and Jaksch model yields benefits
estimates of a pure utility type; and the Hathtech household model captures
only the benefits related to out-of-pocket expenditures. At first glance,
it might be expected that estimates derived from the Cummings study should
provide an upper-bound plausibility check for the Mathtech household
estimates. However, this need not be true because of the different biases
* In fact, many of the criteria reflect concern with the question of bias.
7-56
-------
Table 7-9
BIASES IN MODELS OF SOILING STUDIES
Model
Description of Bias
Expected Impact on
Benefits
Cummings et al.
-j
i
en
1. Not all plausible cleaning tasks considered.
2. Remaining measurement error in BAH data
(e.g., unit cost data).
3. Consideration of differentiated cleaning
tasks.
4. The use of a measure of the opportunity cost
of leisure may be inappropriate since other
activities that are equally undesirable may
be substituted for cleaning when air quality
improves.
5. Adjustments to task frequency data of BAH.
6. Use of linear specifications for frequency,
labor and damage equations.
7. Exclusion of possible relevant explanatory
factors in various specifications.
8. Calculation of benefits not based on varia-
tions in demand and supply curves.
Underestimate of
benefits
Unknown
Overestimate of benefits
(possible jointness)
Overestimate of benefits
Overestimate of benefits
Unknown
Unknown
Underestimate of
benefits
(cont inued)
-------
Table 7-9 (Continued)
Model
Description of Bias
Expected Impact on
Benefits
Watson & Jaksch
Ul
00
1. Uncorrected BAH data employed in analysis.
2. Not all plausible cleaning tasks considered.
3. Consideration of differentiated cleaning
tasks.
4. Assumption of constant unit elastic demand.
5. Assumption of linear, upward sloping
marginal cost curve with multiplicative
shift.
6. Not all adjustment opportunities explicitly
recognized.
7. No adjustment for reallocation of labor time.
8. Estimates of a derived from limited number of
studies. Quality of estimates questionable.
9. Benefits calculated with consumer surplus
measure.
Unknown
Underestimate of
benefits
Overestimate of benefits
(due to jointness)
Possible overestimate
of benefits
Unknown
Underestimate of
benefits
Underestimate of
benefits
Unknown
Small overestimate
of benefits
(continued)
-------
Table 7-9 (Continued)
Model
Description of Bias
Expected Impact on
Benefits
MTII
~J
Ui
1. Problems associated with input data (e.g.,
aggregate data for SMSAs, measurement error,
small number of observations).
2. Limited consideration of possible relevant
explanatory variables.
3. No consideration of location adjustment
opportunities.
4. No consideration of labor-leisure tradeoff
possibilities.
5. No adjustments for reallocation of labor
time.
6. Model does not capture benefits of a pure
utility type.
7. Stone-Geary utility function assumed.
8. Separability assumptions imposed on groups
of market goods.
Unknown
Unknown
Underestimate of
benefits
Underestimate of
benefits
Underestimate of
benefits
Underestimate of
benefits
Unknown
Unknown
(continued)
-------
Table 7-9 (Continued)
Model
Description of Bias
Expected Impact on
Benefits
MTM
-j
i
1. Not all manufacturing sectors treated due to
data limitations.
2. Use of aggregate data.
3. Approximate data on the prices and
quantities of inputs and outputs.
4. Exclusion of potentially relevant variables.
5. Translog cost function assumed.
6. No adjustments permitted for plant
relocation.
Underestimate of
benefits
Unknown
Unknown
Unknown
Unknown
Underestimate of
benefits
-------
associated with each, model. As a consequence, each of the three studies
must be viewed-as contributing individually to the overall stock of
knowledge available for estimating benefits from reduced soiling.
Based on the reviews in this subsection, it is possible to make
subjective judgments on the relative merits of each of the studies.
Because of the model design and the manner in which benefits are calcu-
lated, we believe the MTH study should be weighted most heavily, with the
other two studies providing additional information on the probable range of
soiling benefits. Furthermore, it is our belief that the estimates derived
from the MTH model will likely be conservative estimates of the benefits
from reduced soiling in the household sector. We are less sanguine about
the relationship to "true" benefits for the other two household soiling
studies.
In the manufacturing sector, only one study is available for benefits
estimation. Based on the evaluation in this section, we believe that it is
appropriate to view benefits numbers generated from the MTM model as upper
bounds for the "true" benefit numbers in SIC 344 and SIC 354. Note,
however, in the context of the entire manufacturing sector, these estimates
will likely be conservative estimates of the total benefits from reduced
soiling in the sector.
The above observation is also pertinent with respect to other sectors
of the economy. In particular, benefits from reduced soiling are not
reported for the commercial, government, and institutional sectors. To our
knowledge, no studies have been completed which permit the calculation of
benefits in these sectors. Consequently, a large part of total benefits
from reduced soiling may be uncounted.
7-61
-------
BENEFITS CALCULATIONS
Int r oduct i on
In this subsection, benefits of reduced soiling are calculated for the
household and manufacturing sectors. Benefits in the household sector are
estimated using three studies: Cummings et al. (5), Watson and Jaksch (6),
and the Mathtech household model (7). Benefits estimates in the manu-
facturing sector are developed entirely from the Mathtech manufacturing
model (8).
Benefits estimates are calculated on a county-by-county basis. The
counties included in the benefits analysis are determined as part of a
research effort conducted by another EPA contractor. This other work
examines the costs of implementing alternative PH control strategies in
response to various air quality standards. Any county that is expected to
experience a reduction in ambient levels of PH because of the PH control
programs is included in the benefits analysis.
Although the benefits calculations are performed with county-specific
data, for presentation purposes, benefit estimates are reported by EPA
Administrative Region. The standard ten-region breakdown is employed.
Reporting estimates in this fashion provides an indication of the distribu-
tion of benefits.
Table 7-10 summarizes the scenarios to be evaluated in this sub-
section. As in the other sections of this report, the benefits from
reduced soiling associated with the standards shown in the table are pre-
sented as discounted present values in 1982, in 1980 dollars. Estimates of
annualized benefits are also provided. A discount rate of 10 percent and a
time stream of benefits of 7 years (PH10 standards) or 9 years (TSP
standards) are also used. A complete description of such issues as non-
attainment status for counties, maintenance of ambient levels after control
strategy implementation, and the impact of emissions growth on air quality
is contained in Section 9.
7-62
-------
Table 7-10
AIR QUALITY SCENARIOS
Standard
Primary
Primary
Primary
Secondary
Primary
Secondary
Pollutant
PM10
PM10
PM10
PM10
TSP
TSP
Level
70
250
55
250
55
150
55
75
260
150
Averaging Time
Annual arithmetic mean
24-hour expected value
Annual arithmetic mean
24-hour expected value
Annual arithmetic mean
24— hour expected value
Annual arithmetic mean
Annual geometric mean
24-hour second high
24— hour second high
Implementation
Date
1989
1989
1989
1989
1989
1989
1989
1987
1987
1987
Household Sector
Coning* et al. —
In Cummings et, al. (5), the damage function is reported as:
251.43 + 6.63P
(7.11)
where D is annual expenditures per household in 1980 dollars and P is the
annual arithmetic mean level of TSP. As reported in our review of the
Cummings study, this damage estimate represents both out-of-pocket expenses
and a measure of the opportunity cost of labor in cleaning activities.
As part of our review of this study, an attempt was made to replicate
the results of Cummings. Data were presented in sufficient detail in the
7-63
-------
Cummings report to permit reconstruction of the required variables. One
reason this exercise was undertaken was that Cummings et a_l. do not report
the standard error for the pollution coefficient. In order to obtain a
range of benefit estimates, our plan was to perturb the pollution coeffi-
cient by plus and minus two—standard deviations. If one is willing to
accept the simple regression form used by Cummings, this perturbation
yields an estimate of the 95 percent confidence interval for benefits from
reduced soiling.
The data required for replicating the Cummings et al. equation can be
developed from their Tables 7 and 20. The following steps were carried
out:
Per-household estimates of cost were identified by task and
pollution zone for the households that hired help as well
as households that performed the tasks themselves. Our
estimates for this step coincided with those reported in
Cummings.
A weighted average, based on the distribution of the two
types of households, was calculated. This gave an average
per household expenditure for all houses that performed the
tasks. This weighted average was again estimated by task
and pollution zone.
In order to identify the average expenditures for all
households in a pollution zone, the weighted average
estimates were multiplied by a factor representing the
proportion of households in the zone that performed the
task. In several cases, the Cummings data was such that
greater than 100 percent of households performed a task.
In these instances, an upper bound of 100 percent was
imposed.
Damages were aggregated across tasks by pollution zone to
obtain an estimate of total expenditures per household.
Using this procedure, the following damage equation was estimated:
D = 131.28 + 8.85 ' P (7.12)
(143.67) (1.56)
7-64
-------
Standard errors are given in parentheses. This expression differs from the
one estimated in Cummings et al. Unfortunately, except for the 100 percent
constraint imposed in Step 3, it is not possible to identify other causes
for the observed differences. With the exception of the first step
described above, no additional tables or descriptions that appear in
Cummings are useful for making comparisons.
Equation (7.12) is used in this subsection to calculate benefits. The
point estimate of marginal damages is 8.85, with a two-standard deviation
range of about 5.72 to 11.98. These numbers represent the dollar value
(1980 dollars) of a 1 (ig/m change in annual average TSP levels. This two-
standard deviation range is used to identify a confidence interval for the
benefits estimates reported below. Note the marginal damage estimate
reported in Cummings (6.63) falls within this range. On the other hand,
this range does not include the marginal damage estimate derived from the
Hichelson and Tour in (13) study. Their value (adjusted to 1980 dollars) is
$2.50. This discrepancy is to be expected, however, since the Cummings
study includes an estimate of the opportunity cost of labor in cleaning
activities in their damage relation.
The measure of pollution used in the Cummings study is the annual
geometric mean of TSP. More precisely, the midpoint of the range of TSP in
each of the four pollution zones of Philadelphia is associated with the
cost data for that zone. Thus, the units of the TSP coefficient in
Equation (7.12) reflect a dollar value per unit of average TSP annual
geometric mean concentrations. That is, the index of pollution for the
analysis is an average value. Since the air quality data generated for use
in this study rely on design value monitors that likely reflect the highest
values in a given county, it is appropriate to adjust the air quality data
to provide a representation of population exposure in the county that is
more consistent with the structure of Equation (7.12). This adjustment
procedure is discussed in detail in Section 9.
Tables 7-11 through 7-16 present the benefits estimates for each of
the scenarios listed in Table 7-10. These benefits represent the benefits
7-65
-------
Table 7-11
ESTIMATED BENEFITS FOR: CUMMINGS SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
8.2
289.9
341.0
1768.1
699.0
100.6
235,
2263,
.2
.5
Point
Estimate
0.0
.7
8
329.1
12.
447.
526.6
2730.9
1079.7
155.4
363.2
3496.0
508.3
Maximum
0.0
17.2
605.7
712.3
3693.7
1460.3
210.2
491.3
4728.5
687.5
Total U.S.
6034.7
9320.7 12606.6
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
2415.6
3730.9
5046.1
7-66
-------
Table 7-12
ESTIMATED BENEFITS FOR: CUMMINGS SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
83.2
107.5
565.0
797.6
2613.7
1140.8
270.8
501.5
4384.0
471.7
10935.7
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
128.5
166.0
Maximum
173.9
224.5
1180.3
1666.2
5460.
2383
565.6
1047.6
9158.2
985.3
.1
.1
Total U.S.
16890.2 22844.8
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
4377.3
6760.8
9144.2
7-67
-------
Table 7-13
ESTIMATED BENEFITS FOR: CUMMINGS SOILING STUDY
Benefits Occurring Between 1989 and 199S
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
282.2
202.9
814.9
1047
2998
1365
389
754.8
4954.9
838.6
13648.4
435.8
313.4
1258.6
1617.9
4631.2
2108.3
600.8
1165.8
7652.9
1295.3
589.5
423.9
1702.3
2188.3
6263.8
2851.6
812.6
1576.8
10350.8
1751.9
21080.1 28511.7
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
5463.2
8437.9 11412.6
7-68
-------
Table 7-14
ESTIMATED BENEFITS FOR: CUMMINGS SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
83.2
107.5
565.0
797.6
2601.5
1116.3
269,
501,
4383,
Point
Estimate
128.5
166.0
872.7
1231.9
4018.1
1724.2
416.
774.
6770,
439.2
678.3
Maximum
173.9
224.5
1180.3
1666.2
5434.6
2332.0
562.7
1047.6
9157.6
917.5
Total U.S.
10864.9
16780.9 22696.9
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
4349.0
6717.0
9085.1
7-69
-------
Table 7-15
ESTIMATED BENEFITS FOR: CUMMINGS SOILING STUDY
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AAM/260 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
287.0
261.3
1322.8
1598.6
4835.1
1919.3
671.6
932.4
8023.7
917.2
20769.0
Point
Estimate
443.3
403.6
2043.0
2469.0
7467.8
2964.4
1037.3
1440.2
12392.6
1416.6
Maximum
599.6
545.9
2763.3
3339.4
10100.5
4009.4
1402.9
1947.9
16761.5
1916.1
32077.7 43386.4
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
5808.0
8970.5 12133.0
7-70
-------
Table 7-16
ESTIMATED BENEFITS FOR: CUMMINGS SOILING STUDY
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
853.5
522.7
2007.8
2086.1
5858.1
2304.4
1178.4
1333.8
9149.9
1540.9
26835.5
Point
Estimate
1318.2
807.3
3101.0
3222.1
9047 . 8
3559.2
1820.0
2060.0
14132.0
2379.9
Maximum
1783.0
1091.9
4194.2
4358.0
12237.5
4814.0
2461.6
2786.3
19114.1
3218.9
41447.5 56059.4
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
7504.5
11590.8 15677.0
7-71
-------
that would be achieved when all counties included in the analysis are in
compliance with the standard for all years under consideration.* The first
three tables present estimates for alternative PM10 primary standards. The
order of presentation for the tables is from most lenient to most strin-
gent. The minimum and maximum estimates are obtained by perturbing the
coefficient in Equation (7.12) by minus and plus two standard deviations,
respectively. From the tables, it can be seen that almost two-thirds of
total U.S. benefits will accrue in the East North Central and South Pacific
regions.
Table 7-14 presents estimates for a PM10 secondary standard of 55
^
Hg/m . The table reports the total benefits associated with attainment and
maintenance of this standard. To identify the incremental benefits of the
secondary standard conditional on a primary standard being attained, it is
necessary to net out the benefits of the primary standard. For example, if
the 55 |ig/m secondary standard is associated with the 70/250 |ig/m primary
standard, then the incremental benefits (discounted present value) of the
secondary standard would be $7.5 billion for the total U.S. point estimate
(i.e., 16.8 - 9.3). Although reduced soiling is a welfare benefit, it is
important to note that positive welfare benefits will be generated in the
attainment of the primary standard. This occurs even though the primary
standard is based only on health considerations.
Tables 7-15 and 7-16 present benefits estimates for the current TSP
primary and secondary standards. As before, to identify the incremental
benefits of the secondary standard, it is necessary to net out the benefits
generated through attainment of the primary standard.
Finally, Table 7-17 shows the benefits that accrue under a 70/250 PM10
primary standard when all counties are not in attainment with the standard
throughout the 1989-1995 time horizon. This can occur because available
means of controlling emissions are exhausted prior to standard attainment
* In the language of Section 9, these benefits represent "B" scenario
benefits.
7-72
-------
Table 7-17
ESTIMATED BENEFITS FOR: CUMMINGS SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type A PM10 - 70 AAM/250 24-hr.
Point
Federal Administrative Region Minimum Estimate Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0 0.0 0.0
8.2 12.7 17.2
279.3 431.4 583.5
274.9 424.5 574.2
1267.2 1957.2 2647.1
442.7 683.7 924.7
'89.9 138.8 187.7
220.9 341.2 461.5
1221.3 1886.3 2551.3
103.7 160.2 216.7
Total U.S. 3908.0 6036.0 8163.9
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
1564.3
2416.1
3267.8
7-73
-------
and/or because of emissions growth. This table can be compared to Table 7-
11 where all counties were assumed to be in compliance with the same 70/250
PM10 standard. As expected, the benefits estimates in Table 7-11 exceed
those shown in Table 7-17.
Watson, and Jmkseh —
In the Watson and Jaksch study, benefits estimates are calculated by
task and pollution zone. The expression used to calculate benefits is:
ACWi- = o In
P2i
p-^- ' X41 (7.13)
LPiJ
where ACW-. is the change in welfare in task i for pollution zone j.
?2- is the final level of pollution in zone j.
P,. is the initial level of pollution in zone j.
a is a factor estimated from damage function studies.
X— is the outlay for task i in zone j .
For purposes of extrapolation, it does not seem warranted to estimate
benefits at such a disaggregate level as task and zone. Consequently, the
benefits derived below are calculated on a county basis only and for the
aggregate of tasks. In essence, the subscript i can be deleted in Equation
(7.13), and the j index refers to a county instead of a pollution zone.
There are five bits of information needed to determine benefits. The
two levels of TSP or PM are provided to us through the cost of control
study. These data are representative of the highest value monitors. As
with the Cummings study, average TSP indices are used in WJ. Consequently,
the design value data from the cost analysis must be adjusted to provide a
more consistent match-up with the WJ model.
The third data element that is required is an .estimate of a. This
factor is taken from WJ, and three values are available: 0.56, 1.0, and
7-74
-------
2.0. These three estimates provide the variation needed to obtain a
probable range of benefits for the Watson and Jaksch study. The fourth
data element required is an estimate of per household expenditures.
Earlier in this review, it was mentioned that because of data difficulties
with, the Booz, Allen and Hamilton data base, it would be advantageous if
the Cummings "cleaned" data could be used. Unfortunately, upon closer
examination this did not appear to be feasible. The principal constraint
is that only three of the eight cleaning/maintenance tasks examined in
Watson and Jaksch correspond to those tasks included in Cummings.
Using data provided in Watson and Jaksch, a per-household expenditure
estimate of $155.20 was calculated. This figure is a weighted average
across pollution zones for all eight cleaning/maintenance tasks. Since
this value is in 1971 dollars, it was adjusted to 1980 dollars by the U.S.
average Consumer Price Index for home materials repair. In 1980 dollars,
the annual per household expenditures are estimated to be $333.68.
The final variable needed to calculate benefits is the number of
households per county. This variable and the necessary growth rates are
taken from Reference (29) and Reference (30), respectively.
The final form for estimating benefits from the Watson and Jaksch
study is:
-ACW. = o In
P2j
• X • Nj (7.14)
where X. is the per-household outlays in county j.
N. is the number of households in county j.
The other variables are as defined previously, except that j indexes a
county rather than a pollution zone.
Tables 7-18 through 7-23 present the benefits estimates derived from
the Watson and Jaksch model for the scenarios described in Table 7-10. As
7-75
-------
Table 7-18
ESTIMATED BENEFITS FOR: WATSON AND JAXSCH SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
I
New England
N.Y.-N.J.
Middle Atlantic
South. Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
0.0
5.6
107.5
146.3
905.2
321.3
42.6
106.5
750.8
151.7
2537.4
.9
.2
.5
.7
0.0
10.0
191,
261,
1616.
573.
76.1
190.1
1340.7
270.9
4531.1
.9
.5
,0
0.0
20.0
383.
522.
3233.
1147.4
152.1
380.2
2681.3
541.8
9062.2
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount
Annualized Benefits
Between 1989 and 1995
Total U.S.
1015.7
1813.7
3627.4
7-76
-------
Table 7-19
ESTIMATED BENEFITS FOR: WATSON AND JAKSCH SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
38.8
69.5
242.5
386.0
1373,
570.
.1
.3
126.8
219.
1683,
232.3
Point
Estimate
69.4
124.1
433,
689,
1018,
226,
2452.0
391.2
3006.5
414.8
Maximum
138.7
248.3
866.0
1378.7
4904.
2036.
453,
,0
.7
,0
782.4
6013.1
829.6
Total U.S.
4942.1
8825.2 17650.5
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount
Annualized Benefits
Between 1989 and 1995
Total U.S.
1978.2
3532.5
7065.1
7-77
-------
Table 7-20
ESTIMATED BENEFITS FOR: WATSON AND JAKSCH SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Eegion Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
193.9
131.1
391.2
531.8
1579.5
711.3
193,
339.
1992.2
476.1
.1
.7
346.3
234.1
698.6
949.7
2820.5
1270.2
344.8
606.6
3557.5
850.2
692.5
468.2
1397.1
1899.3
5640.9
2540.4
689.5
1213.2
7115.0
1700.5
Total U.S.
6539.9
11678.3 23356.7
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount
Annualized Benefits
Between 1989 and 1995
Total U.S.
2617.8
4674.6
9349.1
7-78
-------
Table 7-21
ESTIMATED BENEFITS FOR: WATSON AND JAKSCH SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.I.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
38.8
,5
,5
69,
242.
386.0
1366.4
555.4
126.2
219,
1683
214.6
Point
Estimate
69.4
124.1
433.0
689.3
2440.0
991.9
225.4
391.2
3006.3
383.2
Maximum
138.7
248.3
866.0
1378.7
4879,
1983
450.
782.4
6012.6
766.4
.9
.7
.7
Total U.S.
4902.1
8753.7 17507.4
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount
Annualized Benefits
Between 1989 and 1995
Total U.S.
1962.2
3503.9
7007.8
7-79
-------
Table 7-22
ESTIMATED BENEFITS FOR: WATSON AND JAKSCH SOILING STUDY
•Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AGM/260 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
II
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
151.9
163.2
622
827
2586
1015.3
336.5
413
3380
.2
.7
Total U.S.
483.1
9979.6
Point
Estimate
271.3
291.4
1112.0
1476.8
4617.9
1813.0
601.0
737.9
6036.9
862.7
Max imam
542.5
582.7
2223.9
2953.5
9235.8
3626.1
1201.9
1475.7
12073.8
1725.5
17820.8 35641.6
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount
Annualized Benefits
Between 1987 and 1995
Total U.S.
2790.8
4983.6
9967.2
7-80
-------
Table 7-23
ESTIMATED BENEFITS FOR: WATSON AND JAKSCH SOILING STUDY
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
598.0
335.0
1045.2
1150.2
3172.4
1242.4
670.
657
4143.3
938.3
.3
.4
Point
Estimate
1067.9
598.2
1866.4
2054.0
5665.0
2218.6
1197.0
1174.0
7398.8
1675.5
Maximum
2135.8
1196.4
3732.8
4107.9
11330.0
4437.2
2394.0
2347.9
14797.6
3351.0
Total U.S.
