Study of the Economic Effects
of Changes in Air Quality
Final Report
Air Resources Center
Oregon State University
June 1972
-------
The following report summarizes research completed during
the past three fiscal years (May 1969 - June 1972) under a
contract with the Air Pollution Control Office, a unit of
the Environmental Protection Agency, for a Study of the
Eoonamio Effects of Changes in Air- Quality.
The report is concerned with measurement of the economic
changes consequent to the introduction of various air pol-
lution control policies. In particular, the report exam-
ines the impact of alternative policies to restrict open
burning of agricultural fields in the Willamette Valley.
This project was under the joint leadership of Dr. G. W.
Sorenson and Dr. R. C. Vars, Associate Professors of Econ-
omics. Dr. Sorenson is the author of Appendix H of this
report, while Dr. Vars is the author of the remainder of
the report.
R. M. Alexander
Project Director
The work upon which this publication is based was performed
pursuant to Contract No. CPA 70-117 with the Air Pollution
Control Office, Environmental Protection Agency.
-------
STUDY OF THE ECONOMIC EFFECTS
OF CHANGES IN AIR QUALITY
Charles R. Vars, Jr.
Gary W. Sorenson
Oregon State University
Corvallis, Oregon
-------
PREFACE
This report summarizes the research results of a three year study
of the economic effects of changes in air quality. The study specified
a conceptual model within which the benefits of controlling air pollution
can be evaluated, as well as generated new testable hypotheses concerning
the effects of air pollution on consumer activity. Implementation of the
benefit-cost methodology consistent with the conceptual model and its
associated hypotheses focused on the estimation of the various benefits
and costs of controlling smoke produced by open burning of grass seed
fields in the Willamette Valley.
Although the conceptual framework presented in this report was not
originally designed to supplement (and, in certain respects, complement)
the frameworks employed in two recent reports to the Environmental Protec-
tion Agency, we believe that this report does present a useful conceptual
framework intermediate to those proposed by the TRW Systems Group and the
CONSAD Research Corporation.* Our extensions of the theoretical litera-
ture on the economic effects of pollution in appendices A, B, C, and H
provide the foundation to the proposed conceptual framework and its asso-
ciated hypotheses.
In contrast, the benefit-cost analysis of alternative open field
burning control policies presented here contributes new findings for
Oregon State University's continuing research program on field burning.
Although in some respects the questions surrounding field burning are
unique to the Willamette Valley, we would suggest that certain features
of our empirical methodology may be usefully applied to other pollution
problems.
During the course of this study, we received encouragement, guidance,
and review from members of the faculty and staff at Oregon State University,
the Air Quality Control Division of the Oregon Department of Environmental
Quality, county property assessment offices in the Willamette Valley, and
the Environmental Protection Agency. At Oregon State University, we are
particularly indebted to the following individuals: Robert M. Alexander,
Director of the Air Resources Center, for general guidance; Professors
David 0. Chilcote (Agronomic Crop Science) and Frank S. Conklin (Agricul-
tural Economics) for extending our understanding of the Oregon seed indus-
try and its practices, as well as providing data without which Part III of
Kenneth R. Woodcock, A Model for Regional Air Pollution Cost/Benefit
Analysis, a report prepared for APCO, EPA under contract with the TRW Sys-
tems Group (McLean, Virginia: May 1971); and An Economic Model System for
the Assessment of Effects of Air Pollution Abatement Volume I: The OAP
Economic Model System Development and Demonstration), a report prepared for
the Office of Air Programs, APCO, EPA under contract with CONSAD Research
Corporation (Pittsburgh, Pa.: May 15, 1971).
-------
this report could not have been undertaken; Professors Norbert A. Hartman
(Statistics) and Robert G. Mason (Agricultural Economics) for important
contributions to the design and analysis of the research on tourist behavior;
Professors Lars Olsson, Ernest W. Peterson, and William P. Lowry (Atmospheric
Sciences) for their advice, interest, and encouragement during the develop-
ment of a pollution production for the Willamette Valley; Don Poole (Instruc-
tional Resources and Materials Center) for drafting the figures presented
in Parts II and III and appendices A and B; and our colleagues in the Economics
Department, Professors Donald Farness and Fred Miller (on leave to the Oregon
Highway Division for providing the data on recreational activities under-
taken by Willamette Valley residents and tourists that underlies Part IV of
this report. At the Department of Environmental Quality, Air Quality Control
Division, meteorologist Bruce Snyder generously provided assistance and data
without which our work on pollution production functions and, hence, benefit
estimation would literally have been impossible. At EPA, Thomas E. Waddell
has supplied information and steady encouragement over the course of our re-
search. Finally special mention must be given the exceedingly able research
assistance provided by Jagjit Brar, who collected the data, conducted the
statistical investigations, and prepared first drafts of materials which ap-
pear in appendices E and F. We further wish to acknowledge the substantive
assistance of the Air Resources Center secretarial staff in the preparation
of this report. We are especially grateful to Mrs. Susan Wilson who supervised
the production of the report, and Miss Marjorie Lee Howe for her assistance in
typing the report.