13952.6
24915.3 49830.6
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount
Annualized Benefits
Between 1987 and 1995
Total U.S.
3901.8
6967.6 13935.1
7-81
-------
before, these benefits represent the benefits associated with complete
attainment and maintenance of the standards in all counties. The ordering-
of the tables is also the same. In particular, the first three tables
represent alternative PM10 primary standards, Table 7-21 shows the benefits
for the PM10 secondary standard, and Tables 7-22 and 7-23 present benefit
estimates for the current TSP primary and secondary standards, respec-
tively. Finally, Table 7-24 shows the benefits that accrue with the 70/250
PM10 scenario when some of the counties remain in nonattainment due to
emissions growth.
A comparison of the benefits derived from the Cummings and Watson and
Jaksch models reveals that the Watson and Jaksch estimates are lower.
However, because of model differences, it is not possible to compare the
studies in a definitive manner. Recall that the Cummings study accounts
for benefits derived from reduced out-of-pocket expenditures as well as the
value of reduced labor time in cleaning activities, while the Watson and
Jaksch model identifies adjustments in the demand for cleanliness. These
adjustments reflect actions based on "utility" considerations as well as
changes in the cost of cleanliness.
Mmthtech Household Model (MB) —
The basic analysis in MTH is limited to 24 SMSAs, with these SMSAs
accounting for approximately 30 percent of the total U.S. population in
1976. Since the demand systems estimated in MTH use SHSA-specific price,
income, and demographic data, extrapolation to other counties of the U.S.
requires assumptions on these data. For example, since price data are not
available for many parts of the country, proxies must be created. The
proxies take the form of regional averages based on the data available in
the basic study.
The specific assumptions made in performing the extrapolation include:
• Certain data are assigned to counties based on the region
in which the county is located. In particular, the
following steps are taken:
7-82
-------
Table 7-24
ESTIMATED BENEFITS FOR: WATSON AND JAKSCH SOILING STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type A PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
I
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
5.6
103.1
111.0
599.8
192.0
36,
95,
387.0
41.0
.6
.9
Total U.S.
1571.9
Point
Estimate
0.0
10.0
184.1
198.2
1071.0
342.9
65.3
171.3
691.0
73.2
2807.0
Maximum
0.0
20.0
368.2
396.4
2142.1
685.8
130.7
342.6
1382.0
146.3
5614.1
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount
Annualized Benefits
Between 1989 and 1995
Total U.S.
629.2
1123.6
2247.2
7-83
-------
— The country is divided into two major areas:
Northeast plus North Central and South plus West.
— Averages by region are computed from the 24 SMSA data
for the following items:
a) 30-year average temperature
b) family size
c) average annual percent change in the all-item
Consumer Price Index
d) average percent of total consumption expenditures
in the SHSA data
e) disaggregate and aggregate price sets developed
in the basic analysis of MTH.
• Certain data are assigned to counties based on county- or
state-level data. These data include:
— Air quality data
—• Baseline county population numbers are taken from
Reference (29). Conversions to household data are
made by dividing by the regional family size values.
— County population projections are taken from Reference
(30).
— State income projections for 1985 and 1990 (current
1972 dollars) are obtained from Reference (31).
• Certain data on assumptions are relevant for all counties.
These include:
— The parameters of the demand models in MTH.
—• The air quality scenario.
— The parameters of the benefits calculations (i.e.,
discounted present value in 1982, in 1980 dollars; 10
percent discount rate; 7- or 9-year steam of benefits.
Given these assumptions, benefit estimates are developed for each
county that experiences a reduction in PH concentrations. Note that the
extrapolation from the original 24 SMSAs in the MTH study is only
geographic. Recall that the goods included in the basic MTH analysis
account for only about 40 percent of current consumption expenditures.
With available information, it is not possible to extend the scope of the
MTH study beyond this subset of goods.
7-84
-------
Tables 7-25 through 7-30 present the benefit estimates of the MTH
model for the scenarios described in Table 7-10. Since the measure of TSP
used in the MTH analysis is the maximum of all site 24-hour average second-
high readings in a county, it is not necessary to adjust the design value
data to characterize population exposure.
The ranges of benefits reported in the tables reflect two plausible
sources of uncertainty in the MTH benefits equations. First, since the
estimators for TSP are random variables with known standard error, values
of the coefficient plus or minus two standard deviations from the estimated
coefficient provide a range that reflects the stochastic nature of the
estimator. Note that this procedure does not yield a 95 percent confidence
interval for the MTH model. Because of the across-equation constraints
implied by the estimated demand systems, a true confidence interval would
have to account for covariances among the variables in the system. The
procedure described above does not do this. Analytical problems associated
with developing the statistically-preferred measure prevented its use for
this review. Nevertheless, the less rigorous approach does provide an
*
indication of the uncertainty associated with the pollution estimator.
The second way uncertainty in the MTH model is reflected in the
benefits calculations is through variations in prices. In the original MTH
study, price data were available for only 24 SMSAs. An extrapolation to
other counties in the country was made by defining two average price sets
for the United States. One price set is constructed from the prices of
those MTH SMSAs that are in the West and South; the other price set is the
average of prices of SMSAs in the Northeast and North Central parts of the
country. Taken together, these average prices were used to derive the
point estimates of benefits in Tables 7-25 through 7-30.
The minimum and maximum price set estimates are developed by using
price data from the MTH SMSAs that yield the lowest and highest marginal
valuations per unit change in PM levels. The low price set is associated
with Atlanta; the high price set is associated with New York City.
7-85
-------
Table 7-25
ESTIMATED BENEFITS FOR: MATHTECH HOUSEHOLD EXPENDITURE STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
Maximum
0.0
0.2
4.7
5.7
35.8
10.0
1.4
3.5
31.5
6.6
0.0
1.7
35.4
40.3
271.0
72.0
10.2
25.1
231.0
48.0
0.0
3.2
67.4
76.7
513.7
136.5
19.4
47.5
440. 9
91.2
Total U.S.
99.4
734.9
1396.5
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
39.8
294.2
559.0
7-86
-------
Table 7-26
ESTIMATED BENEFITS FOR: MATHTECH HOUSEHOLD EXPENDITURE STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
Maximum
1.1
2.1
8.7
12.6
56.1
15.8
3.7
7.0
60.4
9.2
8.4
16.1
65.5
89.1
423.3
113.1
27.7
50.4
440.3
66.5
15.9
30.5
124.0
168.4
800.0
213.4
52.3
95.3
835.4
126.0
Total U.S.
176.7
1300.3
2461.1
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
70.7
520.5
985.1
7-87
-------
Table 7-27
ESTIMATED BENEFITS FOR: MATHTECH HOUSEHOLD EXPENDITURE STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
.3
,5
Total U.S.
4.7
4.1
13.0
15.9
63,
19.
5.3
10.0
68.9
16.1
220.7
Point
Estimate
35.0
30.5
97.0
112.5
477.8
140.0
39.6
71.6
501.9
116.1
1621.9
Maximum
65.8
57.6
183.4
212.3
902.8
263.9
74.8
135.1
951.1
219.1
3065.9
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
88.3
649.2
1227.2
7-88
-------
Table 7-28
ESTDIATED BENEFITS FOR: MATHTECH HOUSEHOLD EXPENDITURE STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
Maximum
1.1
2.1
8.7
12.6
55.8
15.5
3.7
7.0
60.4
8.3
8.4
16.1
65.5
89.1
421.4
111.1
27.5
50.4
440.3
59.8
15.9
30.5
124.0
168.4
796.4
209.8
52.0
95.3
835.3
113.3
Total U.S.
175.2
1289.5
2440.8
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
70.1
516.2
977.0
7-89
-------
Table 7-29
ESTIMATED BENEFITS FOR: MATHTECH HOUSEHOLD EXPENDITURE STUDY
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AAM/260 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
Maximum
3.4
4.5
19.9
23.0
103.3
28.1
8.9
12.1
108.9
17.3
25.8
33.7
148.2
161.4
in. 6
200.2
66.9
86.5
788.1
124.6
48.8
63.7
280.0
304.1
1467.6
377.0
126.4
163.2
1489.9
235.8
Total U.S.
329.4
2413.1
4556.4
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annnalized Benefits
Between 1987 and 1995
Total U.S.
92.1
674.8
1274.2
7-90
-------
Table 7-30
ESTIMATED BENEFITS FOR: MATHTECH HOUSEHOLD EXPENDITURE STUDY
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Total U.S.
431.1
Point
Estimate
102.7
69.8
224.7
218.2
907.0
247.6
119.9
123.3
937.1
204.0
3154.4
Maximum
193.3
131.7
423.9
410.1
1711.4
466.0
226.0
232.0
1767.6
384.3
5946.4
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
120.6
882.1
1662.9
7-91
-------
With the two adjustments for uncertainty, there are a variety of ways
to define the overall range of benefits. We have chosen the option that
yields the widest range. That is, the low estimate reflects the low price
set in combination with a pollution coefficient that is two standard devia-
tions below the estimated coefficient; the point estimate reflects the
average (two-region) price set in combination with the estimated coeffi-
cient; and the high estimate reflects an evaluation of benefits for the
high price set and an estimator two standard deviations above the estimated
coefficient.
The benefit calculations shown in Tables 7-25 through 7-30 are carried
out with SO* levels assumed constant at the current primary standard level
(260 jig/m ). In the MTH analysis, SO- is part of an interdependent system
of equations so that its value, even if unchanging, can influence alloca-
tion decisions. Consequently, some measure of SOj must be included in the
extrapolation procedure. Because S02 data are not available for many of
the counties included in the PH10 scenarios, the SO- concentrations are
a
fixed at 260 ug/m for all counties. To test the sensitivity of the
benefits calculations to this assumption, estimates were derived for other
S02 levels. The estimates for the alternative SO- levels were quite close.
For example, with all counties presumed to be at 100 ug/m , the estimate of
total benefits corresponding to the point estimate in Table 7-25 is $733.8
million. This represents a difference of only $1.1 million.
Finally, Table 7-31 shows the benefits associated with the 70/250 PM10
standard when some counties remain in nonattainment after the contol
strategy. As before, the benefits for this option are about 65 percent of
the benefits when all counties are brought into attainment.
Synthesis of Household Sector Benefits —
Tables 7-11 through 7-31 of this section report the benefits accruing
to the household sector from reduced soiling. For each study and each
scenario, a range of benefit estimates is presented. This range reflects.
7-92
-------
Table 7-31
ESTIMATED BENEFITS FOR: MATHTECH HOUSEHOLD EXPENDITURE STUDY
Benefits Occurring Between 1989 and 1995
Scenario: Type A PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
Maximum
0.0
0.2
4.5
4.4
25.0
6.1
1.2
3.0
15.9
2.1
0.0
1.7
34.2
31.1
189.7
44.1
9.1
21.4
116.8
15.4
0.0
3.2
65.0
59.3
360.2
83.8
17.3
40.5
223.6
29.7
Total U.S.
62.5
463.5
882.7
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
25.0
185.5
353.3
7-93
-------
in part, the degree of uncertainty in the different studies due to model
biases and/or methodological design.
The level of benefits estimates in the various tables may also be
affected by biases that occur in the extrapolation to national benefits
estimates. Examples of these types of biases are shown in Table 7-32. As
can be seen from the table, most of the identified biases are common to
each of the studies. Furthermore, because the direction of bias is not
known in all cases, it is not possible to determine the impact of extrapo-
lation on the soundness of the estimates. Thus, even if no model biases
were present, there would still be uncertainty as to whether a particular
benefit value was an over- or underestimate of the "true" level of
benefits.
Although it is difficult to make objective judgments on the soundness
of the various benefit estimates, consideration of the evaluation criteria,
the model and extrapolation biases, and the impact of uncertainty can all
help to provide a basis for making subjective judgments on the relative
merits of each of the studies.
Because of model design and the manner in which benefits are calcu-
lated, we believe the HTH study should be weighted most heavily in the
compilation of household benefits of reduced soiling. Since the estimates
from this study fall at the low end of the range, and there is reason to
believe that the HTH benefits are conservative estimates, a reasonable
lower-bound estimate of overall benefits for each scenario is the point
estimate of the MTH study.
The process of choosing an upper—bound estimate is more speculative.
The option chosen involves adding the point estimates from the Watson and
Jaksch (6) and Cummings (5) studies. The rationale for this choice is
based on the types of benefits identified in the two studies. In particu-
lar, Cummings' model is designed to reflect both changes in out-of-pocket
expenditures as well as the opportunity cost of labor time spent in
cleaning activities, while the Watson and Jaksch model captures behavioral
7-94
-------
Table 7-32
EXTRAPOLATION BIASES IN THE HOUSEHOLD SECTOR MODELS
Model
Gummings et al.
•^1
Ul
Watson & Jaksch
Description of Bias
1. The damage function is calculated from data
for Philadelphia only.
2. The damage function is based on a TSP range
that is less than that in U.S. counties
included in extrapolation.
3. Air quality improvements are experienced
only in the counties included in air quality
data file.
4. Population exposure calculated from improve-
ments observed only at design value monitor.
1. Demand and supply curves are based only on
data from Philadelphia.
2. Per-honsehold expenditures are assumed
constant across all counties (equal to per-
household expenditures in Philadelphia).
3. The data underlying the unit elastic demand
curve are drawn from a limited range of TSP
values.
Expected Impact on
Benefits
Unknown
Unknown
Underestimate of
benefits
Probable overesti-
mate of benefits
Unknown
Unknown
Unknown
(cont inued)
-------
Table 7-32 (Continued)
Model
Watson & Jaksch
(continued)
Mathtech
vO
Description of Bias
Air quality improvements are experienced
only in the counties included in the air
quality data file.
Population exposure calculated from improve-
ments observed only at design value monitor.
TSP elasticities for all counties are
generated from a model that includes only 24
SMSAs.
Price data are assigned to counties based on
regional averages.
The household model is estimated for a range
of TSP concentrations that is more narrow
than the range implied in the extrapolation.
Air quality improvements are experienced
only in the counties included in the air
quality data file.
Population exposure calculated from improve-
ments observed only at design value monitor.
Expected Impact on
Benefits
Underestimate of
benefits
Probable overesti-
mate of benefits
Unknown
Unknown
Unknown
Underestimate of
benefits
Probable overesti-
mate of benefits
-------
adjustments based on utility and cost considerations. Since the sum of
benefits across these two studies likely contains double-counting, we
believe that the sum should be viewed as an upper-bound estimate. The
extent to which thesum is an upper-bound also depends on the accuracy of
the underlying studies. Unfortunately, since many of the biases that are
present in the two studies affect benefits in an unknown direction, with an
unknown magnitude, it is not possible to determine whether the studies lead
to over- or underestimates of benefits. However, because of the potential
for double-counting, it is our judgment that the sum of benefits from these
two studies provides a reasonable upper-bound estimate of the benefits from
reduced soiling.
The point estimate of benefits is calculated as the geometric mean of
the lower- and upper-bound estimates. If equal weight were given to each
of the studies, then the logical choice for the point estimate would be
taken from the Watson and Jaksch results, since the results of this study
fall between Cummings and MTU. However, we believe more weight should be
given to the HTH results. Thus, we have chosen to use the geometric mean,
which gives less weight to higher values (relative to the arithmetic mean).
Table 7-33 summarizes the range of benefits estimated for the house-
hold sector for each of the scenarios. For each scenario, the upper-bound
estimate is about 20 times larger than the lower-bound estimate. This wide
range in benefits estimates reflects, in large part, the uncertainty that
is currently present in the modeling and estimation of household soiling
models.
Manufacturing Sector
JUthtech Manufacturing Model (8) —
The basic analysis of the MTM study was limited to the analysis of six
3-digit SIC industries comprising about 8.3 percent of the value added in
the manufacturing sector. The analysis was restricted to this subset of
industries because of data limitations. From this group of six industries.
7-97
-------
Table 7-33
SUMMARY OF BENEFITS FROM REDUCED SOILING IN THE HOUSEHOLD SECTOR*
Standard
PM10 70/250
PM10 55/250
PM10 55/150
PM10 55
TSP 75/260
TSP 150
Benefit
Minimum
0.73
1.30
1.62
1.29
2.41
3.15
Point
3.14
5.68
7.16
5.64
10.74
14.21
Maximum
13.85
25.72
32.76
25.53
49.90
66.36
* Discounted present values in 1982 in billions of 1980 dollars.
PM was found to be an important explanatory variable in the cost relation-
ships for two of the industries. Thus, the coverage in the basic analysis
of MTM is small.
Extrapolation of the MTM results to more complete estimates of
national benefits can be done in several ways. First, the geographic
coverage of the two PM-sensitive industries could be extended to other
areas of the country. However, since the original MTM data collection
effort included all the county-level economic data which were available, no
additional county coverage could be obtained.
The alternative possibility is to extrapolate the results for the two
PM-sensitive industries to other closely related manufacturing industries.
For example, instead of conducting the analysis at the 3-digit level, one
might take the view that all 3-digit industries in a given 2-digit group
can be treated similarly. That is, results for SIC 354 could be treated as
7-98
-------
representative of effects for all 3-digit SICs in the 2-digit SIC 35. If
this approach is taken, extrapolated benefits could be obtained for two 2-
digit SICs: SIC 34 and SIC 35.
The extrapolation to other industries raises at least two questions.
First, can the effects identified in a subindustry (e.g., SIC 344) be
viewed as representative of the effects in the broader industry group?
Second, if so, how should the extrapolation be carried out? The first
question cannot be answered definitively without actually conducting a
specific analysis of other subindustries in each group. Clearly, there are
similarities among the various 3-digit industries within a 2-digit group.
The similarities can include the use of common raw materials, similar
processing techniques, and most importantly, the production of related end
products. However, the industries can also be different in important ways,
and it is the latter fact which guided the selection of an extrapolation
procedure.
One possible extrapolation procedure would be to apply the estimated
models from the MTH basic analysis to data for the corresponding 2-digit
industries. Data and conceptual problems prevented this alternative from
being applied. Consequently, a less formal extrapolation procedure was
adopted. The procedure involves answering the following question: If the
benefits of improved air quality at the 2-digit level were the same as at
the 3-digit level in terms of the percentage savings in production cost for
a given change in air quality, how large would benefits be? Note that this
approach does not necessarily require that the underlying production
technologies be the same — only that air quality benefits, on a percentage
basis, be the same.
In order to calculate the extrapolated benefits using the approach
described above, data are collected on the value of shipments for both 2-
digit and 3-digit industries on a county basis. With estimates of benefits
available for the 3-digit industries, a normalizing factor is estimated for
each county which reflects the dollar benefits, per dollar of value
shipped, per unit change in air quality. In order to establish a probable
7-99
-------
range for the factor, the minimum, maximum and average across all counties
(in the analysis) are calculated. Given county—specific air quality data
and value of shipments at the 2-digit level, the three derived normalizing
factors can be used to calculate a range of benefits for the 2-digit
industries. Note that with the use of the more aggregate 2-digit industry
data, data are available for more counties than in the analysis of 3-digit
industries. Hence, extrapolation to the industry level means that more
counties can be considered in the 2-digit analysis.
Benefits estimates for the two 3-digit SICs are given in Tables 7-34
through 7-45. The first'six tables correspond to SIC 344 and represent
benefits for the six scenarios shown in Table 7-10. As in the household
sector, these numbers represent the case where all counties are brought
into attainment throughout the analysis period. Furthermore, because
county average second-high 24-hour average TSP data were used in the MTM
study, a correction factor to reflect facility exposure relative to the
design value monitor readings has been utilized.*
In the tables for the 3-digit SICs, note that only point estimates are
given. The nonlinear (in TSP) system used to estimate the cost functions
for the various industries make it difficult to compute statistically
correct confidence intervals.
Tables 7-40 through 7-45 present the benefit estimates for SIC 354.
In this group of tables, the more lenient standards report no benefits for
several of the EPA regions. This does not necessarily imply that no TSP
benefits will occur in these areas. The zero entries occur because the
necessary economic data were not available for counties in these areas.
Consequently, no benefits could b.e calculated.
Tables 7-46 through 7-57 show the benefit estimates for the two 2-
digit SICs. As before, six tables are shown for each SIC, corresponding to
* See Section 9 for a description of the process used to estimate this
correction factor.
7-100
-------
Table 7-34
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 344
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Point
Federal Administrative Region Minimum Estimate Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
99.1
67.9
403.7
140.2
0.0
12.3
51.3
32.9
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
807.4
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
323.2
0.0
7-101
-------
Table 7-35
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 344
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.I.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Point
Estimate
7.7
14.7
221.5
225.5
756.9
269.2
39.6
47,
83,
.1
.9
37.1
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
1703.1
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
681.7
0.0
7-102
-------
Table 7-36
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC -344
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
134.1
45.7
299.1
294.9
789.9
309.7
66.9
76.6
105.5
97.2
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
2219.5
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
888.4
0.0
7-103
-------
Table 7-37
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 344
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Point
Estimate
7.7
14.7
221.5
225.5
756.9
266.4
39.6
47.1
83.9
37.1
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
1700.3
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
680.6
0.0
7-104
-------
Table 7-38
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 344
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AAM/260 24-hr.
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
27.3
.1
.4
.5
36,
427.
413,
1482.2
521.0
129.0
73.0
164.6
67.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
3341.6
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
0.0
934.5
0.0
7-105
-------
Table 7-39
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 344
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
225.2
102.1
623.4
599.5
1527:6
502.7
271.0
125.5
116.7
149.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
4243.2
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
0.0
1186.6
0.0
7-106
-------
Table 7-40
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 354
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Point
Estimate
0.0
2.4
34.8
1.2
680.4
10.2
0.0
0.0
1.0
0.0
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
730.0
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
292.2
0.0
7-107
-------
Table 7-41
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 354
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
Point
Estimate
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
1321.3
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
528.9
0.0
7-108
-------
Table 7-42
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 354
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Point
Estimate
94.3
61.2
70.6
4.8
1219.0
23.4
5.4
0.0
1.9
0.0
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
1480.5
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
592.6
0.0
7-109
-------
Table 7-43
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 354
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
Point
Estimate
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
1321.3
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
528.9
0.0
7-110
-------
Table 7-44
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 354
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AAM/260 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Point
Estimate
20^8
47.4
122.1
6.9
2164.3
41.0
20.5
0.0
4.3
0.0
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
2427.2
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
0.0
678.8
0.0
7-lH
-------
Table 7-45
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 354
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Point
Estimate
192.2
.7
.7
107,
157,
10.0
2229.8
40.2
65.1
0.0
5.4
0.0
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0-
Total U.S.
0.0
2808.1
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
0.0
785.3
0.0
7-112
-------
Table 7-46
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 34
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
136.9
97.5
1082.0
166.8
13.4
20.6
332.2
65.2
Point
Estimate
0.0
0.0
251.4
179.0
1986.0
306.0
24.6
37.9
609.9
119.7
Maximum
0.0
0.0
368.8
262.6
2912.9
448.8
36.0
55.6
894.8
175.6
Total U.S.
1914.6
3514.5
5155.1
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
766.4
1406.8
2063.5
7-113
-------
Table 7-47
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 34
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
47,
50.
263,
257.8
1910.7
309.6
78.
- 67,
750.
,5
.7
.4
.0
.4
,1
Point
Estimate
88,
94,
491.9
.5
,1
,0
88.1
481.
3568.
578.
145.6
125.9
1401.0
164.5
Maximum
139.4
148.8
772.6
756.4
5604,
907,
228.
197.7
2201.1
258.5
.5
.7
.7
Total U.S.
3823.3
7139.9 11215.4
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
1530.4
2858.0
4489.3
7-114
-------
Table 7-48
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 34
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
223.0
120.4
385.6
359.1
2519.6
405.7
152.3
142.2
1014.3
185.7
Point
Estimate
391.2
211.2
676.4
630.0
4420.3
711.5
267.2
249.4
1780.2
326.0
Maximum
713.1
384.9
1232.8
1148.2
8056.5
1296.9
487.0
454.6
3244.6
594.1
Total U.S.
5507.9
9663.3 17612.7
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
2204.7
3868.0
7050.0
7-115
-------
Table 7-49
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 34
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
47.
50.
263.
1910,
308.
257.8
78.0
67.
750.
88.1
Point
Estimate
88.3
94.3
489.8
479.5
3553.1
572.7
145.0
125.3
1395.1
163.8
Maximum
139.4
148.8
772.6
756.4
5604.
903
228.
197,
2201.
,5
.3
,7
.7
.1
258.5
Total U.S.
3821.7
7107.1 11211.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
1529.8
2844.8
4487.5
7-116
-------
Table 7-50
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 34
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AAM/260 24-hr.
Federal Administrative Region Minimum
Point
Estimate
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
144.8
121.1
566.
466,
3748.6
585.2
246.4
152.6
1542.9
149.6
,5
.3
267,
223,
1046.2
861.2
6923,
1080.
455.0
281.9
2850.8
276.4
.5
.7
,0
.5
Maximum
385.0
322.0
1505.3
1239.3
9961.4
1554.5
654,
405.
4103,
.7
.7
.2
397.9
Total U.S.
7724.0
14266.3 20528.9
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
2160.0
3989.6
5740.9
7-117
-------
Table 7^51
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 34
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
551.1
267.3
858.4
660.4
4775.9
634.8
457.8
246.6
1987.5
296.6
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
TV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
1008.2
489.1
1570.5
1208.8
8738.0
1161.0
837.7
451.2
3638.9
543.1
Maximum
1663.5
807.0
2591.1
1994.5
14417.1
1915.6
1382.2
744.5
6003 . 8
896.0
Total U.S.
10736.3
19646.6 32415.4
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
3002.4
5494.2
9065.0
7-118
-------
Table 7-52
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 35
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
10.3
119.5
61.4
1148..7
261.0
0.0
26.4
460.8
71.7
0.0
0.0
28.4
327.6
184.8
3148.3
715.2
0.0
72.5
1263.4
45.1
521.3
294.0
5008.4
1137.8
0.0
115.3
2009.9
196.7
313.0
Total U.S.
2165.8
5936.9
9444.8
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
866.9
2376.4
3780.5
7-119
-------
Table 7-53
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 35
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
26.5
152.2
262.3
170.4
2381.9
502.9
10.6
146.2
1054.5
94.2
Point
Estimate
73.9
423.9
730.2
474.8
6631.7
1399.8
29.6
407.1
2937.4
262.6
Maximum
116.5
668.7
1151.9
749.0
10461.5
2208.2
46.7
642.2
4633.6
414.2
Total U.S.
4801.7
13370.9 21092.5
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
1922.0
5352.1
8442.8
7-120
-------
Table 7-54
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 35
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM/150 24-hr.
Federal Administrative Region Minimum
202.6
246.6
341.2
213.6
2704.5
594.6
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
562.6
685.1
947.9
593.8
7514.9
1651.9
121.5
595.2
3372.8
535.0
Maximum
875.0
1065.5
1474.2
923.5
11686.4
2568.9
188.9
925.6
5244.9
831.9
Total U.S.
5966.4
16580.8 25784.8
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
2388.2
6636.9 10321.1
7-121
-------
Table 7-55
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 35
Benefits Occurring Between 1989 and 1995
Scenario: Type B PM10 - 55 AAM
Federal Administrative Region Minimum
Point
Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
26.5
152.2
262.3
170.4
2322.5
480.9
10.6
146.2
1054.5
94.2
73.9
423.9
730.2
474.8
6466.2
1338.4
29.6
407.1
2937.4
262.6
116.5
668.7
1151.9
749.0
10200.4
2111.4
46.7
642.2
4633.6
414.2
Total U.S.
4720.3
13144.1 20734.6
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
1889.4
5261.3
8299.6
7-122
-------
Table 7-56
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 35
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 75 AAM/260 24-hr.
Point
Federal Administrative Region Minimum Estimate
Maximum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
278.9
846.3
1604.4
879.3
13544.3
2363.6
232.6
638.6
5667.0
424.8
472.0
1432.2
2715.2
1488.1
22921. &
4000.0
393.6
1080.8
9590.4
718.9
Total U.S.
11299.5
26479.7 44813.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
3159.9
7405.1 12531.9
7-123
-------
Table 7-57
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 35
Benefits Occurring Between 1987 and 1995
Scenario: Type B TSP - 150 24-hr.
Federal Administrative Region Minimum
797.9
627.0
1022.9
497.6
7297.1
1143.1
332.6
463.1
3049.6
384.7
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
rv
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
1811.8
1424.7
2323.9
1132.4
16577.0
2596.2
756.0
1051.8
6936.7
875.4
Maximum
2949.0
2318.8
3782.5
1843.0
26981.5
4225.8
1230.4
1711.9
11289.7
1424.7
Total U.S.
15615.5
35485.9 57757.4
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1987 and 1995
Total U.S.
4366.9
9923.6 16151.8
7-124
-------
the six scenarios described in Table 7-10. These tables are also based on
complete attainment of the various standards, and the correction factor for
facility exposure has been incorporated in the estimates. In this group of
tables, minimum and maximum estimates are presented in addition to the
point estimates. As described above, these bounding values are calculated
from the minimum and maximum values of the normalizing factor.
Finally, for comparative purposes. Tables 7-58 through 7-61 show the
benefits that would accrue for the two 2-digit SICs and the two 3-digit
SICs when not all counties are assumed to be in attainment throughout the
analysis period. The estimates in these tables are given only for the
70/250 PM10 standard.
At the conclusion of the section describing the benefits models, a
table of model biases was presented. Of the six biases listed for the MTM
study, four were identified as having an "unknown" impact on the benefit
estimates. As a consequence, it is not possible to determine conclusively
whether the estimates shown above represent over- or underestimates of some
"true" level of benefits for the SICs examined in MTM. If production costs
of other industries are also affected by changes in TSP, then the estimates
of this section represent conservative estimates of benefits for the total
manufacturing sector.
In addition to the model biases, there are several biases that arise
in the course of estimating national benefits. One of these biases, the
concern with a limited number of SICs, was mentioned above. Other biases
.of this type include:
• A limited number of counties are included in the extrapola-
tion since confidentiality restrictions do not allow
identification of needed economic data in all cases.
• It is assumed that air quality improvements are experienced
only in the counties included in the air quality data file.
• The procedure for benefits calculation assumes a perfectly
inelastic demand for plant output.
7-125
-------
Table 7-58
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 344
Benefits Occurring Between 1989 and 1995
Scenario: Type A PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
Point
Estimate
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
.2
.2
0.0
0.0
95
65.
255.9
79.6
0.0
12.3
31.3
5.1
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
544.6
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
218.0
0.0
7-126
-------
ESTIMATED BENEFITS FOR:
Table 7-59
MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 354
Benefits Occurring Between 1989 and 1995
Scenario: Type A PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
Point
Estimate
Maximum
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.4
33.5
1.2
459.1
5.8
0.0
0.0
0.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Total U.S.
0.0
502.7
0.0
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
0.0
201.2
0.0
7-127
-------
Table 7-60
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 34
Benefits Occurring Between 1989 and 1995
Scenario: Type A PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
TV
V
VI
VII
VIII
IX
X
New England
N.T.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
0.0
132.9
93.3
784.7
104.4
12.1
20.5
160.9
15.7
Point
Estimate
0.0
0.0
231.8
162.9
1368.8
182.1
21.2
35.7
280.8
27.5
Maximum
0.0
0.0
357.9
251.4
2112.6
281.0
32,
55,
433,
.7
.1
.5
42.4
Total U.S.
1324.6
2310.7
3566.6
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
530.2
924.9
1427.6
7-128
-------
Table 7-61
ESTIMATED BENEFITS FOR: MATHTECH MANUFACTURING EXPENDITURE STUDY
SIC 35
Benefits Occurring Between 1989 and 1995
Scenario: Type A PM10 - 70 AAM/250 24-hr.
Federal Administrative Region Minimum
Point
Estimate
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
REGION
I
II
III
IV
V
VI
VII
VIII
IX
X
New England
N.Y.-N.J.
Middle Atlantic
South Atlantic
E.N. Central
South Central
Midwest
Mountain
South Pacific
North Pacific
0.0
9.6
107.3
56.9
756.2
143.3
0.0
24.3
.239.5
18.6
0.0
27.0
301,
160.
2124.
402.
.6
.1
.7
.5
0.0
68.3
673.0
52.2
Maximum
0.0
43.7
488.7
259.4
3442.6
652.1
0.0
110.7
1090.5
84.6
Total U.S.
1355.7
3809.3
6172.2
Discounted Present Value in Millions of 1980 Dollars in 1982
Using a 10 Percent Rate of Discount.
Annualized Benefits
Between 1989 and 1995
Total U.S.
542.7
1524.8
2470.6
7-129
-------
Each of these biases is expected to contribute to an underestimate of
benefits. There may also be a bias introduced by the assumption that the
dollar benefit per microgram reduction is the same for all industries in a
2-digit SIC grouping. Although the minimum and maximum estimates were
developed to reflect the uncertainty surrounding the assumption, it is not
possible to determine whether the assumption is ezpeccted to lead to a
positive or negative effect on benefits.
Synthesis of Manufacturing Sector Benefits —
A total benefit estimate for industries in the two 2-digit industry
groups was developed as follows:
A lower-bound estimate was calculated as the point estimate
for SIC 354.
An upper-bound estimate was calculated as the sum of the
point estimates for the two 2-digit industries.
A point estimate was calculated as the geometric mean of
the upper- and lower-bound estimates.
The choice of the lower-bound estimate was influenced by the fact that
TSP was not a significant explanatory factor at the 10 percent level for
all specifications in SIC 344. Thus, a very conservative estimate of a
lower-bound value would include benefits derived only from SIC 354. Given
that these estimates are developed directly from the data and equations in
the basic MTM analysis, they are the most defensible of the various benefit
numbers presented for the manufacturing sector.
Several options were possible for selecting an upper-bound estimate.
The choice of the sum of the two 2-digit SIC point estimates reflects our
judgment that the various biases in the MTM model and in the extrapolation
procedures will likely lead to an overestimate of benefits at the 2-digit
level. As was the case with the household sector, the choice of an upper-
bound estimate is more speculative relative to the choice of the lower-
bound estimate.
7-130
-------
Finally, the point estimate is determined as the geometric mean of the
upper- and lower-bound estimates. The geometric mean appears to be a more
relevant statistical index than the arithmetic mean since it weights lower
values relatively greater. This would conform to our belief that more
confidence can be ascribed to the estimates presented as lower-bound values
in the various scenarios relative to those identified as upper-bound
values.
Table 7-62 summarizes the estimated benefits of reduced soiling for
the manufacturing sector. As was pointed out in the review of the MTM
study, these estimates should be interpreted with caution. Although the
analytical model and empirical technique are carefully applied in the MTM
analysis, the results were found to be sensitive to influential observa-
tions in the data.
Table 7-62
SUMMARY OF BENEFITS FROM REDUCED SOILING IN THE MANUFACTURING SECTOR*
Standard
PM10 70/250
PM10 55/250
PM10 55/150
PM10 55
TSP 75/260
TSP 150
Benefit
Minimum
0.73
1.32
1.48
1.32
2.43
2.81
Point
1.30
2.41
2.86
2.41
4.80
5.81
Maximum
9.45
20.51
26.24
20.25
40.75
55.13
* Discounted present values in 1982 in billions of 1980 dollars,
7-131
-------
REFERENCES
1. Beloin N. and F. Haynie. Soiling of Building Materials. Journal of
the Air Pollution Control Association, 24:339, 1975.
2. U.S. Environmental Protection Agency. Air Quality Criteria for Parti-
culate Matter and Sulfur Oxides. Volume IV, Review Draft December
1981.
3. U.S. Environmental Protection Agency. Review of the National Ambient
Air Quality Standards for Particulate Matter. Draft Staff Paper,
1982.
4. U.S. Environmental Protection Agency. Review of the Relationships of
IP10, IP15 and TSP. Memo to H. Thomas of EPA, July 1981.
5. Cummings, R., H. Burness and R. Norton. Methods Development for
Environmental Control Benefits Assessment, Volume V: Measuring House-
hold Soiling Damages from Suspended Air Particulates, A Methodological
Inquiry. Draft Report, January 1981.
6. Watson, W. and J. Jaksch. Air Pollution: Household Soiling and
Consumer Welfare Losses. Forthcoming in Journal of Environmental
Economics and Management, 1981.
7. Mathtech, Inc. Benefits Analysis of Alternative Secondary National
Ambient Air Quality Standards for Sulfur Dioxide and Total Suspended
Particulates, Volume II. Final report to U.S. Environmental Protec-
tion Agency, May 1982.
8. Mathtech, Inc. Benefits Analysis of Alternative Secondary National
Ambient Air Quality Standards for Sulfur Dioxide and Total Suspended
Particulates, Volume III. Final report to U.S. Environmental Protec-
tion Agency, May 1982.
9. Geomet, Inc. Sulfur Dioxide and Sulfates Materials Damage Study.
Draft Final Report prepared for U.S. Environmental Protection Agency,
Research Triangle Park, NC, February 1980.
10. Waddell, Thomas E. The Economic Damages of Air Pollution. Environ-
mental Protection Agency 600/5-74-012, May 1974.
11. Freeman, A. M. The Benefits of Environmental Improvement: Theory and
Practice. John Hopkins University Press, Baltimore, Maryland, 1979.
12. Lodge, J. P., Jr., A. Waggoner, D. Klodt, and C. Crain. Nonhealth
Effects of Airborne Particulate Matter. Atmospheric Environment, Vol.
15, pp. 431-482.
7-132
-------
13. Michelson, I. and B. Tourin. Report on Study of Validity of Extension
of Economic Effects of Air Pollution Damage from Upper Ohio River
Valley to Washington, D.C. Area Environmental Health and Safety
Research Association, August 1967.
14. Ridker, R. Economic Costs of Air Pollution. New York: Frederick A.
Praeger Press, 1967.
15. Narayan R. and B. Lancaster. Household Maintenance Costs and Particu-
late Air Pollution. Clean Air, 7:10-13, 1973.
16. Booz, Allen and Hamilton, Inc. Study to Determine Residential Soiling
Costs of Particulate Air Pollution. APTD-0715, National Air Pollution
Control Administration, October 1970.
17. Liu, B. and E. Yu. Damage Functions for Air Pollutants. Report
prepared for U.S. Environmental Protection Agency, February 1976.
18. Brookshire, D., R. d'Arge, W. Schulze, and M. Thayer. Methods
Development for Assessing Tradeoffs in Environmental Management,
Volume II. EPA-600/6-79-0016, 1979.
19. SRI, Inc. An Estimate of the Non-Health Benefits of Meeting the
Secondary National Ambient Air Quality Standards. Prepared for the
National Commission on Air Quality, January 1981.
20, Courant P. and R. Porter. Averting Expenditure and the Cost of Pollu-
tion. Journal of Environmental Economics and Management 8(4),
December 1981.
21. Rowe, R. and L. Chestnut. Issues in Visibility Benefit-Cost Analysis,
Draft Report for U.S. Environmental Protection Agency, August 1981.
22. Esmen N. A Direct Measurement Method for Dustfall. Journal of the
Air Pollution Control Association, 23:34-36, 1973.
23. U.S. Bureau of Labor Statistics. Consumer Expenditure Survey:
Integrated Diary and Interview Survey Data, 1972-73. Bulletin 1992,
U.S. Government Printing Office, Washington, DC, 1978.
24. U.S. Bureau of Labor Statistics. Average Retail Prices of Selected
Commodities and Services. U.S. Government Printing Office, 1973.
25. Stone, R. Linear Expenditure Systems and Demand Analysis: An Appli-
cation to the Pattern of British Demand. The Economic Journal 64:511-
527, 1954.
26. Diamond P. and D. McFadden. Some Uses of the .Expenditure Function in
Public Finance. Journal of Public Economics, 3:3-21, 1974.
7-133
-------
27. Mathtech, Inc. Benefits Analysis of Alternative Secondary National
Ambient Air Quality Standards for Sulfur Dioxide and Total Suspended
Particulates - Volume VI. Final Report to U.S. Environmental Protec-
tion Agency, Research Triangle Park, NC, May 1982.
28. U.S. Department of Commerce, Bureau of the Census. Census of
Manufactures, Vol. II. 1972.
29. U.S. Department of Commerce, Bureau of the Census. City and County
Data Book. 1977.
30. Bureau of the Census. Current Population Reports Series P-25, Projec-
tions of the Population of the United States, 1977 to 2050. No. 704,
July 1977.
31. U.S. Department of Commerce News. Projections of Personal Income to
the Year 2000. December 9, 1980.
7-134
-------
SECTION 8
BENEFITS OF NATIONAL VISIBILITY STANDARDS
-------
SECTION 8
BENEFITS OF NATIONAL VISIBILITY STANDARDS
INTRODUCTION
Existing assessments of the benefits of air pollution control have to
a large extent focused on the health, soiling, materials damage, and vege-
tation damage effects of air pollution. However, an increasing body of
economic research suggests that the aesthetic effects of air pollution are
also important. Visibility, defined as the ability to see distant vistas
nnobscured, is affected by naturally occurring dust and humidity as well as
by air pollution. Based on psychological.studies of perception and studies
of the value of clean air, visibility appears to be the dominant aesthetic
impact of air pollution. Thus, an appropriate research task is the estima-
tion of the value of preserving or improving atmospheric visibility as a
result of air pollution control programs. An assessment of the national
benefits of visibility protection has not been accomplished to this point.
This section attempts to make a very preliminary estimate of such benefits,
acknowledging that, at this point, critical research still needs to be
completed before such an estimate can be considered to be rigorous. How-
ever, the estimates provided here do suggest the order of magnitude of
benefits to be derived from a program of national visibility protection.
Benefits are calculated for four alternative visibility standards
defined by a minimum annual allowable average visual range of 13 miles, 20
miles, 30 miles, and a 20 percent improvement over existing levels. In
making these benefit calculations we assume, for example, for the 13-mile
standard that all areas in the country which now have an average annual
visibility below 13 miles are brought up to an average annual visual range
°f 13 miles. Obviously, however, reductions in emissions necessary to
8-1
-------
bring the worst areas up to a visual range of 13 miles will, as a side
effect, improve visibility in nearby areas already meeting the standard.
We do not calculate the benefits of these secondary improvements in visi-
bility, although ideally they should be included.
The focus on alternative visual range standards rather than concentra-
tions of PM was required because no rigorous transformation has been
defined between visual range and levels of TSP or PM10. Scientific studies
of the optical properties of PM indicate that fine particles (FP) are the
most dominant size cut influencing light scatter. Consequently, an FP
standard may be most appropriate for identifying the benefits associated
with changes in visual range. Since the scenarios examined in this report
relate to PM10 and TSP only, the benefits reported in this section cannot
be added in a consistent fashion with the benefit numbers reported in the
other sections. This section is most properly viewed then only as a
synthesis of available information of the benefits of alternative national
visibility standards.
For calculating national visibility benefits we utilize three additive
categories of value: 1) residential, 2) recreation, and 3) existence. In
a separate subsection, we discuss in qualitative terms some of the work
that has looked at visibility impairment and effects on air and highway
safety. Residential benefits are defined as those which are derived from
visibility in and around consumers' homes and local communities. Thus,
this category includes, for example, benefits of unobscured views from and
around homes, unpolluted clear blue skies, and the benefits to recreation
undertaken near peoples' homes. These are the benefits presumably captured
by the urban surveys of visibility benefits.* Note that these surveys have
typically asked households for their willingness-to-pay for clean air and
then asked respondents to split their bid into component parts, including
as one component visibility. Also, these studies have compared their
results with property value studies and, in general, have found survey
results to be consistent with the effects of air pollution on property
* See, for example, Brookshire, d'Arge, Schulze and Thayer (1).
8-2
-------
values. However, such studies have been completed only for cities on the
West Coast (Los Angeles and San Francisco). Although visual ranges are of
the same order of magnitude as for eastern cities (less than 20 miles),
extrapolating these results to the East is not a desirable procedure.
Fortunately, results for East Coast cities will be available within one
year from a study underway at the University of Chicago under the overall
management of George Tolley. Preliminary pre-test results from Chicago do,
however, show roughly the same value of visibility per mile as the West
Coast studies done by the University of Wyoming group.*
The second category of benefits, recreation value, is defined as the
benefits of visibility improvement to recreation undertaken away from a
household's home community. For example, a visit to the Grand Canyon
provides enjoyment from being able to see far down the canyon from an
overlook. To avoid overlap with residential benefits, we assume that half
of the average of 60 days of recreation undertaken by families per year
takes place away from the home community. Again, the only available
studies of the value of visibility to recreation have been undertaken in
the western United States. The extrapolation to eastern recreation is
quite weak because visual ranges in the western recreation areas studied
have been greater than 50 miles, while in eastern recreation areas they are
generally less than 50 miles. But, such an extrapolation is necessary to
obtain an order of magnitude estimate of the benefits of visibility
standards. Again, the need for new research is clear. The University of
Chicago study mentioned above should provide some preliminary evidence in
this area for the East Coast.
The final category of benefits, which is examined quantitatively, is
existence value. Existence value is defined as the benefit derived from
just knowing that the environment is preserved. Note that this value is
not associated with use of the environment and probably applies to "natural
wonders". Thus, a household in New York may be willing to pay to know that
visibility at the Grand Canyon is preserved even if they never intend to
* Op. cit. Brookshire et al. (1).