-------
Agricultural field burning ten miles
lorth of Eugene, near the Willamette
River, August, 1970.
(Photo by C. Wilkins)
-------
TABLE OF CONTENTS
I. INTRODUCTION 1
A. The Problem: Open Field Burning in the Willamette 1
Valley
B. An Overview of the Report 7
II. EXTERNAL DISECONOMIES, ENVIRONMENTAL QUALITY, AND 9
BENEFIT-COST ANALYSIS
A. Introduction 9
B. Production, External Diseconomies, and Environmental 11
Quality
C. Consumption Activities and Environmental Quality 27
D. Benefit-Cost Analysis of Environmental Quality 34
III. THE COSTS OF ALTERNATIVE OPEN FIELD BURNING CONTROL POLICIES 39
A. Introduction 39
B. An Analytical Framework for Measuring the Costs of 39
Regulating Open Field Burning
C. Changes in Consumers' Surpluses and Producers' 45
Rents Under Alternative Open Field Burning Control
Policies
D. Changes in Agricultural Land Values under Alternative 53
Open Field Burning Control Policies
IV. THE BENEFITS OF ALTERNATIVE OPEN FIELD BURNING CONTROL 57
POLICIES
A. Introduction . 57
B. Visibility Under Alternative Field Burning Control 59
Policies
C. Resident Outdoor Recreational Activity Under 68
Alternative Field Burning Control Policies
-------
TABLE OF CONTENTS
(continued)
V. A CONTRIBUTION TO THE EVALUATION OF ALTERNATIVE 73
OPEN FIELD BURNING CONTROL POLICIES
APPENDIX A: Notes on Two-Sector General Equilibrium Models A-l
of Production and Distribution
A.I Fundamental Production and Income Distribution A-l
Possibilities
A.2 Changes in Factor Supply and Technology A-9
A.3 Technological External Diseconomies and A-19
Production Possibilities
APPENDIX B: A General Equilibrium Analysis of Production B-l
Process Regulation
B.I Analysis in Model of Closed Economy . B-2
B.2 Analysis in Regional Model B-4
APPENDIX C: Private Production Possibilities and Efficient C-l
Government Production of Public Goods
C.I Mathematical Analysis C-l
C.2 Discussion and Implications C-3
APPENDIX D: The Pollution Production Function D-l
D.I The Importance of the Pollution Production D-l
Function
D.2 The Model D-3
D.3 Statistical Results D-6
D.4 Implications and Tentative Conclusions D-ll
APPENDIX E: The Determinants of Agricultural Land Values E-l
in the Willamette Valley
E.I The Model E-2
-------
TABLE OF CONTENTS
(continued)
E.2 Statistical Results E-3
E.3 Conclusion E-5
APPENDIX F: Demand and Supply Response Functions for Grass F-l
Seeds Raised in the Willamette Valley
F.I Specification of Demand and Supply-Response F-l
Functions
Demand for Utilization F-2
Demand for Ending Stocks F-4
Supply-Response Function F-4
F.2 Estimated Demand and Supply-Response Functions F-6
Demand for Utilization F-6
Demand for Ending Stocks F-14
Estimated Supply-Response Functions F-14
F.3 Calculated Demand and Supply Functions F-16
APPENDIX G: Tables G-l
APPENDIX H: Effects of Air Quality Variation on Tourist H-l
Behavior
I. Introduction H-l
II. Research Design H-4
III. Regression Results H-10
IV. Analysis of Questionnaire Derived Data H-20
Characteristics of Travelers H-20
Characteristics of Parties Changing Plans H-22
-------
TABLE OF CONTENTS
(continued)
The Impact of Air Pollution H-23
Multivariate A.N.O.V. H-26
Change Vs. No Change H-28
Pollution Bad Vs. Not So Bad H-29
Interaction H-30
Data Analysis Extension H-30
V. Summary and Conclusions H-32
VI. Postscript H-37
Appendix 1: The Effects of Changes in Air
Quality on Tourist-Related
Industries in a Region
Appendix 2: Selected Tables from Oregon State
Highway Division Planning Section
Economics Unit
Appendix 3: Interview Instrument
Appendix 4: Questionnaire Derived Data Arranged
by Closed Ended and Open Ended Type
Questions
-------
I. INTRODUCTION
A. The Problem: Open Field Burning in the Willamette Valley
Oregon dominates United States production of certain grass seed crops,
in large measure because Willamette Valley producers adopted the annual prac-
tice of open field burning to control seed diseases and to dispose of harvest
residues.