8-3
-------
visit. Only one study of the existence value of visibility protection has
been completed, and that study focused on the Grand Canyon.* Thus, to
provide even a rough order of magnitude of the benefits of visibility
standards, a weak extrapolation must again be made. We have chosen to
include six particularly well-known national parks for which we extrapolate
existence values for visibility based on the Grand Canyon results. The
results from the Grand Canyon study indicated that the existence value
component of visibility benefits is quite large. Consequently, this
category may be especially important in forming decisions on where future
research may be most valuable.
Prior to reporting aggregate national benefit estimates, we describe
in qualitative terms research on safety aspects of reduced visual range.
In the discussion, it is suggested that cost of delays is probably a more
appropriate measure of better visibility benefits than lives saved. Unfor-
tunately, no usable study has yet to be completed which permits a quantita-
tive evaluation of these safety aspects.
The final part of this section reports on the overall benefits calcu-
lations, giving annual benefits and discounted present values for different
attainment paths for alternative visibility standards. We have also calcu-
lated low, medium, and high estimates for benefits based on differing
assumptions. Table 8-1 presents 1980 annual benefits if the alternative
visibility standards could have been attained in that year using our medium
estimates, except for existence values, where we only include six parks
(our low estimate). Of these estimates, as we have indicated above, the
residential benefits are the most defensible. We would argue that an
appropriate interpretation of Table 8-1 is as follows: "the benefits of a
20-mile visibility standard may be of the order of magnitude of $7 billion
(residential benefits) and could easily be as high as $15 billion (total
benefits).
* See Schulze, Brookshire et al. (2).
8-4
-------
Table 8-1
1980 ANNUAL NATIONAL BENEFITS OF ALTERNATIVE VISIBILITY STANDARDS
(Millions of 1980 Dollars)
Standard
13 miles
20 miles
30 miles
20% Improvement
over current
visibility
Residential
Benefits
1,581
6,986
16,887
7,608
Recreation
Benefits
247
1,595
4,535
12,296
Existence
Value
Benefits
2,172
7,194
14,430
10,172
Total
4,000
15,775
35,852
30,076
ESTIMATING TBR BENEFITS OF URBAN VISIBILITY IMPROVEMENTS
Introduction
During the past 15 years, more than a score of econometric and survey
studies have been conducted to estimate the economic value of air quality.
These studies have usually adopted one of two approaches. First are
studies which use published data on housing values to infer a "price" for
air quality; and second are those studies which ask individuals to estimate
directly their willingness- to-pay for air quality improvements. With the
exception of one study, the resulting valuations are for all aspects of air
pollution reduction, inclusive of perceived human health effects, vegeta-
tion impacts, corrosion, impacts on animals, aesthetic considerations such
as visibility, and other humanly conceived effects.
It is the purpose of this subsection to review briefly the results
obtained in several of the property value and willingness-to-pay approaches
for assessing the benefits of improved visibility in urban-suburban
8-5
-------
locations. In Section 5, property value studies were reviewed and benefits
evaluated for the alternative air quality scenarios being examined in this
report. The discussion of property value studies in this section focuses
on the attainment of visual range standards and is used primarily as a
cross-check on the plausibility of the residential visibility benefit
estimates derived from willingness-to-pay studies. For this reason, the
discussion of alternative property value studies is limited in this
section. There is, however, extended discussion of the various biases that
may be part of the direct willingness-to-pay studies.
In the property value studies, the common procedure is to link
differences in property values to various measures of air pollution concen-
tration. Since measures of various pollutants are often recorded in
different averaging times or units, the various property value studies
cannot be easily compared. In addition, further complications arise in
that benefits are to be estimated for visual range standards. In order to
assess the relative monetary magnitude of the property value estimates, a
set of "standardization" equations are developed here to relate various air
pollution concentrations to visual range. These "standardization"
equations are likely to contain substantial errors for any one site or
pollutant condition and therefore should be viewed with caution and quali-
fications when comparing the various property value studies.
In the review of study results, willingness-to-pay studies are also
compared as to their "implied" willingness-to-pay for visibility and other
aspects of air pollution. These studies are more directly comparable since
in only one case was there a need to use the "standardization" equation
(3).
Given the estimated marginal willingness-to-pay for the two
approaches, we next examine briefly the relationship of the magnitudes of
benefits from the two types of studies. We would expect that the property
value estimates should exceed those derived from the willingness-to-pay
studies.
8-6
-------
Following the cross-check of approaches, prediction equations are
developed for estimating the relationship between visibility and
willingness-to-pay for clean air. These equations are based exclusively on
results obtained in a willingness-to-pay study of Los Angeles air pollution
(1). However, the prediction equations constructed from the Los Angeles
study are shown to be close to estimates derived from other studies of San
Francisco, Chicago, Boston, and Denver.
Property Value Studies
Using differences in property values between polluted and less
polluted locations has become a rather common approach to assess the
implied damages of air pollution. In addition, a large literature has
evolved on how the results of these studies should be interpreted and the
assumptions necessary to make them valid in estimating the marginal
benefits of improved air quality across an urban area (4).* However, no
one has made a comparison of the magnitude of estimates derived from the
various studies for alternative visibility standards. The major problem in
doing so is the difference in air pollution data inclusive of type of air
pollution, measurement methods applied, averaging time, and type of device
used to measure concentration. In this study, a set of "standardization"
equations is developed with reference only to urban visibility. The calcu-
lations are included in Appendix 8A. By assuming a precise relationship
between equivalents of sulfation and particulates, one can be converted to
the other in terms of impact on visibility [see Trijonis (5)]. Next, using
data for Los Angeles, a relationship is obtained between nitrogen oxides
and particulates holding visibility constant such that an equivalent effect
on visibility can be estimated either from concentrations of particulates
or nitrogen oxides. Finally, using isopleth maps for Los Angeles, a rela-
tionship is obtained between the change in nitrogen oxide concentrations
and changes in visibility. Let S, P, and NO- represent concentrations of
* See Section 5 for a detailed description of the theory underlying the
property value studies.
8-7
-------
sulfation, particulate matter, and nitrogen dioxide, respectively, and let
V represent visual range. Then, by calculation:
S = oP
AN02 = PAP
AV =
where a is estimated from several sources, with P and y estimated directly
from Los Angeles data sources (see Appendix 8A). Depending on particle
size and the presence of other pollutants, the $ estimated from Los Angeles
data may over- or underestimate the relationship for other urban areas of
the U.S. It is likely that the estimate for y is larger for Los Angeles
than other areas because of the above average maximum distance of potential
visibility. Thus, the same concentration of N02 in Los Angeles is likely
to have a greater than average impact on visibility. Given the rather
tenuous nature of these calculations, most of the property value studies
can be compared in visibility units of dollars per mile. For example, a
property value study reporting a "hedonic price" of the change in housing
value per unit change in sulf ation, AHV/AS, can be converted to AHV/AV by
simply noting that:
AV = y AS
or
AHV m /AHV\ . /o_
AV Us / UY
Table 8-2 records the implied or direct willingness-to-pay per mile of
visibility for selected property value studies. The last two studies
included in Table 8-2 are especially important, since willingness-to-pay
(contingent valuation) studies were conducted simultaneously. The
remaining nine property value studies represent a sample of the many
8-8
-------
Table 8-2
ANNUALIZED PROPERTY VALUE CHANGES PER MILE OF VISIBILITY BY STUDY AND AREA*
oo
Study
Ridker-
Henning (6)
Anderson-
Crocker (7)
Crocker
(8)
Harrison-
Rubinfeld (9)
Smith
(10)
Nelson
(11).
Zerbe
(12)
Peckham
(13)
Brookshire,
et al. (1)
SRI
(14)
Area,
Base Year
St. Louis,
1960
St. Louis,
1960
Chicago,
1964-1967
Boston,
1970
Wash., D.C.
1971
Wash., D.C.
1970
Toronto,
1961
Hamilton,
1961
Philadelphia,
1960
Los Angeles,
1979
San Francisco,
1981
in Property Value
US/Apoll.)
186.50-245.00/0.25 rag/100 cm2
(sulfation)
300.00-700.00/10 ug/m3+0.1 mg/100 cm2
(part. + sulfation)
350.00-600.00/10 ug/m3 + 1 ppb SO,
(part. + S02)
800.00/1 ppm NOx
430.00-500.00/10 ug/m3
(part.)
576.00-693.00/10 ug/m3 part.
141.00-152.00/0.01 ppm oxident
200.00-450.00/0.25 mg/100 cm2
(sulfation)
580.00-882.00/0.25 mg/100 cm2
(sulfation)
298.00/0.25 mg/100 cm2
(sulfation)
4468.00-6134.00/30% ANO2
14.73 -» 95.42/30% APSI Index
0.68 -» 31.55/30% AOzone
.A in Miles
of Visibility
11.49 mi
9.19 mi
9.19 mi
4.74 mi
(V = 14.22 mi)
4.60 mi
4.66 mi
11.49 mi
11.49 mi
11.49 mi
Fair -> Good 14 mi
Poor -» Fair 10 mi
Fair -» Good
2.4 mi
$ Per Mile
Per Yr. (1981)
5.66
16.38
14.57
38.84
22.30
38.56
8.43
18.97
7.81
39.27
75.47
6.14 -> 39.76
0.28 -» 13.15
* References.
-------
available studies of this type. Note that all nine appeared in the discus-
sion in Section 5.
In each of the property value studies, property value differentials
are related to ambient concentrations of various air pollutants. Using the
procedure documented in Appendix 8A, these ambient concentrations are
converted to NO- equivalents and then to changes in visibility utilizing
observed relationships for the Los Angeles Air Basin. While this conver-
sion procedure is not likely to be highly accurate, it does allow a
comparison between studies on a per mile or average basis. Given the
qualifications underlying the conversion procedure, the range of implied
willingness-to-pay over the various studies is quite small. The mean for
the sample of property value studies is $23.04, with a standard deviation
of $19.94, and a range of $.28 to $40.00 per household/mile. What is
interesting about the property value studies is that with one exception
they all cluster in the $5.00 to $40.00 per household/mile range on an
annual basis. This suggests that a rough order of magnitude inference can
be made; namely, it is unlikely that households value air pollution, in all
of its "perceived" dimensions, at more than $100.00 per mile or less than
$1.00 per mile.
Direct Willingness—to—Pay Studies
Table 8-3 records the estimated annual willingness-to—pay per house-
hold for miles of visibility when families were asked directly their
willingness-to-pay. Since these studies most often use pictures to portray
visual range as part of the survey process, no "standardization" equations
are required. These studies encompass a fairly broad typing of urban areas
ranging from Los Angeles and Chicago down to a small energy impacted
community in New Mexico. The estimates ranged from $1.54 per house-
hold/mile to $40.25. However, the Bresnock study (3), which provides the
lower bound estimate, utilized carbon monoxide as the relevant measure of
pollution. There appears to be some question as to whether carbon monoxide
is a good proxy for visual range so that the estimated willingness-to-pay
may not adequately reflect visibility changes. Also, the range in changes
8-10
-------
Table 8-3
ANNUAL WILLINGNESS TO PAY PER MILE OF VISIBILITY BY STUDY AND AREA
Study
Bresnock
(3)
Brookshire,
et al. (1)
SRI
(14)
EPRI/Blank,
et a^L. (15)
Tolley-
Randall (16)
Area,
Base Year
Denver ,
1977
Los Angeles,
1979
San Francisco,
1981
Farm ing ton,
1978
Chicago,
1981
WTP Per Year
5.70 for 18 pptra CO (low)
186.24 for 239 pptm CO (high)
312.00
70.56
i
ES- 91.08
CS-856.97
$200*
A in Miles
of Visibility
5.12 (low)
67.96 (high)
Fair -* Good 14 mi
Poor -* Fair 10 mi
2.4 mi
50 mi
9 mi
(V = 9)
$ Per Mile
Per Year (1981)
1.54 (low)
3.79 (high)
Fair -» Good 28.64
Poor -» Fair 40.25
29.40
ES- 2.35
CS-22.11
22.22
00
This is a preliminary estimate based on a pre-test and given in a telephone conversation with one of the
principle investigators.
-------
in visibility are perhaps too non-standardized across the various studies
to allow direct comparison between studies. However, the range of
estimates is within that postulated earlier as reasonable from the property
value studies. The mean is $18.57 with a standard deviation of $14.57.
According to theory, the property value estimate should exceed the
willingness-to-pay estimate because the property value differences are
represented along a rent gradient, while individual willingness-to-pay is
represented by a movement along an indifference curve.* Thus, individuals
are willing to pay less than the current market price because they have
already taken into consideration their individual "perceived" effects of
air pollution. Comparing the means of the property value studies of
$23.04, and $18.57 for the willingness-to-pay studies, at least the results
conform to expectations from theory. However, this is a very weak test
since we are comparing means of estimates derived from different times and
locations.
Biases in. Direct Willinpiess~to~PHT Stiulies
Economists have argued that valuing public goods such as visibility
through a direct demand revealing process such as willingness-to-pay would
yield biased results. The principal theoretical support for this conten-
tion is the possibility of strategic bias. However, as survey techniques
to elicit contingent behavior or bids have come into use — in part because
development of energy resources in formerly pristine environments allows no
other techniques to be used — other types of bias have come to be regarded
as just as important. These include information bias, instrument bias,
hypothetical bias and traditional problems of sampling, interviewer, and
nonrespondent bias.
Beginning with Samuelson's seminal work on public goods, it has been
supposed that direct revelation of consumer preferences for such goods —
and, of course, environmental quality is a public good — would be
* See D. S. Brookshire, M. A. Thayer, W. D. Schulze, and R. C. d'Arge (17).
8-12
-------
impossible (18). In particular, the free-rider problem would give indivi-
duals incentives to misstate their preferences. For example, if residents
in Farmington, New Mexico were asked how much they were willing to pay to
clean up the air near a power plant, and if they suspected that control
costs would be borne by consumers and owners elsewhere, these local resi-
dents would have an incentive to overstate their willingness-t-o-pay.
Alternatively, if residents suspected that they would be individually taxed
an amount equal to their own willingness-to-pay, then a clear incentive
would exist to understate their own true value, hoping that others would
bid more.
Each approach for asking willingness-to-pay will potentially generate
its own bias. Thus, if recreators are told that the average of their bids
to prevent construction of a nearby power plant would be used to set an
entrance fee, those individuals who suspected their bid to be greater than
the average bid would have an incentive to overstate their willingness-to-
pay. They, in fact, would have an incentive to raise the average bid as
close as possible to their own true bid. In other words, individuals would
have incentives to misstate their own preferences in an attempt to impose
their true preferences on others. It would require a substantial amount of
information to actually behave in this manner. Of course, if the respon-
dents to such a survey do not believe the survey would have any impact on
policy or outcomes, then no incentives for bias exist. The hypothetical
nature of such surveys may then, in actuality, aid in eliciting bids which
are not strategically biased. Alternatively, since payment is not
required, a tendency to exaggerate willingness-to-pay for a preferred
outcome might also exist.
Empirical evidence thus far does not support the existence of
strategic bias among consumers. Bohm (19), in an experimental approach
utilizing actual payments for public television, failed to find strategic
bias significantly affecting the outcome. Scherr and Babb (20) utilized
three different mechanisms for valuing public commodities and found little
evidence supporting the existence of strategic bias. Smith (21), in
laboratory experiments, also failed to find strategic bias as a significant
8-13
-------
problem. Studies at the University of Wyoming have not encountered
problems with strategic bias.
Since contingent behavior or valuation is hypothetical, it is clear
that answers obtained through surveys are not based on information similar
to that which would apply if consumers based answers on real experiences.
One is an ex ante response while the other is an ex post statement.
Typically, consumers do reevaluate decisions on the basis of experience and
gained knowledge. Thus, an individual or household might respond to a
hypothetical decrease in environmental quality at one location with a low
bid, thinking that other nearby sites would make good substitutes.
However, in a real situation the individual might have found that other
sites involved more travel costs and were less satisfactory than imagined.
The information presented to the respondent in a survey situation relating
to substitution possibilities and alternative costs may alter the stated
willingness-to-pay. The respondent must be made aware of proposed alterna-
tives in terms of quality or quantity. Other variants of information bias
might include giving the respondent information as to how other respondents
behaved, whether in the aggregate their bid was sufficient to achieve (or
not achieve) the stated goal (e.g., prevention of visibility deterioration)
or alternative sequencing of questions.
Related to information bias is instrument bias, whereby characteris-
tics of the mechanism for obtaining willingness-to-pay possibly influence
the outcome. Two characteristics of the survey bidding approach are
vehicles for payment and a starting point for initiation of the bidding
process. Studies have recognized that the mechanism used to collect the
bid or pay compensation may influence its magnitude (22). That is, if the
recreator pays a higher park entrance fee rather than another type of tax,
his bid for an environmental attribute may differ. From economic theory,
the bid should differ, if the price of the commodity represented by the
bidding vehicle changes, provided the recreator's substitution possibili-
ties associated with alternative payment mechanisms are different. When a
payment vehicle allows the individual to substitute over a wider range of
current commodities purchased, then the bid or compensation should be
8-14
-------
related to adjustments in disposable income or wealth, where the individual
has the greatest latitude for potential substitution. Practically,
however, a believable payment mechanism related to income adjustment, in
general, cannot be applied. For example, surveys are often taken at
recreation sites away from the individual's locale or state. In this case,
a wage tax may not be viewed as realistically payable by the recreator.
Thus, there is a tradeoff between accuracy associated with a less than
ideal method of payment and the believability of the vehicle for payment or
compensation. The reduction in substitution possibilities for a more
believable payment mechanism is likely to reduce the contingent expenditure
or increase the compensation estimate.
A second type of instrument bias is starting point bias. The contin-
gent valuation approach commences with questions on payment (and/or compen-
sation) for hypothetical changes in environmental attributes. Contingent
bidding surveys to date have asked the recreator (or any type of inter-
viewee) a question with a "yes" or "no" answer rather than a question
requiring explicit calculations. It is presumed that the recreator can
more accurately respond to the yes/no question framework, although to our
knowledge this proposition has not been formally tested for individuals
responding to contingent valuation questions. Given the proposition that
yes/no responses are desirable, often a starting bid or minimal level of
compensation has been suggested. The potential bias arises in suggesting a
starting point from at least two possible sources. First, the bid itself
may suggest to the individual the approximate range of "appropriate bids".
Thus, the individual may respond differently depending on the magnitude of
the starting bid. Second, if the individual values time highly, he may
become "bored" or irritated with going through a lengthy bidding process.
In consequence, if the suggested starting bid is substantially different
from his actual willingness-to-pay, the bidding process may yield
inaccurate or only roughly approximate results. The effect of these two
types of starting point biases may substantially influence the accuracy of
contingent valuation and therefore the usefulness of this approach for
assessment of environmental preferences.
8-15
-------
Tie discussion on information bias suggested that the contingent
valuation approach will give answers dependent upon the information or
"state of the world" described. The contingent valuation approach requires
postulating a change in visibility such that it is believable to the indi-
vidual and accurately depicts a potential change. The change must be fully
understandable to him; i.e., he must be able to understand most, if not
all, of its ramifications. The individual also must believe that the
change might occur and that his contingent valuation or behavioral changes
will affect both the possibility and magnitude of change in visibility. If
these conditions are not fulfilled, the hypothetical nature of contingent
valuation approaches will make their application utterly useless. A test
of hypothetical bias would require that the perturbation proposed would
occur and then the respondents' actual reaction would be evaluated in terms
of the previous hypothetical statements of willingness-to-pay. This, how-
ever, makes it extremely difficult to measure the extent of hypothetical
bias within a contingent experiment since it depends not only on the
structure of the experiment, but also on the "uncontrolled" factors of the
future.
Any survey approach, including the contingent valuation approach, is
subject to sampling bias, nonrespondent bias, and interviewer bias. Any of
these certainly can subject the results of an experiment to question even
if all previously mentioned biases are nonexistent. In a study by Rowe,
d'Arge and Brookshire (23), it was found that from 40 to 50 percent of the
forecasted differences between "true" bids and reported bids was due to
various types of biases. That is, some types of bias can potentially lead
to a 50 percent error in reported bids. Given these qualifications, it is
interesting to note that estimated willingness-to-pay per mile tends to
cluster around $10.00 to $25.00 per household.
We do not attempt to evaluate the extent of the various biases
described above for the five direct willingness-to-pay studies identified
in Table 8-2. While some of the studies did test for the presence of
alternative biases, it is difficult to reach general conclusions on the
likely magnitude of particular biases for contingent valuation studies
8-16
-------
since so much depends on questionnaire design. Detailed discussions of the
biases and the manner in which they are dealt with in questionnaire design
can be found in Brookshire et al. (1) and Rowe e_t al. (23).
Before we present the estimates of visibility benefits for urban
areas, we want to highlight some of the major assumptions which accompany
the development of a benefits number from the contingent valuation studies.
These include the following:
• Willingness-to-pay estimates derived primarily for West
Coast cities are appropriate for the rest of the nation.
(Probably leads to upward bias.)
• That the depictions through photographs in western condi-
tions are appropriate for other regions of the country.
(Depends on locale, but probably contributes an upward bias
because of greater vistas in the West.)
« That one-half of the total willingness-to-pay was due to
aesthetic considerations of poor air quality throughout the
nation. (Bias is probably upward since health considera-
tions may be most important in many eastern urban areas.)
• That average willingness-to-pay does not differ markedly
from marginal willingness-to-pay. (Empirical estimates
suggest average is above marginal, so bias is likely to be
upward for visibility improvements.)
• That the techniques used to obtain empirical estimates are
unbiased. (Bias, if present, is unknown quantitatively but
likely to be upward since there were no budgetary restric-
tions imposed in discovering willingness-to-pay.)
In summary, there are many assumptions underlying the extrapolation of
some experimental estimates to nationwide benefits assessments. Many of
the biases can be at least partially evaluated as to direction, and in most
instances tend to be upward.
Prediction Relationships for Urban Areas
In order to develop national estimates of visibility values in urban-
suburban areas, estimated prediction equations were developed relating
8-17
-------
willingness-to-pay for improved visibility to income and visibility
improvements. The general forms of the equations were:
B = tiYaVP (8.1)
B' = iiYa(AV)P (8.2)
where B = willingness-to-pay (in 1981 dollars) per family per year
for visibility of V rather than no visibility.
B' = willingness-to-pay (in 1981 dollars) per family per year
for an air quality improvement that leads to a visibility
change of V.
Y = yearly family or household income in 1981 dollars.
V = miles of visibility in an urban setting.
AV - improvement in visibility in miles, urban and suburban.