"In 1968 Oregon produced 41 percent of the total U.S. grass
and legume seeds on approximately 308,000 acres. The value of
Oregon's total seed crop to farmers was about 31 million dollars.
The Willamette Valley accounted for 25 million dollars, or nearly
81 percent of the total state seed crop value. Nearly 43 percent
of the total state seed crop value was in Linn County.
Grass seeds were grown on about 250,000 acres at a value of
26 million dollars. Of this amount 86 percent, or 22.4 million
dollars, was in the Willamette Valley. About one-half (13 million
dollars) was in Linn County.
Ryegrass seed was grown on 134,000 acres in 1968 at a value
of nearly 13 million dollars—almost one-half of the total value
of grass seed sales. Essentially all of the U.S. ryegrass seed is
produced in the Willamette Valley, with 75 percent being grown in
Linn County. Most ryegrass seed is grown on poorly drained soils
on which alternative uses are quite limited. About 65 percent
(87 thousand acres) of all ryegrass acreage is annual ryegrass.
Perennial ryegrass is grown on the remaining 35 percent (47 thou-
sand acres).
Cleaning, sacking, and handling added about 12 percent, or
3.7 million dollars, to the value of the 1968 tocal seed crop.
Blending and small packaging of grass seeds has been increasing
within Oregon in recent years."*
Unfortunately, although field burning provides an effective and low-cost
method of residue disposal and disease control, it also produces smoke in suf-
ficient volumes to constitute a major air pollution problem for Oregon authori-
ties. In fact, the Department of Environmental Quality has recently reported
that, despite the decline in complaints from 1969 to 1971, field burning remains
Middlemiss, W- E. and R. 0. Coppedge, Oregon's Grass and Legume Seed
Industry in Economic Perspective3 Special Report 284, Cooperative Extension
Service, Oregon State University, Corvallis, April 1970. p. 7.
-------
its major source of citizen complaint. Although most complaints reported
in Table I.I came from individuals and groups in the Eugene area, locations
in the mid-Willamette Valley - e.g. Salem, Corvallis, and Lebanon - generated
relatively large increases in complaints in 1970 as field burning south of
these cities was purposely increased over past years when southerly winds
were forecast. Overall, the DEQ reported that smoke and visibility reduction
were the most frequent causes of complaint.*
Table 1,1
Field Burning Complaint Summary
Received by: 1968 1969 1970 1971
Department of Environ-
mental Quality 11 1645 306 113
Mid-Willamette Valley
Air Pollution Authority 6 88 186 81
Lane Regional Air
Pollution Authority 127 3409 1241 591
Total 144 5142 1733 785
Source: Field Burning in the Willamette Valley - 1971, Department of Environ-
mental Quality, Air Quality Control Division Report, .April 30, 1972,
Table 2, p. 9.
More objective evidence of visibility restriction, smokiness, and air
quality deterioration during recent summers is reported in Table 1.2 and
appendix D of this report. Table 1.2 shows that smoke occurs with increasing
frequency at both Eugene and Salem as summer progresses. It also provides
evidence that the 1970 DEQ burning program reduced smokiness and increased
visibility in Eugene while producing opposite effects for Salem in August —
a finding consistent with the statistical results we report in appendix D ol
this report.
Since field burning takes place between mid-July and late September, the
visibility reductions associated with this agricultural practice occur during
the height of the tourist and resident-vacation seasons in the Willamette Val-
ley. Given the popular and widely prevalent belief that the desirability of
outdoor recreation-type activities varies directly with visibility, it is
generally believed that field burning adversely affects the Valley tourist
industries. It is presumed that air quality deterioration leads tourists and
residents to select different activities, alter their expenditures, etc.
Department of Environmental Quality, Air Quality Control Division,
Field Burning in the Willamette Valley3 1970 3 staff report dated April 8, 1971,
pp. 3-5.
-------
TABLE 1.2
Smokiness in Salem and Eugene
SALEM EUGENE
Year -'68 '69 '70 '71 '68 '69 '70 '71
JULY
Smoky Days 3644 3533
Smoky Hours
Visibility 6 mi. or less 10 8 8 16 10 12 8 12
Visibility 3 mi. or less 0000 0442
Visibility 1 mi. or less 0000 0011
AUGUST
Smoky Days 5 10 10 5 4 11 7 4
Smoky Hours
Visibility 6 mi. or less 11 16 53 14 15 40 14 8
Visibility 3 mi. or less 0 3 16 2 8 30 3 3
Visibility 1 mi. or less 0000 01001
SEPTEMBER
Smoky Days 15 8 6 6 17 963
Smoky Hours
Visibility 6 mi. or less 92 66 50 19 170 51 35 9
Visibility 3 mi. or less 18 16 10 1 62 42 1 1
Visibility 1 mi. or less 0000 6400
OCTOBER
Smoky Days 11 13 10 11 16 15 10 3
Smoky Hours
Visibility 6 mi. or less 53 85 65 59 67 39 47 5
Visibility 3 mi. or less 5 35 16 8 50 25 3 0
Visibility 1 mi. or less 0000 8300
SEASON TOTAL SMOKY DAYS 34 32 30 26 40 40 26 13
Source: Department of Environmental Quality, Air Quality Control Division,
T?-ie1d Burning -In the Witlconette Valley - 19713 staff report dated
April 30, 1972, Table IV, pp. A-ll.
Note: Smoky days are those days showing a restriction to visibility at the
airport by smoke only, haze only, or smoke and haze on one or more
hourly observations.
Smoky hours are those hourly observations showing restrictions to
visibility by smoke only, haze only, or smoke and haze.
A weather element is listed as restricting visibility when it re-
duces prevailing visibility to six miles or less.
-------
Consequently, many argue that a complete ban or some other stringent regula-
tion of open field burning may raise incomes of the owners of specialized
resources in the tourist industries while simultaneously improving air quality
in the Willamette Valley.
Of course, some of the adverse effects of visibility reductions have
already been somewhat reduced by the smoke management policy adopted in recent
years by the Oregon Department of Environmental Quality, Air Quality Control
Division. The DEQ has attempted to minimize visibility losses attributable
to open field burning smoke in Eugene and Salem by controlling daily acreages
burned according to prevailing and predicted meteorological conditions. This
program has been reasonably successful (see Table 1.2 and appendix D) in
decreasing, though not eliminating, one major and well-established effect of
open field burning smoke without directly affecting production practices or
total acres burned.*
In contrast to smoke management, improvements in visibility obtained
by partially or completely eliminating open field burning would increase total
costs per acre in producing grass seed (see Table 1.3) and/or significantly
reduce yeilds per acre (see Table 1.4). In addition to raising production
costs, however, published and unpublished studies suggest that in the absence
of extensive burning there would be (1) increases in the incidence of seed
diseases, (2) decreases in seed purity, and (3) increases in seed cleansing
costs-** Since the superior quality of Oregon seeds has been a major factor
in the successful growth of the Oregon seed industry, increases in disease and
impurities suggest that traditional markets for Oregon seeds would decline
This judgment should not be interpreted as indicating that either the
DEQ staff or the authors of this report see no room for improvement in the
DEQ smoke management program. Evidence presented in Part IV and appendix D
suggests improvements are possible but unlikely to prove as substantial in the
future as those achieved over the 1970 and 1971 burning seasons unless all
burning in the southern half of the Valley is done under southerly wind
conditions.
Agricultural Field Burning in the Willamette Valley, Air Resources
Center, Oregon State University, Corvallis, January 1969; John R. Hardison,
Justification for Burning Grass Fieldss in Proceedings; 24th Annual Oregon
Seed Growers Conference, Corvallis, 1964; and David 0. Chilcote, Department
of Agronomic Crop Science, personal letter with accompanying tables dated
May 30, 1972.
-------
TABLE 1.3
A summary of increases in total costs per acre over open field burning
with alternative residue removal techniques on selected grass seed crops.
Annual Perennial Highland Fine Merion
Alternative Residue Ryegrass Ryegrass Bentgrass Fescue Bluegrass
Removal Techniques $/A $/A $/A $/A $/A
A. Incorporation of residues
into the soil $21-$26
B. Mobile field incinerator $ 5-$10 $ 5-$10 $ 5-$10 $ 5-$10 $ 5-$10
C. Mechanical removal of
residues followed by
field sanitation a/
1. Bunching and Field
Bucking $12-$16 $11-$14 $10-$12 $10-$12 $ 9-$ll
2. Stack Former and
Mover $15-$25 $13-$21 $11-$17 $11-$17 $11-$15
3. Chopper-Blower
and Hauling b_/ $43 $34 $25 $25 $22
4. Baling and Hauling cj $24-$39 $18-$28 $18-$28 $18-$28 $16-$24
5. Field Cubing and
Hauling d_/ $34-$68 $34-$52 $25-$37 $25-$37 $22-$31
aj Costs include an $8/acre charge for use of a mobile field incinerator but
exclude any expenses which may be required for residue utilization or disposal.
b/ Due to a lack of data, no range in costs were calculated. Only a custom rate
of $10/ton for chopping, blowing, and hauling, and $8/acre for field sanitation
were used.
cj Projecting a range in baling and hauling costs of $6 to $12/ton with no swathing
required.
d/ Projecting a range in cubing and hauling costs of $10 to $17/ton.
Source: Conklin, Frank S. and R. Carlyle Bradshaw, Former1 Alternatives to
Open Field Burning: An Eoonomia Appraisal, Special Report 336, Agricultural
Experiment Station, Oregon State University, Corvallis, October 1971, pp. 13.