TJ = a constant term.
a, P = elasticities of income and visibility associated with
willingness-to-pay for improved visibility.
Two methods were developed to estimate the coefficients t\, a, and p.
Both involved starting with results from the Los Angeles experiment (1).
In Method I, (J was estimated by using an econometric estimate of AB/ANO*
from Brookshire et al.. and converted to AB/AV using the results in
Appendix 8A, and then recomputed in elasticity form using the means of the
Los Angeles data. The coefficient a was obtained from econometric
estimates found in Brookshire et al. Finally, using the means for Los
Angeles of B, V, and Y, the coefficient r\ was computed. The complete
equation using Method I is then:
B = o.497Y°-566V°-399 (8.3)
While this equation tended to predict well around the means for Los
Angeles, it tended to not predict well for large changes in visibility for
Los Angeles or other cities examined. The Boston study (9) contained an
8-18
-------
equation where B = 0.279Y1*00V°'87 (although property value differences
were used), while the Denver study (3) had an implied form of B =
0.204T ' *V * (although because of the rather arbitrary conversion of
CO into N02 the 0 elasticity for Denver should be suspect).
Method II involved using the average of the natural logarithms of the
means of the household bid and changes in visibility for the Los Angeles
experiment to estimate (J, the elasticity for visibility. Thus, 0 equals In
312/ln 13 = 2.239. Then this value along with a = 0.566 was substituted
into Equation (8.2), again using the means for Los Angeles to estimate the
constant term, n. The resulting equation was:
B' = 0.0039Y°-566(AV)2'239 (8.4)
This equation tended to predict more accurately the value of changes
in visibility across various studies. Using Method II, 0 was calculated to
be 2.239 for Los Angeles, 4.864 for San Francisco (14), 2.411 for Chicago
(16), and 2.854 for Boston where property value differences were used (9).
Thus, the Los Angeles elasticity is the lowest using Method II for the
studies where willingness-to—pay and visibility measures could be inferred
from reported results, and also lower when compared with the Harrison and
Rubenfeld property value study for Boston.
For estimation purposes, both Equations (8.3) and (8.4) are utilized.
The fundamental difference between them is that Method I utilizes the
differences of the logarithms to predict, while Method II uses the
logarithm of the difference (or change in visibility), both with different
weights because of the magnitude of the constant term. As proposed visi-
bility improvements become larger. Method I would tend to predict less than
Method II, while for very small changes the reverse would be true. Method
I is calibrated to the Los Angeles data set, but tends to underpredict for
Boston, overpredict for Denver, and be reasonably close for San Francisco.
Alternatively, Method II consistently underpredicts, for small visibility
improvements, for San Francisco, Chicago, and Boston.
8-19
-------
Table 8-4 records national estimates using the equations resulting
from Methods I and II. Note that Method I yields higher estimates for
small predicted visibility improvements, yet lower ones when the visibility
change becomes larger, as was postulated. Furthermore, both methods tend
toward identical predictions around the 20 percent visibility improvement.
For prediction purposes. Method II is probably most accurate for
intermediate predictions of visibility improvement across all cities, while
Method I would be more accurate for extreme changes and where the base
visibility was very high or very low.
It is also a useful exercise to check how the benefit estimates
derived from Equations (8.1) and (8.2) compare with some gross estimates of
visibility developed simply from the information given in Table 8-3. It
appears reasonable to presume that, in the aggregate within an urban
context, where visibility is typically in the range of seven to 25 miles,
the value that residents place on improved visibility ranges from $5.00 to
$40.00 annually per mile for an average household. A "best estimate" value
would be about $19.00. Given that there are approximately 76 million U.S.
urban-suburban households and that the average improvement of visibility of
20 percent means a mileage increase of seven to ten miles, then urban
residential benefits for a 20 percent improvement would be from $133.00 to
$190.00 per household each year (in 1981 dollars). Nationally, this
translates into an estimate of from $10.0 billion to $14.4 billion per
year. Thus, in terms of averages across studies, it is clear that benefits
to households in urban areas are likely to be in the billions per year for
a 20 percent improvement in visibility per year. Note, however, that this
is willingness-to-pay for all effects inclusive of health effects as
represented by visibility. In the Los Angeles study, only about one-half
of willingness-to-pay can be attributed to aesthetics. Therefore, on a
national basis the aesthetic benefit of improved visibility of 20 percent
would range from $5.0 to $7.2 billion per year based on the aggregate
estimates just given. Note also that the ranges for a 20 percent improve-
ment (medium estimates) are entirely included in the range cited earlier
for the total effects case of gross national averages.
8-20
-------
Table 8-4
LOW, MEDIUM AND HIGH ESTIMATED RESIDENTIAL BENEFITS PER YEAR IN 1980
MILLIONS OF DOLLARS FOR VISIBILITY POLICIES OF 13, 20, 30 MILES AND
A NATIONWIDE 20% IMPROVEMENT*
Policy
13 Miles
20 Miles
30 Miles
20%
Low
Eq. 1**
260
4,230
13,242
6,743
Eq. 2+
39.8
2,618
24,028
6,335
Medium
Eq. 1
1,581
6,986
16,887
7,608
Eq. 2
126
4,010
30,637
14,391
High
Eq. 1
1,893
7,908
18,534
8,555
.
Eq. 2
364
6,094
38,887
27,657
* Let:
B. = Benefits per year for region i.
Y. = Average annual household income for region i.
V- = Initial visibility in region i.
V? - New visibility imposed by the policy alternatives in
H.
region i.
Number of households in region i.
'i = vi-vi-
B = Benefits per year =
defined by Map 1.
t where the regions are
Then Low, Medium and High estimates were calculated by assuming a
different V* (B^ = 0 for unaffected regions).
°-566(V1i)0-399)
** Calculated using:
Bi = (0.497Y.0-566(V2)0-399 - 0.497Y.
+ Calculated using:
BJ^ = (0.0039Yi°'566(AV.)2'239) x Hi<
Note: These numbers must be reduced by at least 50 percent to calculate
the aesthetic benefits of improved visibility.
8-21
-------
It is important to note that the values obtained here are for all
perceived effects of air pollution by those who reside in urban areas. It
was impossible to separate out the aesthetic value of visibility from
perceived human health and other effects of reduced visibility. However,
one study (1) demonstrated that from 22 to 55 percent of the total per-
ceived effect due to visibility changes is associated with aesthetic value.
This proportion would likely be lower in urban areas with low visibility
and higher in suburban areas with little or no perceived health effect. In
consequence, any national estimate of the aesthetic benefit must take this
into account. For example, the benefits estimates recorded in Table 8-4
for a 20 percent improvement in visibility nationwide, for lust aesthetics.
must be adjusted downward to 22 to 55 percent of their values. Using
Equation (8.1), a 20 percent improvement would yield national benefits
estimates of from $1.47 billion to $4.7 billion annually. This is the
approximate range observed for the aggregate estimates discussed earlier,
after the correction had been introduced.
Conclusions
The purpose of this brief analysis was to develop approximate yet
plausible estimates of the aesthetic value of improved urban-suburban
visibility for the U.S. accomplished by using the results of recent direct
willingness-to-pay studies. The results were also cross-checked using
estimates derived from property value studies. From the studies analyzed,
it is clear that urban-suburban visibility values are positive and are
likely to be in the billion dollar plus category. Given the substantial
uncertainties as to precise willingness—to-pay, regional variations in
tastes and preferences, and topography that are not taken into account,
little more can be reasonably proposed.
RECREATION BENEFITS
In this subsection, the five existing original empirical studies
devoted to establishing measures of the value of atmospheric visibility to
outdoor recreators and tourists are critically reviewed and synthesized.
8-22
-------
The results of this review are used to arrive at magnitudes for the
constant b and the exponents p^ and 02 in the multiplicative expression:
In(TWP) = ln(b) + px ln(Y) + 02 ln(V) , (8.5)
where TWP is the total willingness-to-pay for a single outdoor recreation
activity day of the representative U.S. household, Y is that household's
annual income in thousands of dollars, and V is atmospheric visual range in
miles. The absence of any site-specific attributes in this expression
implies that visibility is to be valued independently of its location.
Only three of the five studies have thus far appeared in the refereed
literature. To generate their data, all five studies employ the contingent
valuation techniques variously reviewed, evaluated, and defended in
Brookshire and Crocker (24) and Schulze ejt ajL (25).
Review of Studies
Randall et ml. (22) —
This 1974 study was the first to apply the contingent valuation
technique to the problem of assessing the economic benefits of atmospheric
visibility. The study focused on the 1972 environmental damages that a
2,080 MW coal-fired power plant in Fruitland, New Mexico, and its
associated raw material and transmission facilities, imposed on residents
of and visitors to the Four Corners region of the Southwest. More than 700
people were asked to value the "high," "intermediate," and "low" environ-
mental damages represented by three sets of photographs. The three photo-
graphs in each set depicted power plant stack emission conditions, mine
spoil bank conditions, and transmission line conditions.
The format and reporting of Randall et. al. severely limits its rele-
vance to a discussion of the value of atmospheric visibility to outdoor
recreators. Although the authors assert (p. 141) that respondents con-
sidered reduced visibility to be "... far and away the most serious" of the
three forms of damage, they provide no information supporting this
8-23
-------
this conclusion. Moreover, if the conclusion refers to the stack emissions
pictures, the reader cannot know how respondents separated the visual
impact of the power plant and its stacks from the visibility impact of its
emissions. Further limiting the usefulness of the study for determining
the separable value of visibility is the authors' failure to relate stack
emissions to ambient pollution concentrations; the responsibility to form
this relation was left to the interviewee who did not have to report where
his imagination led him. Having established a measure of the compensating
surplus values the interviewees attached to various representations of
environmental damages, the authors, when they reported these values, chose
to group residents and nonresidents, thus making it impossible to distin-
guish the residential portion of the reported total valuation from its
outdoor recreator component. Finally, because respondents were asked to
express their willingness—to—pay for environmental damage reductions in
terms of electricity bills and/or sales taxes rather than income, the
reported values are, by the Le Chatelier principle [Silberberg (26),
Chapter 9], likely to be biased downward somewhat. Limiting the individual
to varying only his consumption of electricity, or the subset of marketed
goods subject to sales taxes, reduces the alternative ways he has available
to maximize his gains from the visibility improvement.
Whatever the interpretive difficulties Randall .e_t .§_!. pose, they do
provide for our purposes some useful insights. In particular, they convey
information on the signs of 0, and 0~ "* Equation (8.5).
When those Randall et al. resident respondents who did not live on an
Indian reservation were requested to express their willingness to pay
increased sales taxes for reductions in environmental damages, they were
willing to pay $35 annually for a reduction from high to intermediate
damages, and $50 annually for a reduction from intermediate to low
8-24
-------
damages.* Given that the visual range and color contrast intervals are
equal across the three damage levels, this result implies that the incre-
mental benefits of environmental quality improvement are positive and
increasing, which means that fk will be positive.
Only when reporting income elasticities (p. 146) do Randall e± al.
provide separate results for their sample of 526 nonreservation area resi-
dents and ISO recreators and tourists from outside the area. For the
nonreservation residents, they estimated an income elasticity of 0.65 with
a standard error of 0.10 for a move from high to intermediate damages, and
an income elasticity of 0.65 with a standard error of 0.08 for a move from
intermediate to low damages. The 150 recreators and tourists, who were
asked to bid in terms of user access fees, had an estimated income elasti-
city of 0.09 with a standard error of 0.15 for a move from high to inter-
mediate damages, and an income elasticity of 0.16 with a standard error of
0.11 for a move from intermediate to low damages. These results indicate
the possible range in £,, for subgroups of the survey population.
Brookshixe et. ml. (27) —
As its authors note, this study was inspired by Randall et al. (22).
Starting from a pair of photographs representing an undisturbed Lake Powell
environment, 83 outdoor recreators and tourists in the region were asked,
in 1974, to state their willingness to pay daily access fees to prevent the
construction of a 2,400 MW coal-fired power plant having no visible
emissions from 700-foot stacks, and their willingness to pay daily access
fees to prevent the same power plant with readily perceived emissions and
* On p. 147. Randall et_ al. state: "Mean individual household willingness
to pay for abatement, ..., was about $50 annually to achieve situation B
(intermediate damages) and $85 annually to achieve situation C (low
damages)." This statement is inconsistent with their detailed presenta-
tion of results. The $35 and $50 figures we use are consistent with
their detailed presentation.
8-25
-------
visibility impacts from being built.* Equivalent surplus measures of value
were thus obtained.
Some of the features limiting the usefulness of Randall e_t .§_!. (22)
are also found in Brookshire et al. Again, although the high damage repre-
sentation depicted both stack emissions and resulting visibility impair-
ment, no information is provided about the degree of impairment. In
addition, because respondents had to express their bids in terms of access
fees and were thereby constrained to adjust their access to Lake Powell
rather than their consumption of some other good, bids are likely to be
biased upward.** This bias could be accentuated by the requirement that
respondents state their bids in terms of per day access fees: degraded
visibility may result either in a reduced willingness-to-pay for a visit,
reduced visit length, and/or reduced visits. If, contrary to what they
would have preferred to do, some respondents presumed that they were not
supposed to consider the latter two adjustment possibilities, their stated
utility losses from the environmental degradation would be exaggerated.
Since respondents remained uninformed about what they were to assume with
respect to visit lengths and frequencies, an element of noncomparability is
present across respondents' valuations.
Brookshire et al. (p. 338) only present willingness-to-pay magnitudes
for two cases: 1) a movement from a depiction of an undisturbed environ-
ment to one with a power plant having no visible emissions; and 2) a
movement from the same undisturbed environment to a representation of a
power plant with readily perceived visibility impacts. Given additivity
across the cases, the difference in willingness-to-pay between these two
cases is the willingness-to—pay to stop a movement from the no—visible-
emission power plant to the visibility-impacted power plant case. The
difference would then approximate the stated value the 83 recreators and
* The published form of Brookshire et al. only contains line drawings of
the photographs that were used.
** Although one cannot tell from the published version, the same bias is
probably present in the user access fee results of Randall e_t al. (22).
8-26
-------
tourists attached to visibility degradation independently of the visual
impact of the power plant. Upon taking the difference, a weighted mean bid
of $1.33 per day results.* This figure is less than half the weighted mean
bid for stopping a movement from the undisturbed environment case to the
no-visible-emission power plant case.** Brookshire £t al. therefore imply,
contrary to the finding in Randall e_t a_l. [(22), p. 141], that the visual
impact of the power plant contributed more to recreator losses than did the
reduction in atmospheric visibility. However, if respondents associated
the no-visible—emission impact case with unavoidably small reductions in
visibility, and if respondents were indifferent to the presence of the
power plant, then the weighted mean daily bid to stop movement from the
undisturbed environment case to the "small" reduction in visibility case
was $1.81. Thus, as in Randall et al. (22), the results in Brookshire gjt
al. are consistent with f^ being positive. As in Randall et al. (22), no
inferences can be made from Brookshire et al. about the magnitude of fU
since a quantitative measure of the visibility change is not provided the
reader.
Although they do not explicitly provide income elasticity measures for
changes in environmental quality, at least two of the Brookshire et al.
findings are consistent with a small magnitude for p^. First, they test
for differences due to. an income effect between compensating and equivalent
surplus measures of value. No statistically significant differences were
found. Second, they varied the income distributions across the four group
classes of their respondents, including residents. The variations had
"little impact" (p. 344) on an aggregate bid consisting of the product of
visitor days and bids summed across the four respondent groups.
* Treating remote campers, developed campers, and motel visitors separ-
ately, the weights are the product of each group's sample size and mean
visitor days relative to the sum of these products for the same three
groups.
** The standard deviations of the mean bids for each group-case combination
are always less than 30 percent and generally less than 20 percent of
the corresponding daily mean bid. The tails of the distributions of the
bids across cases thus have very little overlap.
8-27
-------
Rove et al. (23) —
This study is the immediate successor to the previous two studies.
Its method (a contingent valuation technique) of acquiring data on the 1976
value of atmospheric visibility and its geographical focus (the Four
Corners region) was the same as its ancestors. In its particulars, how-
ever, Rowe et al. displays considerable evolutionary progress. The three
sets of two photographs employed to generate data for obtaining equivalent
surplus measures of atmospheric visibility contained no prominent man-made
features. A fourth set of photographs was employed to obtain an equivalent
surplus bid on the value of the visual impact of the same power plant used
in Randall et al. (22). At the three times the visibility photos were
taken, visual range in the region was estimated by air traffic controllers
at the Farmington, New Mexico airport to be approximately 75, SO, and 25
miles.
The study samples consisted of 26 outdoor recreationists at a large,
man-made reservoir near Farmington, and 93 Farmington residents who were
interviewed in their homes. One cannot tell from the published report
whether the former group was allowed to adjust visitation rates and daily
access fees simultaneously. The authors separately report (p. 10) the mean
equivalent surplus bids for the two groups. Outdoor recreationists were
willing to pay daily user access fees of $3.00 (12.00 cents per mile) with
a standard deviation of $0.77 to maintain the visibility represented by the
75-mile photographs rather than that represented by the 50-mile photo-
graphs. They were willing to pay $2.53 (10.12 cents per mile) per day with
a standard deviation of $0.65 to maintain the 50-mile rather than the 25-
mile visibility. Thus, although the bid distributions across the photo-
graph sets do overlap to a substantial degree, improved visibility, just as
is probably true in Randall et. al. (22) and Brookshire et_ al. (27), causes
a progressively greater willingness-to-pay for further improvements. The
equivalent surplus bids of Farmington residents in Rowe et al. displayed
the same pattern. Thus, if the outdoor recreationist and the resident
samples in Rowe et al. are considered as separate studies, four out of four
reviewed studies present evidence that fa i* positive.
8-28
-------
Not all the evidence reported in Rowe e_t *!_. is so comforting to the
interpretations earlier given to Randall et_ al. (22) and Brookshire et. al.
(27). Outdoor recreation respondents were only willing to pay $4.06 per
day with a standard deviation of $1.05 to maintain 75-mile rather than 25-
mile visibility. This mean bid is only 73 percent of the sum of the two
bids over the 75-to 50-mile and the 50- to 25-mile intervals. These
results clearly raise doubts about the validity of the bid additivity
assumption earlier made for the Brookshire et al. (27) study.
Rowe et al. provide empirical evidence supporting Randall's et al.
(22) contention that the impact of emissions on visibility plays a far
greater role in respondents' bids than does the visual impact of the power
plant and its stacks. The equivalent surplus bid of outdoor recreationists
for maintaining 75-mile visibility rather than having 25-mile visibility
and a power plant in the middle—distance was only $4.56, a figure only
$0.50 greater than the offer of the same respondents for 75-mile rather
than 25-mile visibility. Similar results were obtained for Farmington
residents.
Rowe et al. do not report income elasticities for outdoor recrea-
tionists. As calculated for a one-step move from 75-mile to 25-mile visi-
bility, the elasticities they report for Farmington residents range from
0.25 to 0.36. Thus, as do the two previously reviewed studies, Rowe et al.
find that stated willingnesses-to-pay for improvements in atmospheric visi-
bility are quite unresponsive to increases in respondent incomes; i.e., (J^
is again estimated to be positive but small. The authors do not present
the income ranges to which these elasticities are meant to apply. It is
quite conceivable that the low reported elasticity is due more to the small
income range that the respondents represented.
R*c (28) —
Though not yet published, Rae adds some new dimensions to the use of
contingent valuation techniques to estimate the economic value of atmos-
pheric visibility. This 1980 study was also placed in the Four Corners
8-29
-------
region, specifically Mesa Verde National Park in southwest Colorado. Visi-
bility impairment due to plume blight was distinguished during the data
generation exercise from that due to regional haze. For the first time,
telephotometer measurements made at the same times as the photographs
allowed the 54 respondents' statements to be related to objective measures
of visual range and color contrast. No prominent man-made features
confounded respondent interpretations of the depicted natural landscapes.
If differences in the natural landscapes are neglected, the valuations
obtained from the two sets of regional haze pictures can be compared with
the values established with the three Rowe et al. (23) picture sets that
did not include a power plant. The estimated visibility valuations corres-
ponding to the two sets of plume pictures used in Rae are perhaps crudely
comparable to the visibility conditions represented by the power plant
pictures in Randall et al. (22), Brookshire et al. (27), and Rowe et al.
(23).
The three contingent valuation studies reviewed earlier say nothing
about the atmospheric visibilities and other attributes at alternative
outdoor recreational sites: the respondents were allowed to form their own
premises about the attributes of these other sites and to keep these formu-
lations to themselves. Thus the range of implicitly considered substitu-
tion possibilities for the depicted sites in these three studies could vary
widely from one respondent to another.* By the Le Chatelier principle, the
wider the range of adjustment possibilities, the less the value the indivi-
dual respondent will attach to impaired visibility at the depicted site.
Rae (pp. 3.12-3.14) placed each of his 54 respondents on nearly the same
footing by explicitly limiting the available substitutes to nine, including
Arches, Bryce Canyon, and Petrified Forest National Parks, San Juan
National Forest, Glen Canyon National Recreation Area, Canyon De Chelly
National Monument, Salt Lake City, Las Vegas, Nevada, and the individual
* Since the range of substitution possibilities the respondent weighs in
his answer is likely to be that which he considers in his actual
behavior, this variation across respondents can be considered an analyti-
cal advantage of the direct asking approach used in Randall e_t al. (22),
Brookshire et al. (27), Rowe et al. (23), and Schulze e_t al. (25).
8-30
-------
respondent's most frequently visited site. The simultaneous influences
upon the value of atmospheric visibility at Mesa Verde of the combinations
of respondent-perceived attributes at each of these sites were captured by
means of a multinomial logit model (29). Included among the stated attri-
butes of Mesa Verde was a daily access fee which formed the basis for
estimating compensating surplus measures of improvements at Mesa Verde in
atmospheric visibility. Unfortunately, the author did not make clear
whether this fee was to be an increment to or a substitute for existing
access fees. Brookshire et al. (27) suffers from the same confusion.
In contrast to the direct inquiries into marginal willingness-to-pay
that the earlier reviewed studies employed, the multinomial logit approach
in Rae infers these willingnesses-to-pay from respondents' stated changes
in choice behavior, including visitation rate changes. Rae postulated that
each of his 54 utility—maximizing respondents possessed a utility function
U(s,x) + w(s,x). where s is a vector of respondent attributes, x is a
vector of site attributes, v is a nonstochastic function reflecting "repre-
sentative tastes", and w is a function that varies randomly in the sample.