-------
TABLE 1.4
Etfccts of Various Soud Crop Residue Managumoiu 1'rogrnms on Seed Yield,
by Species, Treatment, ami Sdmplo Plot
Sj>ec lee
Chcwlngs Fescue
Red Fescue
Bentgrass
Orchardgrass
Bluegraas
Perennial Ryegrass
Location 1
Location 2
1966
BE
BL
CR
CR
S
CRP
BE
BE
CR
S
CRP
BE
CRP
CR
R
S
BE
BE
BE
BE
CR
R
BE
CRP
BL
BL
CR
S
CS
1967
BE
BE
CR
CR
S
BE
BE
CR
CR
BE
S
CRP
BM
BM
CR
S
CRP
BE
CRP
CR
R
S
BE
BE
BE
BE
CR
R
BE
CRP
BE
BM
CRP
BM
S
BE
BE
BM
CRB
CR
CSB
CR
CS
CS
1968 I/
BE
BL
CRP
CR
S
BE
BE
BL
CR
R
S
CRP
BE
S
CR
S
CRP
BM
CR
CR
R
S
BE
R
S
BM
S
CS
BL
S
BL
CRB
CRP
CS
CR
CS
CR
1969
BE
S
CRP
CR
S
BE
S
CR
CR
R
S
CR
BE
BL2/
CR
S
CRP
BE
CRP
CR
R
S
BE
BM
S
S
CR
S
100
82
68
56
38
100
96
88
78
78
46
102
100
87
75
58
101
100
98
84
78
64
100
95
86
84
69
62
100
100
99
84
76
70
60
101
87
100
94
86
85
65
60
49
Source: David 0. Chllcote, Department of Agronomic Crop Science, personal letter with accompanying tables,
May 30, 1972.
]/ 1968 was a vet sunuier and burns were not as complete but subsequent yields were good.
2J Poor burn and esser-tlally like no removal. *CRP - chop removed straw and most of S - residue
BE - burned early (Aug.) stubble and propaned (early to
BM - burned mid seisun (Sept.) mid season)
*CR - chop removed straw and moat of stubble *CS - chopped & spread (similar to
no removal)
-------
without burning. For this reason as well as others, the development of the
mobile field sanitizer must be regarded as crucial.*
Finally, we should note that Oregon's dominant position in the produc-
tion of certain grass seed crops derives in large measure from soil and climate
conditions wonderfully well-suited to grass production. Grass seeds are raised
on poorly drained Amity and Dayton soils in the Willamette Valley. The top
soils are excellent but have severe drainage problems caused by an almost
impermeable layer of subsoil. During the winter rains (November to April),
top soils are waterlogged, commonly creating cultivation problems in the spring.
Grasses in contrast to other crops stand these wet-soil and spring conditions
well, and their seeds mature nicely during the Willamette Valley's typical warm,
dry summers. These special conditions mean, however, that few alternatives
exist for these soils without substantial capital investments and/or special-
ized management skills.**
B. An Overview of the Report
For all of the reasons indicated in the previous section, the pollution
problem created by open field burning in the Willamette Valley is somewhat
unusual and poses interesting analytical and empirical questions. In particu-
lar, the seasonal nature of field burning, the special soil and climatic condi-
tions in the Valley, Oregon's dominant position in the U.S. grass seed industry,
and the populations (resident and tourist) affected by field burning smoke com-
bined to suggest difficulties not readily analyzed in previously developed
conceptual frameworks for evaluating alternative policies for controlling air
pollution. Consequently, we undertook to construct a conceptual framework
appropriate to study the open field burning problem in Oregon. In the process
of doing this, we developed both a conceptual framework and testable hypotheses
Since 1970, the Department of Agricultural Engineering, Oregon State
University, has pioneered in the development and testing of a mobile field
sanitizer, which is pulled by a tractor and has an auxiliary engine to power
the blower and hydraulic units. It is desinged to burn all residue and stubble
in its path within a self-sustaining combustion chamber, thereby reducing smoke
emissions 80-90 percent and unburned hydrocarbon emissions by 99 percent when
compared with open field burning. Disease and impurity problems would probably
not arise if mobile field sanitizers were generally used in the Willamette Valley.
This is the conclusion reached in a careful study by Frank S. Conklin
and R. Carlyle Bradshaw, Farmers Alternatives to Open Field Burning: An Econo-
mic Appraisal, Special Report 336, Agricultural Experiment Station, Oregon State
University, Corvallis, October 1971, p. 14.
-------
with more broad applicability than was our original intention. Part II of
this report presents that framework and its associated hypotheses in a very
general and non-specific fashion intended to demonstrate: (1) the merits of
a general equilibrium framework for the analysis of certain air pollution
control problems; (2) new testable hypotheses concerning air pollution and
consumption activity; and (3) how benefit-cost analytical techniques are ap-
plicable in a general equilibrium context where a public good (i.e., air quality)
affects individual welfare.