Assuming that the choice set of interest is B = (x^, Xj, ... , XQ), and
that the w values are independently and randomly distributed over B, the
probability that a respondent will choose any x can be shown to be (29):
v(s,x)
p(xls.B) = —
yeB
The choice probabilities are invariant to increasing monotonic transforma-
tions of UO) [McFadden (29)]. Given a parametric specification of the
utility indicator v, the expression for p is used to obtain parameter
estimates and is ultimately a basis to predict respondent behavior, given
the set of respondent-stated individual and recreational site attributes.
Because only two sets of slides depicting regional haze conditions
were presented to the respondents, the results of Rae, by themselves, can
provide no insight about the magnitude of pV However, because the "clear"
8-31
-------
conditions represented a visual range of 160 miles, the interval of visual
ranges for which valuations have been estimated in' the Four Corners region
is considerably extended by this study. Rae's 54 respondents in the summer
of 1980 were willing to pay $8.45 (8.88 cents per mile) in daily access
fees for an improvement in visual range from 75 miles to 160 miles.
Rae's results are in agreement with Randall's et al. (22) and Rowe's
e_t JLi- (23) finding that power plants and their associated plumes impose
lesser insults than does general visibility impairment. An improvement
from a severe plume condition (color contrast = -0.04) to a clear condition
(color contrast = -0.24) was worth $4.00 in daily access fees to Rae's
representative respondent.
In his preferred specification from his Table 4-2, Rae shows that
respondents with an annual income of $10,000 were willing to pay a daily
access fee of $4.55 for an improvement in visual range from 75 miles to 160
miles; respondents with annual incomes of $25,000 were willing to pay $8.45
for this improvement; and respondents with annual incomes of $40,000 were
willing to pay $10.74. For each income category, the standard deviation of
the bid estimate was about 30 percent of the bid. The average 1980 income
of the respondents was $25,000. .Using the lower income and bid in each
interval as the base, these results imply that over the $10,000-$25,000
annual income interval, the income elasticity of demand for the visibility
improvement is 0.58, while over the $25,000-$40,000 interval it is 0.45.
However, when the upper end of each interval is used as the base, these
elasticities increase to 0.77 and 0.55, respectively.
Schulze et al. (2) —
This study is the latest entry in the rapidly evolving literature on
the economic value that outdoor recreators attach to altered atmospheric
visibility. Although it too employs a direct asking technique to obtain
compensating and equivalent surplus measures, it adds some significant new
dimensions to earlier treatments. Teleradiometer measurements of the
apparent green contrast in the five sets of five photographs that are used
8-32
-------
to represent various levels of Grand Canyon and a regional visibility for
outdoor recreators are provided. Tie photographs used for outdoor recrea-
tors represent visual ranges from 78 to 235 miles where no man-made
features are present. Separate willingness-to-pay for visibility in the
Grand Canyon alone and in the entire Southwest, and for hazes and for
plumes, are obtained. Rather than limiting the sample to an arbitrary
group of self-selected current users of the Grand Canyon, a stratified
random sample of respondents was interviewed in the summer of 1980 at their
homes in Albuquerque (61 people), Los Angeles (60 people), and Denver (45
people).* Respondents were explicitly told that their bids were to be
interpreted as an increment to a $2.00 daily access fee, thus avoiding the
possibility that the bid might be regarded as a substitute for existing
daily access fees, but unfortunately retaining the possible Brookshire et
al. (27) and Rowe et al. (23) confusions between daily access fee
adjustments and visitation rate adjustments. Visibility improvements in
the "Grand Canyon only" were tied solely to access fee increases in that
Park. Regional visibility changes were said to involve identical daily
access fee alterations for all regional National Parks.
All compensating surplus bids for the "Grand Canyon only" referred to
the willingness-to-pay in excess of a $2.00 daily access fee for increased
visual ranges rather than having a 78-mile visual range. Thus, if one is
to obtain the incremental bids across ranges, one must assume additivity of
the bids. Making use of this assumption. Table 8-5 presents the mean
incremental bids in cents per mile for each of the three cities. The
visual ranges were calculated from the Schulze e_t .al. photographs by Dr.
Eric Walther of the Visibility Research Center at the University of Nevada-
Las Vegas. Each visual range in the table is an unweighted mean of the
"Grand Canyon haze only" visual ranges in the three photographs in each of
the five sets.
* Another 334 people in these cities were interviewed about their
willingness-to-pay to preserve Grand Canyon and regional visibility inde-
pendently of their desire to visit the area.
8-33
-------
Table 8-5
MEAN INCREMENTAL WILLINGNESS-TO-PAY IN 1980 CENTS PER MILE*
78 to 87 miles
87 to 140 miles
140 to 189 miles
189 to 235 miles
Albuquerque
16.22
1.38
1.84
5.00
Denver
16.33
1.96
2.57
5.52
Los Angeles
22.78
2.79
2.61
8.65
* Calculated from Schulze et al. [(2), p. 66].
With, the exceptions of the move from the 87-140 mile interval to the
140-189 mile interval for Los Angeles respondents, and the moves from the
78-87 mile interval to the 87-140 mile interval for respondents in all
three cities, the rep-orted incremental willingnesses-to-pay of Table 8-5
display the positive and increasing incremental valuations for improved
visibility that appear with varying degrees of believability in the
previously reviewed studies. The relatively high incremental bids reported
for the 78-87 mile interval are nevertheless quite disconcerting. They
have no obvious and secure explanation. However, in distinct contrast to
the four other contingent valuation studies reviewed here, the standard
deviations of the bids in Schulze et al. are consistently about as large as
the mean bids. The use of the 78-mile visual range as the initial range in
the bidding sequence might be another clue to the extraordinarily high 78-
87 mile bid. If, for purposes of the interview, respondents chose to
interpret the 78-mile visibility as complete darkness, then the mean incre-
mental willingness-to-pay over the 0-87 mile visual range interval would be
1.68 cents, 1.71 cents, and 2.36 cents per mile for Albuquerque, Denver,
and Los Angeles, respectively. In the absence of a definitive answer, we
choose to treat the incremental values attached to the 78-87 mile interval
8-34
-------
as an aberration: the evidence for 02 being positive from the four earlier
studies and the last three rows of Table 8-5 dominates.
Schulze &_t .§_!. also provide equivalent surplus statements of 1980
willingness-to-pay in excess of $2.00 daily access fee to prevent plume
blight in what would otherwise be (~ 150-200 miles) visual range Grand
Canyon scene. The plume appears as a horizontal brown band lying just over
some distant mountains. The sky in the photographs is otherwise mostly
blue with some patchy cirrus clouds. No man-made objects such as stacks or
power lines interfere. Albuquerque respondents were willing to pay an
average of $3.18 with a standard deviation of $3.26 to prevent the plume;
Denver respondents were willing to pay $4.80 with a standard deviation of
$6.98. Since these willingness-to—pay magnitudes are similar to the same
respondents' expressed offers to move from haze-impaired 78- to 189-mile
visual ranges, support is offered for the findings of Brookshire et al.
(27) on the relative contributions to total bids of stacks and plumes
versus atmospheric hazes. However, Randall et al. (22) and Rowe et al.
(23) are still in place: they contradict Brookshire et al. (27) and
Schulze et al. (2). No dominating evidence exists; the question of the
relative values of the visual impacts of power plants and stacks and the
visibility impairments of atmospheric hazes remains unsolved.
By adding photographs of vistas in Mesa Verde and Zion National Parks
to the already depicted Grand Canyon haze-impacted vistas, and asking the
respondent to consider the set o'f five photos as representative of visi-
bility in the entire Southwest, Schulze et al. obtain an equivalent surplus
measure of the value of regional atmospheric visibility. The same 166
Albuquerque, Denver, and Los Angeles respondents were asked their
willingness-to-pay to prevent a decline in regional visual ranges from a
five-photograph unweighted average visual range of 135 miles to a similar
unweighted average of 102 miles. The highest and lowest visual ranges in
the 135-mile set were 170 miles and 99 miles, respectively. In the 102-
mile set, the highest range was 129 miles and the lowest was 82 miles.
8-35
-------
1 1
All bids for regional visibility were to be expressed by respondents
as an addition to a daily $2.00 access fee being charged at all regional
National Parks. Albuquerque respondents were willing to pay $3.16
(standard deviation = $3.55) or 9.58 cents a mile to keep regional visual
ranges at the 135-mile average; Denver respondents were willing to pay
$4.93 (standard deviation = $14.83) or 14.94 cents a mile; and Los Angeles
respondents were willing to pay $4.77 (standard deviation = $7.60) or 14.45
cents a mile [Schulze et al. (2), p. 68]. These per-mile of regional
visual range equivalent surplus valuations are five to seven times larger
than the "Grand Canyon only" 87- to 140-mile visual range valuations of
Table 8-5. This is to be expected since all regional site substitutes for
the Grand Canyon are also supposed to have had their visual ranges
similarly impaired. Most importantly perhaps, the comparison between the
regional calculations and those of Table 8-5 provides one piece of
empirical evidence of the caution that must be exercised when studying the
value of atmospheric visibility, when extrapolating valuations from one or
more independent site-specific studies to a regional or national valuation.
Extrapolations of this sort will seriously underestimate the value of
atmospheric visibility.
Measures of jjj, the income elasticity of demand for visibility, are
also given in Schulze e_t a_l. [(2), p. 78], provided that the expressions
were estimated in multiplicative form.* When this interpretation is
attached to the expressions in their Table 18, the income elasticity demand
for the "Grand Canyon only" visibility impairments was only 0.05; for the
regional impairments it was 0.10. These elasticities, if indeed they are
such, are much lower than those reported in the earlier reviewed studies.
Given the uncertainty about their actual meaning, we choose to disregard
them.
* Schulze ^t. il.. [(2), pp. 77-78] fail to indicate the functional form in
which the bid expressions in their Table 18 were estimated. It seems
unlikely that the expressions were linear in the original variables since
b (Mean income/Mean bid at 140 miles) = 0.05 (28,590/$2.77) = 516.
8-36
-------
A Synthesis —
In the preceding pages, we have reviewed the only five original
studies of which we are aware that attempt to estimate the value to outdoor
recreationists and tourists of atmospheric visibility. Though all focus on
one or more areas in the Southwest and all employ a contingent valuation
technique to generate data, they can hardly be regarded as perfect repli-
cates. Differing sorts of confounding problems are present in each study:
some include man-made objects in the photographs presented to respondents,
while others do not; some refer only to qualitative visibility measures,
while others provide actual measures of the visual ranges and color con-
trasts the photographs represent; and residents and outdoor recreators are
considered separately in most but not all studies. Some studies obtain
compensating surplus measures of value, while others choose the equivalent
measure. If income effects are large, substantial discrepancies in valua-
tions of the same visual range interval would be expected; however, the
empirical findings in at least four and perhaps five of the studies imply
that income effects are small and perhaps even trivial.
Several fairly obvious faults are common to most and occasionally all
of the studies. Host serious perhaps are several sources of bias in the
estimated magnitudes of willingnesses-to-pay. The sign of the bias will
differ according to whether the respondent is confronted with having to
minimize his losses or maximize his gains. In no study are respondents
asked to state their valuations in terms of an income equivalent: access
fees, sales taxes, and electricity bills are instead the bidding curren-
cies. Adjustment possibilities are thus unrealistically low; the expected
utility losses from visibility impairments will therefore be exaggerated
and the utility gains from visibility improvements will be biased downward.
Similar biases could arise in all five studies if respondents, contrary to
their preferred adjustment mode, assumed the number of outdoor recreational
activity days to be invariant.
A third source of bias arises because only local visual ranges are
perturbed in all but the two most recent studies. If the physical reality
8-37
-------
in the Southwest is that local visibility impairment due to hazes implies
regional impairment, respondents were then given unrealistically high sets
of adjustment possibilities in all studies but Rae (28) and Schulze et al.
(2). These excessively large sets meant that valuations of visibility
impairments were biased downward, with improvements being biased upward.
If throughout most of North America visibility impacts tend to be regional
in scope, the Southwest regional willingness-to-pay in Rae (8.88 cents a
mile for an improvement from 75 to 160 miles) and Schulze et al. (a
weighted three-city mean of 12.90 cents a mile to avoid an impairment from
135 to 102 miles) would seem most apt for extrapolation to other regions.
Because the Schulze e_t a_l. study involves potential, past, and current
outdoor recreators in the region, whereas Rae samples only currently active
recreators, the 12.90 cents per mile valuation of the former study could
include an option value element. In fact, the 4.02 cents per mile differ-
ence between the two studies is 45 percent of the 8.88 cents per mile
valuation of Rae. The sole existing empirical study of the magnitudes of
option values in outdoor recreational activities of which we are aware is
Greenley et al. (30). In a contingent valuation study of water quality in
the Platte River Basin of Colorado, they found that option values increased
the current recreation use values of water quality by 40 percent. Thus, if
a single per-mile value for atmospheric visual range is to be used, we opt
for the Schulze et al. figure of 12.90 cents per mile in 1980 dollars.
Other potential sources of confusion are present, though their exact
impacts upon valuations are not readily identified. With the sole excep-
tion of Schulze e_t .§_!., those studies that state values in terms of daily
access fees fail to indicate to the respondent and/or to the reader whether
the stated fee is to be regarded as an increment to or as a substitute for
existing fees. Everyone treats incremental valuations as being strictly
additive across visual range intervals even though there is weak empirical
evidence from Rowe et al. that this may be untrue. Again, with the sole
exception of Schulze et al.. the sample respondents have been self-selected
current users. This may account for the wide disparity in the standard
deviations of the valuations that Schulze e_t a_l. report and the standard
deviations reported by the other studies.
8-38
-------
In spite of some confoundings, biases, and confusions, the five
studies hold several patterns in common. Most obviously, the income
elasticity of demand by outdoor recreators and tourists for atmospheric
visibility, though certainly positive, is quite low. It appears to lie
somewhere between 0.1 and 0.5, with the weight of evidence leaning more
toward 0.1. A reasonable compromise for p, is thus perhaps 0.2.
Only because of a couple of aberrations in Schulze .e_t al. and a few
rather strained but nevertheless supportive inferences that must be drawn
from some other studies, the empirical evidence that incremental
willingnesses-to-pay increase with respect to improving visibility is only
slightly less convincing. Though highly inconvenient for much of air
pollution policymaking (31), this empirical finding is hardly surprising.
It is well known that visual range and color contrast in a relatively clean
atmosphere is affected much more dramatically by a given addition of fine
particles than is already dirty air. The empirical finding that incremen-
tal valuations increase with improved visibility simply shows that any
tendency toward decreasing incremental utility of visibility improvements
is insufficient to cancel out the contributions that increasing incremental
physical effects make to outdoor recreator and tourist valuations.
Assuming that the responsiveness of willingness-to-pay for local visi-
bility improvements is similar to that for regional improvements, several
of the reviewed studies provide enough information to yield rough estimates
of the magnitude of {J~. Defining f^ as
- _ A(Villingness-to-pay) . Visual range
2 A(Visual range) Willingness-to-pay '
Table 8-6 gives estimates for this parameter, the visibility valuation
elasticity, as calculated from those studies that allow it. Clearly, the
willingness-to-pay for a given increase in visual range increases as visual
range increases. Once visual range has reached 150 miles or so, $2 almost
certainly exceeds unity. There is no credible empirical evidence whatso-
ever from either the economics or the physical science literature on atmos-
8-39
-------
Table 8-6
VISUAL RANGE VALUATION ELASTICITIES (P2>
Using Lower Bound
of Mileage
Interval as Base
Using Upper Bound
of Mileage
Interval as Base
From Rowe e_t al,. (23)
75 to SO miles
0.38
0.48
From Rowe et al. (23),
and Rae (28)*
75 to 160 miles
0.73
0.81
From Schulze et al.
[(2). p. 66]
Albuquerque
87 to 140 miles
140 to 189 miles
189 to 235 miles
Denver
87 to 140 miles
140 to 189 miles
189 to 235 miles
Los Angeles
87 to 140 miles
140 to 189 miles
189 to 235 miles
0.82
1.17
2.96
1.15
1.43
2.67
1.18
1.03
3.46
0.87
1.12
2.45
1.08
1.27
2.00
1.53
0.96
2.25
* Rowe £t .§_!. (23) values were converted to 1980 dollars. The Rowe et al.
figure of $4.38 was used as the value for the 75-mile visual range, and
the Rae (28) figure of $8.45 was used as the value for the 160-mile
visual range.
8-40
-------
pheric visibility that contradicts the pattern displayed in Table 8-6. In
regions with fairly high background visibility levels, small percentage
increases in ambient pollution concentrations will cause even larger per-
centage losses in the economic value attached to atmospheric visibility.
If background visual ranges are extremely lengthy, the percentage losses in
economic value can exceed the percentage increases in ambient pollution by
factors of 2 to 3. Even, where background visual ranges are relatively
low, reductions in visual range will generate economic losses, though the
percentage loss is unlikely to be more than half the percentage reduction
in visual range.
The visibility valuation elasticities of Table 8-6 do not extend down
to 13-, 20-, and 30-mile median summer visual ranges. Nevertheless, given
the behavior of these elasticities display, it is not a fearsome task to
extend them. Two points can be made. First, there probably is not much
variation in the positive magnitude of (^ over the 13-30 mile interval.
Such variation as there is, is likely exceeded by the noise factors
inherent in any economic benefits analysis. Second, the positive magnitude
of $2 over the 13-30 mile interval is small. A magnitude of 0.2 is as
reasonable as any. Thus, we propose using the following expression to
estimate the value per outdoor activity day that the representative U.S.
household attaches to improvements in visual range over the 13-30 mile
interval:
In(TWP) = 0.2 ln(Y) + 0.2 ln(V) (8.6)
where TWP is the household's total willingness-to-pay for a single outdoor
recreation activity day, T is annual household income in thousands of
dollars, and V is visual range in miles. To obtain household annual
willingness-to-pay, any answer obtained to the above expression should be
multiplied by 30, on the assumption that the collection of adults in the
representative U.S. household, whether it has complete darkness or 235-mile
visual range, will annually engage in 30 days of outdoor recreational
activity at sites well removed from its immediate home environs. As
earlier noted, there is nothing whatsoever in the literature giving the
8-41
-------
slightest hint of the response of household outdoor activity days to varia-
tions in visual range.
A nationwide increase of 20 percent in median summer visual ranges
would result in 10- or 11-mile visual ranges throughout most of the Ohio
Valley and Southeast and 100-mile ranges in the Four Corners area. The
rest of the nation would fall between these extremes. On the basis of
Table'8-6, we take the following expression to be a reasonable representa-
tion for a 20 percent increase in visual range over that part of the
country currently experiencing median summer visual ranges in excess of 30
miles.
In(TWP) = 0.2 ln(T) + 0.8 ln(V) (8.7)
As before, the values obtained from the above expression should be multi-
plied by 30 in order to obtain annual household willingness-to—pay. It
should be noted that the 0.8 value attached to ln(V) is the upper bound of
the values from Table 8-6 that one might reasonably attach to a nationwide
average (based on land area) visual range of, say, 50 miles.
Table 8-7 summarizes the low, medium, and high estimated annual
recreational benefits for the four alternative visibility policies. For
the policy involving a nationwide 20 percent improvement in visual range,
Equation (8.6) is used when the initial level of atmospheric visibility is
less than 45 miles, while Equation (8.7) is used when the initial visual
range is greater than or equal to 45 miles.
THE EXISTENCE VALUE OF PROTECTING VISIBILITY
The notion of existence value was first developed by John Krutilla
(32). Krutilla argued that consumers may value preservation of a pristine
natural environment even if they do not expect to use that particular
environment. The classic example is preservation of the large species of
whales. Even though an individual may, for example, never see a blue whale
in person, knowledge that the species survives is of real value and, as an
8-42
-------
Table 8-7
LOW, MEDIUM AND HIGH ESTIMATED RECREATIONAL BENEFITS PER YEAR IN 1980
MILLIONS OF DOLLARS FOR VISIBILITY POLICIES OF 13, 20, 30 MILES AND
A NATIONWIDE 20% IMPROVEMENT*
Policy
13 Miles
20 Miles
30 Miles
20%
Low
59
1,048
3,566
7,392
Medium
247
1,595
4,535
12,296
High
455
2,201
5,606
17,448
* In the following equation let:
B.- = Benefits per household per outdoor recreational activity day
in region i.
Y- = Household income in region i.
V* = Initial visibility in region i (see Table 8-4).
V? = New visibility imposed by the policy alternatives in region
i.
Then Low, Medium and High estimates were calculated by assuming a
different V*.
Estimates calculated using:
B. = ^.0.2^2)0.02 _
and
Bi = (*i°'2(V?)0'8 * Y9-
for y < 45 mi
for V 1 45 miles
To obtain annual benefits per household, B^ was multiplied by 30, pre-
suming therefore that the representative U.S. household participates in
30 outdoor recreational days annually. This result was then multiplied
by the number of households per region and then summed over regions to
obtain total national benefits.
8-43
-------
economic measure, an individual may be willing to pay to preserve the
species. Existence value can include two components. The first is derived
from knowledge that the environment is preserved in the present. The
second is derived from knowledge that the environment will be preserved in
the future for the benefit of future generations. The latter of these two
types of existence values has been termed bequest value.
On a purely subjective basis, existence values in the present are
probably associated with natural wonders of the environment which are
widely known even by people who cannot visit them. Thus, a resident of New
York State may well know of the Grand Canyon and of Yellowstone, yet never
intend to visit either park, but still be willing to pay to preserve these
natural wonders, desiring simply to know that they remain pristine.
The bequest type of existence value may apply to far more ordinary
environments. Thus, parents may wish to know that a particular stretch of
river will remain in a natural state after their own death for use by their
own children or future generations. Hence, existence value may be a more
.#
common phenomenon than is generally supposed.
Unfortunately, to our knowledge, only two studies to date have
attempted to estimate existence value. The first of these studies values
water quality and is only indirectly relevant to the problem of estimating
the existence value of visibility (30). The second study attempts to value
preservation of air quality in the Grand Canyon and surrounding parks and
is directly relevant for estimating the national benefit of visibility
protection (2). We review these two studies in the following subsection.
Studies of Existence Value
The Greenley, Walsh and Young study (to be referred to henceforth as
the G.W.Y. study) investigated the benefits of preventing irreversible
water quality degradation due to mining activity in the South Platte River
Basin of Colorado. G.W.Y. specifically included questions in their survey
to estimate existence and bequest values for nonusers. Two hundred and two
8-44
-------
households in Denver and Fort Collins, Colorado were interviewed using
either a sales tax or a residential water-sewer fee as the vehicle for
collecting the hypothetical payment to preserve the river. The recreation
derived user values (including option value) averaged $79 annually to
preserve the river for the 80 percent of the sample that used or expected
to use the river. This same group was willing to pay an additional $67
annually for existence and bequest values of preservation. The 20 percent
of the sample who were nonusers were willing to pay, on average, $42
annually for preserving the South Platte, a pure measure of existence and
bequest value. This latter group provides a better measure of pure exis-
tence value, since individuals who do not use and do not plan to use a
particular environment can only have existence value for that environment.