Parts III, IV, and V employ the conceptual framework and hypotheses
presented in Part II as the mortar to bond the bricks of our benefit-cost
analysis of alternative open field burning control policies together. Part III
examines the costs of alternative control policies and provides estimates of
their magnitudes for the various groups adversely affected by these policies.
New and traditional measurement techniques are employed to determine the im-
pacts of alternative control policies.
Part IV investigates the benefits of improved visibility in the
Willamette Valley to residents and tourists. An optimal burning policy with no
reduction in total acres burned is specified by employing a simple welfare func-
tion, and we develop an index to measure improvements in visibility under each
control policy investigated. Further, we estimate the resident outdoor recrea-
tional benefits associated with visibility improvements under each policy. Rather
surprisingly, a multi-variate statistical analysis of tourist survey data revealed
no support for the hypothesis that tourist behavior is affected by air quality
deterioration due to smoke from open field burning^ Appendix H reports this study
in detail and should be regarded as a major contribution (albeit negative) of this
report.
Part V concludes the main text of the report by drawing together the
"most reasonable" estimates of the benefits and costs of each policy. Although
no evaluative indexes such as benefit-cost ratios are presented in Part V, the
implicit net total, average, and marginal cost of each policy is presented.
Undoubtedly our most interesting finding is that improvements in Willamette Valley
visibility are produced under conditions of decreasing average costs.
Finally, a word concerning the appendices to this report. As already
indicated, appendix H reports a major investigation of the effects of air quality
changes on tourist and resident recreation behavior. Although other appendices
are less massive than H, they provide theoretical and empirical materials without
which the main report could not have been written. New or generally unappreciated
theoretical results are reported in appendices A, B, and C; these results greatly
influenced the structure and content of Part II. In contrast, appendices D, E,
and F report empirical investigations which tested important hypotheses and
ultimately served as the basis of our benefit and cost estimates in Parts III and
IV.
-------
II. EXTERNAL DISECONOMIES, ENVIRONMENTAL
QUALITY, AND BENEFIT-COST ANALYSIS
A. Introduction
Although economists in recent years have devoted considerable
attention to the economic consequences of external economies and dis-
economies, the rapidly expanding literature remains incomplete in a
variety of respects. Given the vast array of situations where exter-
nalities may conceivably be important, as well as the variety of analy-
tical methodologies potentially applicable, no one could realistically
expect the literature to cover all facets of the subject with equal
facility and completeness. As we reviewed the literature in an attempt
to develop a useful framework for investigating the consequences of con-
trolling the pollution created by open field burning ift the Willamette
Valley, we discovered that the literature had examined pollution problems
in situations hardly comparable to the one with which we were concerned.
Thus, we developed a framework incorporating new hypotheses into a modi-
fied version of the standard general equilibrium model to establish a
methodological framework within which benefit-cost analytical techniques
may be applied to evaluate alternative public policies to improve envir-
onmental quality.
Comments on the three special features of this framework appro-
priately precede detailed specification of the production, consumption,
and benefit-cost relationships in the model. The following observations
are also intended to provide some perspective on why the model is pre-
sented as it is.
(1) The model was designed to focus attention on the effects of
regulating pollution generated by an industry (as opposed to a firm)
in a general equilibrium framework that explicitly allowed for inter-
industry flows of materials. This type of model (particularly when
extended to become a genuine regional model) is appropriate in the present
context because the Oregon seed industry produces an intermediate good on
more than one-fourth of the agricultural land in the Willamette Valley for
U.S. markets which it dominates. A general equilibrium methodology must
underlie an evaluation of alternative programs for restricting open field
burning because such programs appear likely to change production costs for
individual Oregon seed growers, market prices of certain seeds, supplies
of seed produced outside Oregon, and agricultural land values in the
Willamette Valley.
Of course, other economists have examined external diseconomies in
similarly specified models previously. In those presentations, however,
the analysis differs from that presented below. Goetz aad Buchanan, for
example, usefully analyze technological external diseconomies contained
within a single industry in a simple general equilibrium model without
intermediate goods; they assert shifts in transformation curves which we
derive, as well as ignore effects of external diseconomies on other
-------
Industries and Lhv "«LaLe of tin- environment." * Ayrr.s and Kmu'Ho proHcmt
a model which allows for interdependent-lo;a on a conaideriibly greater scale
than our model, but they also assert rather than demonstrate the importance
of a general equilibrium approach that allows for interindustry flows of
materials.** Here and in Appendices A, B, and C, we extend the theoretical
literature on pollution control policy matters by deriving new results and
emphasizing the application of benefit-cost analysis to environmental
quality matters.
(2) Our model postulates that individuals have preferences among
various "states of the world" which differ in terms of (a) consumption
activities produced and consumed by individuals and (b) the state of the
environment. Since activities which may or may not require the purchase
of market goods are among the objects of individual concern, the model has
direct applicability to the loss of visibility associated with open field
burning smoke between mid-July and late-September in the Willamette Valley.
This loss of visibility occurs at the height of Oregon's tourist season,
fouling the environment and obscuring the scenic views that attract the highly
mobile tourist population, and attracted many present residents, to the valley.
Presumably the typical tourist's and resident's activities are affected by the
visibility deterioration. Our model provides new hypotheses concerning en-
vironmental quality and consumer activities, as well as a methodological
framework for conc.ucting benefit-cost analysis of alternative public policies
to improve environmental quality.
(3) The model includes a "state of the environment" function that
relates pollution directly to environmental quality. This function permits
derivation of the environmental quality frontier associated with alternative
pollution control policies and alternative private goods outputs, as well as
simultaneous consideration of all major impacts of pollutants —on production,
consumption, and the environment. While fairly complex diagrams are employed
here to derive this frontier, such diagrams are pedagogically useful in re-
vealing relationships not clearly specified in the literature referred to
above or elsewhere. Furthermore, as the empirical research presented in later
sections of this report indicates, environmental quality frontiers can be
estimated to show the relevant trade-offs necessary for rational decision-
making on environmental quality matters.
* C. J. Goetz and J. M. Buchanan, External Diseconomies in Competitive
Supply, American Economic Review, Vol. LXI, No. 5 (December 1971), pp. 883-
890.
** Robert V. Ayres and Allen V. Kneese, Production, Consumption,and
Externalities, American Economic Review, Vol. LIX, No. 2 (June 1969).
10
-------
B. Production, External Diseconomies, and Environmental Quality
We postulate an economy with two outputs, x and y, each of which requires
the employment of the two primary factors, capital (K) and labor (L), and the
other good as inputs. Perfect competition, perfectly inelastic factor supplies,
full employment, constant returns to scale, diminishing marginal rates of sub-
stitution between primary factors, and fixed proportions in intermediate input
use are also assumed. It is further assumed that each industry imposes Pareto-
relevant or technological external diseconomies on itself and/or firms in the
other industry through their influence on technological production relation-
ships, as well as on consumers, and thereby themselves and firms in the other
industry, through their influence on the "state of the environment," as
measured by some index Q.*
The two production functions are:
py, Qy- V V
where K^ and L^ are respectively the capital and labor inputs (i = x,y), y is
the amount of y used as input in the production of x, and xy is the amount
of x used as input in the production of y, where the. bars Indicate that inters
mediate inputs are used in fixed proportions. Let X and Y represent the net
supplies of x and y available for consumption and/or trade, and let a and av
respectively denote the fixed requirement of x per unit of y and of y per unit
of x. Then, the productive capacity of the economy is defined by
(3)
(4)
(5)
(6)
X
Y
L
K
= x - axy
= -ayx + y
= Lx + Ly
— V -L. V
- &x f K.y
*Although our model could easily be extended to cover situations
where one or more consumption activities generate external diseconomies in
production and/or consumption, expositional simplicity and irrelevance to
the problem under consideration (open field burning in the Willamette Valley)
suggested excluding consumption generated external diseconomies from the pre-
sentation here. We do indicate in a later footnote how such externalities
could be introduced into the supply side of our model.
11
-------
Following Vanek and others, a± is assumed constant, and the product axay is
assumed to be less than one. (The latter assumption is the Hawkins-Simon
condition required to ensure that net outputs are positive.)
The supply side of our model is not completely specified by equations
(3)-(6). At this point, equations (1) and (2) have not been interpreted, and
three additional equations remain unspecified. Consequently, prior to com-
pleting the specification of our model, we interpret the meaning and signifi-
cance of production functions (1) and (2).
The functional subscripts in production functions (1) and (2) are
intended to indicate that pollution associated with the production of x and
y, Px and Py, and the "state of the environment," Q, are parameters of these
functions. The subscripts indicate the existence of what here are termed
technologically relevant, parametric external diseconomies. Such external
diseconomies negatively affect firm and industry production functions in a
fashion indistinguishable from what might be termed "negative technological
progress." More precisely, this means that changes in PX, P , and Q shift
the production functions such that
*
for Pi <. P. >. 0,
d ' ff- \> 0 for Q < Q* < Qn
= 0 for Q > Q* <
ax
where i = x, y,
I = K, L, x or y,
P* = threshold amount of pollution beyond which pollution
generates parametric external diseconomies in production,
Q* = threshold value of "state of the environment" index below
which further environmental degradation generates parametric
external diseconomies in production, and
Q = maximum attainable value for environmental quality on the
"state of the environment" index.