If we accept $42 per year as the existence value for all of the sample,
including users, then existence values are about 53 percent of user values,
a significant increase in the total preservation value. Perhaps the most
surprising aspect of the 6.W.7. study is that positive existence values
were obtained at all. The South Platte is not a wonder of the world, nor,
for that matter, recognized on a national basis as an area for environ-
mental concern. Thus, the ratio of existence value to user value
(including option value) of 0.53 may provide a lower bound estimate of
existence value for preservation of environmental quality for more well-
known natural environments.
In contrast to the G.W.Y. study of the South Platte, the study by
Schulze, Brookshire, Walther, and Kelley (to be referred to as the S.B.W.K.
study) of the value of visibility in the Grand Canyon region deals with an
internationally recognized natural wonder.
The S.B.W.K. study can be summarized as follows. During the summer of
1980, over 600 people in Denver, Los Angeles,' Albuquerque and Chicago were
shown sets of photographs depicting both clear visibility conditions and
regional haze conditions. Each set consisted of five photographs ranging
from poor to excellent visibility. The middle picture in each case
approximated average visibility during the summer (the season of peak
visitation). The vistas were three different views from the Grand Canyon,
8-45
-------
one view from Mesa Verde, and one view from Zion. The 8 x 10 inch textured
prints were placed on display boards, each vista a separate row, and each
row arranged with five photographs from left to right in ascending order of
visual air quality (i.e., photograph A = "poor" visibility and photograph E
= "excellent" visibility).
The survey participants were asked either 1) how much they would be
willing to pay for visibility as shown in the five sets of photographs from
worst to best on the day of a visit to the Grand Canyon (an estimate of
user value), or 2) how much they would be willing to pay in higher electric
utility bills to preserve the current average condition — middle picture
— rather than allow visibility to deteriorate, on the average, to the next
worst condition as represented in the photographs of the Grand Canyon or of
the region (an estimate of total preservation value). They were also asked
about their willingness-to-pay in the form of higher monthly electric power
bills to prevent a plume from being seen in a pristine area. To represent
plume blight, two photographs were taken from Grand Canyon National Park.
Both photographs are essentially identical except one has a plume, a narrow
gray band, crossing the entire vista in the sky. The source was not
industrial or municipal pollution, but a controlled burn in the area around
the Grand Canyon. However, the effect was comparable to what a large
uncontrolled industrial source might produce.
The bidding game was designed to reveal the household's willingness-
to-pay for preserving visibility in specific locations as represented in
the photographs. For the interviewees asked the preservation value
questions in the survey, the bids include both existence value and user
value.
The survey had few refusals, partly because of the nature of the
interviews. Typically, interviews were conducted in the late afternoon or
early evening hours in residential neighborhoods. Due to the large size of
the display boards, most interviews were conducted on the front lawn of the
respondent's home. Often, both husband and wife participated jointly in
answering the questions. This was viewed by S.B.W.K. as appropriate since
8-46
-------
the principal question was "how much would you be willing to pay in higher
monthly electric utility bills to preserve visibility at the Grand Canyon
or in the entire Grand Canyon region?" Household members would often
engage in existence discussion before giving a dollar amount. Individual
bids ranged from an average of $3.72 per month in Denver to $9.00 per month
in Chicago for preserving visibility at the Grand Canyon. These average
bids were increased from $2.89 to $7.10 per month per household in the four
cities if visibility preservation was to be extended to the Grand Canyon
region as a whole. Prevention of a visible plume in the Grand Canyon was
worth on the average between $2.84 and $4.32 per month for the four cities
surveyed.
The validity of these survey results depends on the perception by
individuals of visibility conditions as represented by photographs. The
S.B. W.K. study argued that a linear relationship has been shown to exist
between perceived visibility as quantified by individuals and with scien-
tific measures of the apparent contrast in the vista by a multiwavelength
teleradiometer. This close linear relationship between perception of an
actual vista and the apparent contrast of the vista also extends to percep-
tion of visibility conditions represented by slides or 8 x 10 inch color
photographs as was shown in the research presented in Chapter 4 of the
report by S.B.W.K.
The benefit estimates derived from the interview results were extrapo-
lated from the sample population to the country as a whole by applying
statistical techniques to the results of the survey. The bids offered by
interviewees to preserve visibility were statistically related to income as
well as other demographic characteristics. Using an estimated linear
relationship of bids to population characteristics, the study estimated the
value of benefits to residents for the entire nation. This was done by
substituting the average value for these characteristics for each state
into the relationship and calculating the average value of the bid of a
person in that state. This value was then multiplied by the number of
households in the state to get a total bid or benefit.
8-47
-------
The estimates of aggregate national benefits for preserving visibility
from the S.B.W.K. study are given in Table 8-8. The estimates shown in
Table 8-8 include both user and existence values. To estimate the
existence value component alone, we must net out the user value component.
Fortunately, the S.B.W.K. study does provide individual estimates of user
values for the Grand Canyon which are tied to entrance fees per carload.
Household bids per visit to maintain visibility at level C (the current
summer average) rather than have a worse condition, B, on the day of a
visit, averaged $1.08 (from Figure 15, p. 66 of the S.B.W.K. report) for
the three cities (Albuquerque, Los Angeles and Denver) where the Grand
Canyon user value question was asked. Noting that the Grand Canyon had
2,131,700 individual visits in 1979, or about 761,300 household entrances
(we assume one household is equivalent to one carload), a total annual user
value bid would be $1.08 x 761,300, or $822,204. Thus, maintaining visi-
bility at current average levels (C in the study) as opposed to allowing a
deterioration to level B, a decrease in visibility as represented in the
photographs of 31.6. miles, is worth less than $1 million to users. This
contrasts with a total preservation value estimated at $3,370 million per
year for preventing the downward change in visibility. User value is, at
least given the results of the S.B.W.K. study, a negligible component of
Table 8-8
NATIONAL EXISTENCE VALUE BENEFITS FROM S.B.W.K. STUDY
Yearly Benefits From
Total
($ million)
Preserving visibility at the Grand
Canyon (maintaining C vs. B)
Preserving visibility in the Grand
Canyon region (maintaining C vs. B)
Preventing plume blight at the
Grand Canyon
3,370
5,760
2,040
8-48
-------
the benefits of preserving visibility at the Grand Canyon. Clearly, the
Grand Canyon mnst be an exceptional case in that virtually all of the
measured benefits of visibility preservation are derived from existence
value. This is in direct contrast to the G.W.Y. study of the South Platte
which found existence values to be only a fraction of user value. However,
these results are consistent with * priori expectations in that the' rela-
tive size of existence values would logically be related to the fame of a
particular environmental asset. In crude terms, willingness-to-pay for
existence value depends on knowledge of the state of an environmental asset
independent of the use of that asset. It is unlikely that most residents
of the United States would be aware of the state of water quality in the
South Platte, while it is highly likely that many individuals across the
nation would be aware of a decline in air quality at the Grand Canyon.
In summary, the two available studies of environmental existence
values provide a clear indication that existence values may be important.
However, it should be recognized that this is a new area for benefits
research, and extrapolations based on the one study done for visibility in
the Grand Canyon can only be highly tentative at best.
In assessing existence values for visibility we assume that such
values apply only to nationally recognized recreation areas. Table 8-9
presents a list of all National Parks and in addition those national
recreation areas which have at least one million visits per year. The
table also gives visits for 1979 and approximate summer visibility in
miles.* Also included in the table is an indication as to whether or not
•visibility is classed as important to the recreation area.**
* These data are taken from the National Park Statistical Abstract (33).
** The visibility "importance" ranking and data are taken from
Protecting Visibility; An EPA Report (34).
8-49
-------
00
I
Ul
o
Table 8-9
RECREATION AREAS USED IN EXISTENCE VALUE ANALYSIS
Name of Area
1 Ac ad i a
2 Am is tad
3 Assateague Island
4 Blue Ridge Parkway
5 Boston
6 C & O Canal
7 Cabrillo
8 Cape Cod
9 Cape Hatteras
10 Chickasaw
11 Colonial
12 Delaware Water Gap
13 Fred Spot
14 Gateway
15 Geo. Washington Memorial Pkway.
16 Glacier
17 Glen Canyon
18 Golden Gate
19 Grand Canyon
20 Grand Teton
21 Great Smoky Mountains
22 Gulf Islands
23 Hawaii Volcanoes
24 Hotsprings
25 Independence
26 Indiana Dunes
27 Jefferson Memorial
28 Jefferson NEM
29 JFK Center for the Performing Arts
30 J.D. Rockefeller, Jr. Parkway
31 Lake Mead
32 Lake Meredith
33 Lincoln Memorial
34 Mammoth Cave
35 Mount Rainier
36 Mount Rushmore National Memorial
37 Muir Woods
Classification
NP
NRA
NS
NHP
NHP
NM
NS
NS
NRA
NHP
NRA
NMP
NRA
NP
NRA
NRA
NP
NP
NS
NP
NP
NHP
NL 51
NHS
NRA
NRA
NP
NP
NM
State
ME
TX
MD-VA
VA-NC
MA
MD-DC-WV
CA
MA
NC
OK
VA
PA-NJ
VA
NY-NJ
VA-MD
MT
AZ-UT
CA
AZ
WY
NC-TN
FL-MS
HI
AR
PA
IN
DC
MO
DC
WY
AZ-NV
TX
DC
KY
WA
SD
CA
1979
Individual
Visits
in 1000' s
2,787.4
1,261.8
1,674.9
11,700.2
1,777.4
2,832.6
1,196.4
3,922.0
1,516.9
1,434.5
7,172.1
2,002.5
1,010.7
8,773.1
4,678.0
1,446.1
1,656.0
11,321.1
2,131.7
2,446.2
8,019.8
2,965.0
1,632.6
1,118.8
2,002.3
1,606.2
2,328.1
1,859.4
4,130.6
1,356.0
6,155.1
1,849.4
3,352.3
1,384.9
1,516.7
1,245.4
1,227.2
Summer
Visibility
in Miles
38
29
10
12
15
10
15
15
9
17
8
14
12
10
12
48
72
14
75
70
10
8
15
9
10
12
15
12
75
70
34
12
9
25
55
14
Visibility
Classed as
Important
Yes
NO
No
No
No
No
No
NO
No
NO
No
No
No
No
No
No
No
No
Yes
Yes
Yes
No
Yes
No
No
No
No
NO
No
No
No
No
No
Yes
Yes
No
No
(continued)
-------
Table 8-9 (Continued)
oo
i
Ul
Name of Area
38 Natchez Trace Parkway
39 National Capital Parks
40 Olympic
41 Ozark
42 Point Reyes
43 Rocky Mountain
44 San Juan
45 Shenandoah
46 Statue of Liberty
47 Valley Forge
48 Washington Monument
49 Whiskey town
50 White House
51 Yellowstone
52 Yosemite
53 Zion
54 Arches
55 Badlands
56 Big Bend
57 Bryce Canyon
58 Canyonlands
59 Capitol Reef
60 Carlsbad Caverns
61 Crater Lake
62 Everglades
63 Guadalupe Mountain
64 Haleakala
65 Isle Royale
66 Kings Canyon
67 Lassen Volcanic
68 Mesa Verde
69 Mount McKinley
70 North Cascades
71 Petrified Forest
72 Red Wood
73 Sequioa
74 Theodore Roosevelt
75 Virgin Islands
76 Voyageurs
77 Wind Cave
Classification
NP
NSR
NS
. NP
NHS
NP/4
NM
NHPl
NRA
NP
NP
NP
NP
NP8/
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
NP
State
MS-TN-AL
DC-MD
WA
MO
CA
CO
PR
VA
NY-NJ
PA
DC
CA
DC
WY-MT-ID
CA
UT
UT
SD
TX
UT
UT
UT
NM
OR
FL
TX
HI
MI
CA
CA
CO
AK
WA
AZ
CA
CA
ND
VI
MN
SD
1979
Individual
Visits
in 1000 's
9,475.3
2,811.0
2,078.8
1,453.0
1,489.1
2,568.5
1,458.8
1,521.5
1,661.4
3,107.4
1,304.8
1,162.4
1,312.0
1,892.9
2,350.8
1,040.5
269.8
858.0
282.9
558.1
74.5
288.9
721.6
410.7
718.1
110.5
674.0
14.8
804.2
380.0
473.7
251.1
743.0
671.6
413.9
799.6
591.6
549.7
195.3
457.4
Summer
Visibility
in Miles
9
12
25
10
14
70
10
10
11
12
56
12
75
44
72
73
55
29
72
73
73
61
44
15
45
19
80
44
74
15
33
75
14
14
45
20
55
Visibility
Classed as
Important
No
No
Yes
No
Yes
Yes
No
Yes
No
NO
No
No
No
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
-------
The method used to approximate national existence values for visi-
bility can be described as follows. First, the monthly household existence
value bid in dollars for visibility at the Grand Canyon from the S.B.W.K.
study can be approximated as:
Monthly bid = 2.69 Y°'22
where Y is annual household income in thousands of dollars. This bid is
for preserving visibility conditions C as opposed to B or an increment in
visibility of 31.6 miles.* Thus, an estimate of an annual household bid
per mile of increased visibility at the Grand Canyon is:
Annual bid per mile = 2'*? * 12 Y°'22 = 1.02 Y°'22
o 1. o
Since we have argued that existence values are likely to be dependent on
the fame of a particular environment or perhaps cultural asset, we use as
an index of knowledge by consumers of a recreation area the ratio of visits
to each site in Table 8-9 to visits to the Grand Canyon to weight the value
of the visibility presented in the formula above. Hence, we effectively
assume that if a site has one-tenth the visitation of the Grand Canyon,
existence values for. visibility at that site are worth one-tenth of the
value per mile of the Grand Canyon. While crude, we feel that some
adjustment of this sort is essential to approximate even an order of magni-
tude for existence values.
These assumptions imply that each household in the U.S. will have an
existence value for visibility improvement equal to
n Ri
I AV. =— 1.02 Y°'22
L=l ^C
* Conversion from apparent contrast of the photographs used in S.B.W.Z. to
visibility in miles obtained from Eric Walther by personal communication.
8-52
-------
or the sum over n sites of the changes in visibility at each site i (AV-)
times the ratio of recreation visits at site i to the recreation visits at
the Grand Canyon (R^/RQ^) times the annual value per mile of increased
visibility at the Grand Canyon.
Table 8-10 presents results for this calculation aggregated across
households with varying income levels across the United States. The calcu-
lation is made first for all recreation areas listed in Table 8-9. Note
that this calculation includes presumed benefits for urban parks and areas
where visibility has not been deemed an important value as indicated in
Table 8-9. However, one must question the evaluation that visibility would
not be important at the Statue of Liberty, the Washington Monument, or Lake
Powell. In any case, this provides an upper bound estimate for the
benefits of policies where a 13-, 20-, or 30-mile national visibility
standard is imposed or where visibility is improved everywhere by 20 per-
cent. Probably a more realistic assessment of existence value benefits is
presented in the second column of Table 8-10 where only those areas where
visibility is deemed an important factor are included in the calculation.
Finally, it may be true that existence values are only important for the
most famous of national parks, for which national pride or value to future
generations become the dominant values. Using only Acadia, the Grand
Canyon, the Grand Tetons, the Great Smoky Mountains, Yellowstone and
Shenandoah gives column three of Table 8-10. Note that the possible range
of benefits exceeds one order of magnitude from the high to the low
estimates shown in Table 8-10. On the basis of one study of the existence
value of visibility, such a wide range is an honest reflection of our
current knowledge. At this point, we would recommend use of our lower
bound estimate as the most appropriate assumption of existence value
benefits.
NONAESTBETIC BENEFITS OF VISIBILITY IMPROVEMENT
To this point, the discussion has focused primarily on the aesthetic
benefits associated with visibility improvements. In this section, we
briefly review several areas where benefits of a nonaesthetic type may be
8-53
-------
Table 8-10
ESTIMATED EXISTENCE VALUE BENEFITS PER YEAR IN 1980 MILLIONS OF DOLLARS
PER VISIBILITY POLICIES OF 13. 20 AND 30 MILES AND
A NATIONWIDE 20% IMPROVEMENT
oo
i
Ul
Policy
13 Miles
20 Miles
30 Miles
20%
All Recreation
Areas
21.418
81,131
173.330
56,653
All Recreation Areas Where
Visibility is a Factor
4,717
13.227
27,419
23,656
Only the Most
Famous Parks*
2,172
7,194
14,430
10,172
* Only Acadia, Great Smokies, Shenandoah, Yellowstone, Grand Teton and Grand
Canyon are included in this calculation.
-------
generated because of better visibility. Specifically, we consider various
safety aspects of improved visibility. Unfortunately, no empirical studies
of the nonaesthetic benefits of visibility are available. Conceptually,
however, estimation of such benefits is likely to be straightforward.
Clearly good visibility is a requirement for the safe operation of vehicles
on roads, water-ways, or in the air. Additionally, good visibility may be
necessary for productive activities such as surveying or possibly laser
transmission of messages.
Focusing on safety issues, it is clear that the operation of vehicles
may be hindered by bad visibility. Econometric techniques can be applied
to estimate the effect of visibility conditions on takeoffs and landings of
aircraft, road traffic, and water traffic. If it can be shown that the
level of these activities is significantly affected by visibility (and
probably in a very nonlinear way), it is a simple matter to estimate the
lost economic value associated with a decrease in these activiites from a
reduction in visibility from air pollution. It is important to note that
although reductions in transportation activities result from safety con-
siderations, a direct calculation based on lives saved from better visi-
bility is probably inappropriate. This results from the obvious behavioral
response to delay travel rather than lose lives. Obviously, the cost of
delay is far less than the value of lives possibly lost in a transportation
accident. The cost of delays induced by poor visibility is thus an appro-
priate measure of the benefits of better visibility. Finally, it is
readily apparent that only when visibility is reduced in the extreme are
significant costs likely to be imposed. Thus, the possible role of air
pollution in increasing the frequency and intensity of fogs is likely to be
the most important component for benefits assessment.
Some empirical estimates along these lines should be available
shortly. The University of Chicago group under the direction of George
Tolley (with Alan Randall and Glen Blomquist, both of the University of
Kentucky, playing important roles in the research) is attempting to econo-
metrically relate takeoffs and landings of aircraft to visibility
8-55
-------
conditions.* Their model includes safety as well as aesthetic impacts in
that the desire to fly may be increased under good visibility conditions
since the flight may be more pleasurable. This latter aspect may be of
specific importance for private aviators.
NATIONAL BENEFIT ESTIMATE FOR ACHIEVING A 13-, 20-, 30-MILE AND A NATION-
WIDE 20 PERCENT STANDARD
This section discusses the procedures and data utilized in the calcu-
lation of national benefits for alternative visibility standards. Problems
that required attention were identifying the areas affected by each pro-
posed standard, specifying a mechanism to set upper and lower bounds to the
benefit estimate, defining the appropriate unit of analysis for data
collection, choosing discount rates and time horizons, and choosing attain-
ment goals. The subsections that follow discuss each of these issues in
turn and present the benefit estimates for each proposed standard.
VisTiil t**ngo Regions Utilized in the Benefit Calculation
In calculating the willingness-to-pay for the proposed 13-, 20-, and
30-mile standards, as well as nationwide improvement of 20 percent in
visual range, the U.S. was divided into visibility regions. Figure 8-1
presents the median summer visual range in miles, designated by isopleths
for the years 1974-76. The isopleths divide the nation into visual range
regions of 10 miles and less, 10-15 miles, 15-25 miles, 25-45 miles, 45-70
miles, and 70 miles or greater. These visual range regions form the basis
of the results to be presented later in this section. Also presented on
the map in Figure 8-1 are the visual range levels at a multiple number of
sites in the U.S. This information was utilized in delineating the regions
that represent an improvement from the existing level of visibility to a
proposed standard.
Consider first the delineation of the regions affected -by a proposed
nationwide standard of 13 miles. Inspection of Figure 8-1 reveals that
* These comments are based on a telephone conversation with Alan Randall.
8-56
-------
CO
(J1
A A
P: Based on photographic
photometry
It; basfcU on neplielometry data
*; baitd on uncertain extrapolation of
visibility frequency distribution
Additional isopleths have been added to
the figure. See the text for an expla-
nation and underlying assumptions.
II1
Source: U.S. Environmental Protection Agency (34).
Figure 8-1. Median Summer Visual Range (Miles) and Isopleths for Suburban/Non-Urban
Areas, 1974-76 (35)**
-------
many areas of the country already enjoy visibility levels in excess of 13
miles. For instance, all of the West, excluding coastal California, and
much of New York and New England have current visibility ranges of 15 miles
or greater. Thus the task for the 13-mile national standard was to
identify the counties that had current visual range levels below 13 miles.
In identifying the counties and then the number of households in each
county, a new set of isopleths was drawn. Consider the area marked Region
I in Figure 8-1, which includes major portions of the Southeast. In this
region visual range, as indicated by the various data points on the map, is
at best 10 miles. Thus for Region I, all of the counties were identified
as an area that would experience an improvement in visual range as a result
of setting a nationwide minimum standard of 13 miles.
Examining the states of Illinois, Missouri, Arkansas, and Louisiana,
one can see that the current distribution of visual ranges is from 10 to 13
miles. Thus for the 13-mile proposed standard a new region was designated
— Region II in Figure 8-1 — which identifies areas which would experience
improvement from 10 miles to 13 miles. Again, the counties were identified
in Region II. Looking'at Florida and parts of Georgia and South Carolina,
Region II again represents an area which would improve from 10 to 13 miles
in visual range. While not drawn on Figure 8-1, similar "10— to 13-mile"
visual range regions were designated for the Southern California region,
especially in the South Coast Air Basin area. Again, these counties are in
areas that would experience an improvement in visibility under a national
13-mile standard. In this process the counties that would experience an
improvement in visual range were identified nationwide and utilized in the
calculation of the dollar benefits of a national visibility standard of 13
miles. Care was taken to identify major cities in designating a county as
an improved area. If a major city within a county fell across the new
isopleths, the county was arbitrarily placed into the lower visual range
category.
The process for identifying the counties that would experience an
improvement under a proposed 20-mile standard was identical to that
8-58
-------
followed for the 13-nile standard. However, the counties identified for
the 13-mile standard are already isolated and one only has to assume the
improvement in visual range from the initial level to 20 miles rather than
from the initial level to 13 miles. Looking just west of Region II, a
Region III was identified whereby the assumed initial level of visual range
was the "13-mile isopleth" chosen in identifying affected counties for the
13-mile standard. Thus, Region III in Figure 8-1 represents those counties
that would experience an improvement from 13 to 20 miles under a 20-mile
standard. Revion IV represents the counties that would experience a visual
range improvement from 15 miles to 20 miles. New England and California
also have areas where the same procedure was followed in identifying
counties that would be affected by a 20-mile nationwide visual range
standard. In New England, the corresponding regions are denoted by a
" ' ". In the West, the initial isopleths were too narrow to illustrate
the boundaries.
In identifying the counties affected by a 30-mile standard, Regions I,
II and II" were assumed to improve to 30 miles from their initial levels.
Region III was assumed to have a visual range increase of 13 to 30 miles,
while Region IV's improvement was from 15 to 30 miles. Region V went from
20 to 30 miles. As before, however, _a new region was identified for
counties that had a visual range of at least 25 miles and resulted in a new
isopleth representing the 30-mile standard — designated as Region VI.
Again, new isopleths as represented by Region VI were identified for the
West Coast and New England regions (denoted by " ' ").
For the regions identified thus far for the 13-, 20- and 30-mile
proposed national visual range standards, they were in each case evaluated
for the visibility change that would accompany a 20 percent nationwide
improvement in visibility. That is. Region I was assumed to undergo a
change of 10 to 13 miles; Region II, 10 to 13 miles; Region II, 13 to 15
miles, etc. This potentially introduces biases, but given the available
visual range data base it would have been inappropriate to attempt a
further refinement. In addition to assuming that the already designated
regions realized a 20 percent improvement, the procedures for identifying
8-59
-------
affected counties not previously identified was continued. Thus a Region
•
VII, just to the west of Region VI was identified as counties receiving a
20 percent improvement from an initial base visual range of 30 miles.
Again, the appropriate areas in the West Coast and in New England were
identified. Finally, Region VIII was assumed to represent a 20 percent
improvement from a base of 45 miles, while counties in Region IX experience
20 percent improvements from a base of 70 miles.
In designating visual range regions for the calculation of benefits of
alternative standards, an implicit assumption is made for each region.
That is, within a region the visual range is homogeneously improved
throughout the whole region according to the initial base visual range
level. Consider Region IV, where the base isopleth visual range level is
15 miles and counties falling into this region are assumed under the 20-
mile national standard not to exceed this level of visual range. In calcu-
lating the benefits, the incremental visual range improvement for the
region as a whole would be assumed to be 5 miles. Clearly some counties in
the region have initial values of visual range greater than 15 miles. Thus
under a 20-mile national standard these counties would experience less than
a 5-mile visual range improvement. Consequently, if a base visual range
for the region as a whole of 15 miles were utilized in the calculation of
benefits for a 20-mile standard, the resulting dollar benefits would be
viewed as biased upwards. Again, this is due to the non—homogeneous nature
of the visual range in the region.
The initial value problem illustrates the need to place upper and
lower bounds on the visual range benefit estimates. Table 8-11 presents a
range of initial visual range values for the nine regions. These were the
values utilized in the benefit calculations. Since Region I is the only
area with visual range values below the surrounding isopleths, a "low
value" or base value of 8 miles was chosen. For Region II through Region
IX the low value corresponds to the lowest visual range in the region and
thus is the lower bound isopleth. Again, the use of the "low value" will
represent an upper bound of the visual range benefit estimates for each
proposed standard in each region. The values chosen as "high" for each
8-60
-------
Table 8-11
LOW, MEDIUM AND HIGH VISIBILITY VISUAL RANGE VALUES BY REGION
Region
I
II
III
IV
V
VI
VII
VIII
IX
Low Initial Value Assumed
for Regions Visual Range-
Yields Upper Bound
Benefit Estimate
8
10
13
IS
20
25
30
45
70
Medium Initial Value Assumed
for Regions Visual Range-
Yield Intermediate
Benefit Estimate
9
11.5
14
17.5
22.5
27.5
37.5
57.5
80
High Initial Value Assumed
for Regions Visual Range-
Yield Lower Bound
Benefit Estimate
10
13
15
20
25
30
45
70
89
00
I
-------
region in Table 8-11 represent the highest level of assumed initial overall
1 •
visual range in the region. This represents a lower bound to the benefit
estimate. The medium values represent an approximate average of visual
range values for the region. The results of utilizing the bounds discussed
above is to enable the calculation of a range rather than a point estimate
for each proposed standard.
Procedures for Data Collection
The estimation of a national household benefit value for improving
visibility to 13, 20, and 30 miles and for a 20 percent nationwide improve-
ment required data on average household income and the number of house-
holds. Because the three mile-specific policies were not required for all
areas of the U.S., nine basic visibility regions were defined (discussed
previously) and data were collected by region. Unavoidably, the regional
boundaries did not correspond to state boundaries. Therefore, the appro-
priate unit for observation was determined to be counties (and independent
cities where appropriate). This also caused a problem, in that the visi-
bility boundary did not always correspond to a county boundary. The place-
ment of questionable counties was determined by the location of the major
cities, thus ensuring that the majority of the households would be placed
in the appropriate region. In the rare instances where the visibility
boundary actually bisected a major city, the city (and corresponding
county) was placed in the lower visibility region.
The counties of each region were identified by U.S. Bureau of the
Census Maps (36). Unfortunately, the number of households and average
income per household was unavailable at the county level. This required
that county data be collected on the number of people and the income per
capita (37). This information was aggregated by state (note: a single
state could contain several regions, thus requiring several aggregates).
These aggregates were then transformed into the desired information by
using the state's ratio of individuals per household (38). From these, a
regional average was calculated for each of the nine regions.
8-62
-------
All tlie data collected was for 1974. In order to get approximations
for 1980, the U.S. percentage change in households and household income was
applied to each of the regional averages (39). Hopefully, the described
procedures were able to preserve interstate population and income patterns;
however, intrastate changes were unavoidably lost.
National Benefits for « 13-. 2O-. 30-Mile and 20 Percent Improvement
Standard
The previous sections have discussed the delineation of counties into
visual range regions. This section will present the results by proposed
standard for three benefit categories — residential, recreation and exis-
tence values. The underlying assumptions and benefit expressions utilized
in the calculation of benefits have been discussed in earlier sections.
Table 8-12 presents the residential annual household benefits for the
four standards for the lower to higher bound benefit estimate. Following
the discussion in the earlier subsection, two functional forms were
utilized. (See footnotes in the table.)
For achieving the 13-mile standard the range of annual residential
benefits is from $20 to $947 million. The implementation of a nationwide
standard of 20 miles yields a yearly range of $2,309 to $3,954 million,
depending upon the calculation procedure. For the 30-mile proposed
standard, the range of yearly values is $6,621 to $19,443 million. If the
goal of a 20 percent nationwide improvement is sought, the annual household
benefit range is $3,168 to $13,289 million.
Table 8-13 presents the range of estimates for household recreation
benefits for the alternative standards. Note that different functional
forms were utilized for regions above and below an initial visual range
level of 45 miles (see footnote in the table). The lower bound benefit
estimate for the 13-mile standard is $59 million and $967 million for the
20 percent improvement goal.
8-63
-------
Table 8-12
LOW, MEDIUM AND HIGH ESTIMATED RESIDENTIAL AESTHETIC BENEFITS
PER YEAR IN 1980 MILLIONS OF DOLLARS FOR VISIBILITY POLICIES OF
13, 20, 30 MILES AND A NATIONWIDE 20% IMPROVEMENT*
Policy
13 Miles
20 Miles
30 Miles
20%
Low
Eq. !••
130
2,115
6,621
3,372
Eq. 2
20
1,309
12,014
3,168
Medium
Eq. 1
791
3,493
8,444
3,804
Eq. 2
63
2,005
15,319
7,196
High
Eq. 1
947
3,954
9,267
4,278
Eq. 2
182
3,047
19,443
13,829
* Table 8-4 did not represent an "aesthetics only" set of residential
values. In this section, the residential values from Table 8-4 have
been reduced by 50 percent and thus only represent aesthetic residential
values. In the following notes, let:
B^ = Benefits per year for region i.
Y = Average annual household income for region i.
2
AV-
BY
= Initial visibility in region i (see Table 8-4).
= New visibility imposed by the policy alternatives in region
i.
= Number of households in region i.
='V2-V1.
9
= Benefits per year - E B^ where the regions are defined
by Map 1. i=l
Then Low, Medium and High estimates were calculated by assuming a
different V* (Ei - 0 for unaffected regions).
** Calculated using:
B. = [(0.497Y.°-566(V2)0-399 - 0.497Y.°-566(VJ)0-399) zH.]/2
+ Calculated using:
B._ = [(0.0039Yi°-566(AVi)2'239) x E^/2
8-64
-------
Table 8-13
LOW, MEDIUM AND HIGH ESTIMATED RECREATIONAL BENEFITS PER YEAR IN 1980
MILLIONS OF DOLLARS FOR VISIBILITY POLICIES OF 13, 20, 30 MILES AND
A NATIONWIDE 20% IMPROVEMENT*
Policy
13 Miles
20 Miles
30 Miles
20%
Low
59
1,048
3,566
7,392
Medium
247
1,595
4,535
12,296
High
455
2,201
5,606
17,448
* In the following equation let:
B- = Benefits per household per outdoor recreational activity day
in region i.
Y. = Household income in region i.
V* = Initial visibility in region i (see Table 8-4).
V? = New visibility imposed by the policy alternatives in region
Then Low, Medium and High estimates were calculated by assuming a
different V.
Estimates calculated using:
B. = 2(V2)0'02 - Y?-2^)0'2) for V* < 45 miles,
and
B. = (Y.°'2(Y2)0-8 - Y?-2(V^)°'8) for V\ >. 45 miles.
To obtain annual benefits per household, BI was multiplied by 30, pre-
suming therefore that the representative U.S. household participates in
30 outdoor recreational days annually. This result was then multiplied
by the number of households per region and then summed over regions to
obtain total national benefits.
8-65
-------
Tables 8-14 and 8-15 present the existence value benefits. Tie
results in Table 8-14 are, as discussed earlier, dependent upon an exis-
tence value index. The process of calculating the index is shown in the
footnote in Table 8-15. Utilizing the existence value index, the category
"Selected National Parks" represents, for each standard, the lower bound
existence value benefit estimate. For instance, a 20 percent nationwide
improvement existence value benefit is $10,172 million.
A complete range of benefit values for residential, recreation and
existence value is presented in Table 8-16. Utilizing the values low.
medium and high for the initial assumed visual range 1'evels by region by
standard and the alternative functional forms, a "low-low" to a "high-high"
annual estimate can be constructed. For the 13-mile standard, the lower
bound annual estimate is $2,251 million and the upper bound is $3,574
million. These values in Table 8-16 thus represent the extreme bounds on
the annual visibility estimates.
In addition to the benefit numbers reported in Tables 8-12 through 8-
16, a variety of other calculations were made to reflect different
discounting time horizons, discount rates, and attainment periods.
Estimates are available for discounted present values as well as annualized
numbers. Additionally, a variety of alternative aggregation procedures are
also utilized. Tables reflecting these combinations of assumptions are
available from the authors upon request.
1. Brookshire, D. S., R. C. d'Arge, W. D. Schulze and M. A. Thayer.
Methods Development for Assessing Tradeoffs in Environmental Manage-
ment - Vol. II. EPA-600/6-79-00Ib, 1979.
2. Schulze, W. D., D. S. Brookshire, E. Walther and K. Kelley. The
Benefits of Preserving Visibility in the National Parklands of the
Southwest - Vol. VIII of Methods Development for Environmental Control
Benefits Assessment. A draft report for the U.S. Environmental
Protection Agency, University of Wyoming, Laramie, WY, 1981.
3. Bresnock, A. E. Housing Prices, Income and Environmental Quality in
Denver. Unpublished paper.
8-66
-------
Table 8-14
ESTIMATED EXISTENCE VALUE BENEFITS PER YEAR IN 1980 MILLIONS OF
DOLLARS PER VISIBILITY POLICIES OF 13, 20, 30 MILES AND A
NATIONWIDE 20% IMPROVEMENT*
Policy
13 Miles
20 Miles
30 Miles
20%
Selected National Parks**
2,172
7,194'
14,430
10,172
* Let:
B = Benefits.
A = Total existence value index.
H = Number of households in the U.S.
Y = Household income in thousands.
Then the above were calculated as:
B = 1.0215AH0>22H where A varies according to Table 8-15.
** Arcadia, Great Smokies, Shenandoah, Yellowstone, Grand Canyon, and Grand
Teton.
8-67
-------
Table 8-15
TOTAL EXISTENCE VALUE INDEX*
(1980 Millions of Dollars)
Policy
Selected National Parks**
13 Miles
20 Miles
30 Miles
20%
13.42
44.75
89.51
63.27
A
A =
AVi
Si
2131.7
Di
LAV.
Si
ID..
V2131.7/
Total existence value index.
Change in visibility at site i compared to standard visi-
bility.
Visits to site i in thousands.
0 if (standard visibility - Vj) < 0.
Visits to the Grand Canyon in thousands.
Dummy variable, depending upon importance of visibility:
D£ = 0 Visibility is not a factor
D - 1 Visibility is a factor.
**
D£ = 1 for Arcadia, Great Smokies, Shenandoah, Yellowstone, Grand Canyon
and Grand Teton.
8-68
-------
Table 8-16
RANGES FOR THE LOW. MEDIUM AND HIGH TOTAL BENEFITS PER YEAR IN 1980
MILLIONS OF DOLLARS FOR VISIBILITY POLICIES OF 13, 20, 30 MILES
AND A 20% NATIONWIDE IMPROVEMENT*
oo
Policy
13 Miles
20 Miles
30 Miles
20%
Low
Low
2.251
10.357
24,617
20,732
High
2,361
9,551
30,010
20,936
Medium
Low
2,482
10,794
27,409
26,272
High
3,210
12,282
34,284
29.664
High
Low
2,809
12,442
29,403
31,898
High
3,574
13,349
39,479
41,449
* Calculated from Tables 8-4, 8-7, and 8-10. The Low, Medium, and High
categories were obtained by assuming different initial visibility
levels. The Low and High subcategories indicate the range of the
estimates within each of the primary categories.
-------
4. Freeman, A. M. The Benefits of Environmental Improvement: Theory and
Practice. Baltimore, MD: Johns Hopkins University Press, 1979.
5. Trijonis, J. Visibility in the Southwest: An Exploration of the
Historical Data Base. Prepared by Technology Services Corporation for
U.S. Environmental Protection Agency, 1977.
6. Ridker, R. 6. and J. A. Henning. The Determinants of Residential
Property Values with Special Reference to Air Pollution. Review of
Economics and Statistics 49. 1967.
7. Anderson, R. J. and T. D. Crocker. Air Pollution and Residential
Property Values. Urban Studies 8. 1971.
8. Crocker, T. D. Urban Air Pollution Damage Functions, Theory and
Measurement. University of California, Riverside, available through
NTIS:PB, 1970.
9. Harrison, D., Jr. and D. Rubinfeld. Hedonic Housing Prices and the
Demand for Clean Air. Journal of Environmental Economics and Manage-
ment 5. 1978.
10. Smith, B. A. Measuring the Value of Urban Amenities. Jo-urnal of
Urban Economics 5. 1978.
11. Nelson, J. P. Residential Choice, Hedonic Prices, and the Demand for
Urban Air Quality. Journal of Urban Economics 5. 1978.
12. Zerbe, R., Jr. The Economics of Air Pollution: A Cost-Benefit
Approach. Ontario Department of Public Health, Toronto, Canada, 1969.
13. Peckham, B. Air Pollution and Residential Property Values in
Philadelphia. 1970.
14. SRI International. Measuring the Benefits of Air Quality Improvements
in the San Francisco Bay Area. SRI Project No. 8962, 1980.
IS. Blank, F. et al. Valuation of Aesthetic Preferences: A Case Study of
the Economic Value of Visibility. Electric Power Research Institute,
No. RP7855-2, 1978.
16. Tolley, G. and A. Randall. Personal communication, November 1981.
17. Brookshire, D. S. ert !_1. Valuing Public Goods: A Comparison of the
Survey and Hedonic Approaches. American Economic Review (forth-
coming) .
18. Samuel son, P. A. The Pure Theory of Public Expenditures. Review of
Economics and Statistics 36(4). November 1954.
19. Bohm, P. An Approach to the Problem of Estimating Demand for Public
Goods. Swedish Journal of Economics 73. March 1971.
8-70
-------
20. Scherr B. A. and E. M. Babb. Pricing Public Goods: An Experiment
with Two Proposed Pricing Systems. Public Choice 23. Fall 1975.
21. Smith., V. L. The Principle of Unanimity and Voluntary Consent in
Social Choice. Journal of Political Economy 85:1125-1139. December
1977-
22. Randall, A., B. Ives and C. Eastman. Bidding Games for Valuation of
Aesthetic Environmental Improvements. Journal of Environmental
Economics and Management 1. 1974.
23. Rowe, R. D., R. C. d'Arge and D. S. Brookshire. An Experiment in the
Economic Value of Visibility. Journal of Environmental Economics and
Management 7. 1980.
24. Brookshire, D. S. and T. D. Crocker. The Advantages of Contingent
Valuation Methods for Benefit-Cost Analysis. Public Choice 36. 1981.
25. Schulze, W. D., R. C. d'Arge and D. S. Brookshire. Valuing Environ-
mental Commodities: Some Recent Experiments. Land Economics 27.
1981.
26. Silberberg, E. The Structure of Economics: A Mathematical Analysis.
New York: McGraw Hill, 1978.
27. Brookshire, D. S., B. C. Ives and W. D. Schulze. The Valuation of
Aesthetic Preferences. Journal of Environmental Economics and Manage-
ment 3. 1976.
28. Rae, D. A. Visibility Improvement as Mesa Verde National Park: An
Analysis of the Benefits and Costs of Controlling Emissions in the
Four Corners Area. Interim Report, Research Project 1742, Boston, MA:
Charles Rivers Associates, Inc., 1980.
29. McFadden, D. Conditional Logit Analysis of Qualitative Choice
Behavior. In P. Zarembka, ed.. Frontiers in Econometrics. New York:
Academic Press, 1973.
30. Greenley, D. A., R. G. Walsh and R. A. Young. Option Value:
Empirical Evidence from a Case Study of Recreation and Water Quality.
Quarterly Journal of Economics (forthcoming).
31. Crocker, T. D. and B. A. Forster. Decision Problems in the Control of
. Acid Precipitation: Nonconvexities and Irreversibilities. Journal of
the Air Pollution Control Association 31. 1981.
32. Krutilla, J. Conservation Reconsidered. American Economic Review 57.
1967.
33. National Park Service. National Park Statistical Abstract. U.S.
Department of the Interior, Statistical Office, Denver Service Center,
1979.
8-71
-------
34. U.S. Environmental Protection Agency. Protecting Visibility: An EPA
Report to Congress. EPA-450/5-79-008, Office of Air, Noise, and
Radiation, Office of Air Quality Planning and Standards, Research
Triangle Park, NC, 1979.
35. Trijonis, J, and D. Shepland. Existing Visibil ity Levels in the U.S.
Prepared by Technology Service Corporation for U.S. Environmental
Protection Agency, Research Triangle Park, NC, 1979.
36. U.S. Bureau of the Census. County and City Data Book. U.S.
Government Printing Office, Washington, DC, 1978.
37. U.S. Bureau of the Census. Estimates of the Population of Counties
and Metropolitan Areas: July 1, 1974 and 1975. Current Population
Reports. Series P-25, No. 709, U.S. Government Printing Office, 1977.
38. U.S. Bureau of the Census. Statistical Abstract of the United States:
1980. 101st edition, Washington, DC, 1980.
39. U.S. Bureau of the Census. Characteristics of the Population - Vol.
1. 1970.
8-72
-------
APPENDIX 8A
HOW VISIBILITY CHANGES 1ERE CALCULATED
FOR THE PROPERTY VALUE STUDIES
The dollar per mile per year figures were calculated in the following
manner:
a. The figure for A in property value was capitalized with an
interest rate of 10 percent and a life span of 30 years, which
gives a per-year bid for the given level of pollution change.
b. The figures for visibility were calculated by the following
equations:
i) 1 mg/100 cm2 = 10 ug/cm2 (1). (8A)
>y
Assume 10 ug/m part = 0.1 mg/
since 0.1 mg/100 cm = 1 ug/cm
100 cm2 sulfation (2)
3 2
10 ug/m part = 1 ug/cm sulfation (8B)
2
ii) Given a A in sulfation of 0.25 mg/100 cm , we can convert
it to 2.5 mg/cm by Equation (8A).
2'5 m/C1°2 by Equation (8B).
10 ug/m xug/m
x = 25 mg/cm .
2.5 ug/cm of sulfation is equivalent to 0.25 ug/m of
particulates.
iv) To convert particulates (TSP), the following conversion
was used (3) :
AN02 = A part x 0.097
AN02 = -(25 ug/m3) (0.097)
AN02 = -2.43 ug/m3.
v) Finally, to convert to visibility (3):
AN0 -2 43
- AV = n'49 miles
8-73
-------
c. To calculate the WTP per mile per year:
The yearly WTP was divided by A in miles and then adjusted to
1981 dollars.
d. To convert the oxidant measure of pollution to N02 — * V (3):
.... ,. .. . , ,, ^ 1.41(Aoxidant) _ AV
AN02/Aoxidant = 1.41 =? - _. 211 - - AV
since AN02 = -0.211 AV.
To convert C02 -* N02 -» V (3,4):
0.06 ' AC02
0.006 • AGO, ' 10 = NO, =5> - jrTTi - = AV
To convert SO- -» part -» V:
10 ug/m3 = 1 ppb S02 (5) =5> 10 (ig/m3 + 1 ppb S02
= 20 (ig/m3
AN02/ Apart = 0.097
0.097 Apart
-0.211
REFERENCES
1. Department of Health, Education and Welfare. Air Quality Criteria for
Sulfur Oxides. No. AP-50, p. 166, 1969.
2. Anderson, R. J. and T. Crocker. Air Pollution and Residential
Property Values. Urban Studies 8:171-180. 1971.
3. Derived using the ambient pollution components for Los Angeles as
reported in: Brokshire e_t a_l.. Methods Development for Assessing
Tradeoffs in Environmental Management - Vol. II. EPA-600/6-79-001b,
February 1979.
4. Bresnock, A. E. Housing Prices, Income and Environmental Quality in
Denver. Unpublished paper.
5. Crocker, T. Urban Air Pollution Damage Functions, Theory and Measure-
ment. University of California, Riverside, available through NTIS:PB,
pp. 197-668, 1970.
8-74
-------
|