Note that this characterization of parametric external diseconomies explicitly
introduces the often-observed threhold phenomena: namely, certain threshold
levels must be attained before an externality has a discernible impact on the
marginal productivity of inputs, I±. Thus, our model explicitly allows for
situations where particular pollutants are technologically irrelevant when
produced in small quantities but technologically relevant when produced in
12
-------
amounts exceeding their respective thresholds. Similarly, it should be
noted that we have deliberately refrained from any statement concerning
rates of change in marginal products once thresholds are attained because
this would involve assumptions concerning what are fundamentally empirical
matter.
Three additional equations are required to complete the specification
of the supply side of our model. Equations (7) and (8) are the pollution
production functions associated with the production of x and y:
(7) Px.j = PXJ (Kx, Lx, yx, x) j - 1, ..., n
(8) Pyk = Pyk (Ky, Ly, xy, y) k = 1, ..., m
Notice that here we place no restriction on the number of types of pollu-
tants associated with the production of x and y. Moreover, as before, we
explicitly allow for the existence of thresholds
i
r > 0 for M. > M* > 0
3M. i 1
i
= 0 for H± < M* _> 0
where i = x, y, r = j, k, and M = K, L, x, y, x, y, without specifying in
detail the characteristics of the pollution production functions.* Our own
empirical investigations (reported in Appendix D) suggest that considerable
caution in this regard is essential.
Finally, equation (9) indicates the relationship between some unspeci
fied index measuring "the state of the environment," Q, and the various
types of pollutants generated by the production of x and y:
(9) Q = Q(Pxl, ..., Pxn; Pyl, ....
where "the state of the environment" is a pure public good (in the Samuel -
sonian sense) enjoyed by producers and consumers alike. Although our
"state of the environment" production function is similar to the usual
production function in the sense that alternative combinations of pollutants
(inputs) can produce the same "state of the environment" (output) , we impose
no restrictions on the function other than the following:
*
The inclusion of final outputs available for consumption, X and Y,
as arguments in equations (7) and (8) would introduce consumption-generated
externalities potentially affecting production and the "state of the envir-
onment ."
13
-------
and
plr
Q = Q_ where all P. < P** > 0.
Tnax ir — ir —
We assume that (a) there is some maximum attainable value for any realistic
and well-defined "state of the environment" index but that (b) empirical
investigations are required to determine relevant pollutant thresholds, mar-
ginal rates of substitution among pollutants, and the negative environmental
returns to scale of pollution.
Equations (1) - (9) and their respective conditions establish the func-
tional relationships underlying the supply side of our model of an economy
where pollution is uncontrolled. However, before investigating the impact
of various pollution control policies on supply conditions, we derive the
transformation curve (in X-Y-Q space) for an economy with no pollution abate-
ment activity. For expositional and geometric simplicity, we assume hence-
forth that (1) the "state of the environment," Q, is affected by two pollu-
tants, P , and P , , generated by production of x and y and that (2) industry
x generates a technologically relevant external diseconomy for industry y.
Given these assumptions, Figure II. 1 is used to derive the X-Y-Q transfor-
mation curve for an economy without pollution controls, U'U, present in
Figure II. 2.
The northwest quadrant of Figure II._1 indicates the dimensions of the
Edgeworth-Bowley production box diagram, OKO'L, and the positions of the
gross and net output transformation curves, ZPZ' and FP'F', respectively, for
our model. Each point on the net output curve corresponds to a particular
point on the gross output curve (for example, net output combinations F, P1,
and F1 require grots output combinations Z, P, and Z', respectively), while
the shape of the gross output curve results from our assumption that indus-
try x generates a technologically relevant external diseconomy for industry y.*
Thus, the net output transformation curve, FP'F1, allows for inter-industry
flows of materials and a production-generated external diseconomy despite its
apparent normal properties.
*
The first section of Appendix A provides derivations of the gross and
net transformation, or production possibility curves, while the third section
of Appendix A specifies the impact of technologically relevant external dis-
economies on production possibilities and derives a curve with a shape similar
to the gross output curve in Figure II. 1. Note that the unusual shape of the
gross output transformation curve does not result in an unusually shaped net
output curve in Figure II. 1 because we have assumed a to be relatively large.
Here, and in Appendix A, production functions are assumed to be homogeneous of
degreee one in order to facilitate graphical presentation of our principal
results.
14
-------
FIGURE E. 1
Y
FIGURE U 2
15
-------
The northwest and southeast quadrants of Figure II. 1 depict the
assumed pollution production functions associated with the production of
y and x, respectively. Although our assumption that Pyi per unit of y
is constant at all levels of output probably corresponds to the normal
view of pollution, we assume that Pxi per unit of x steadily increases
the total amount of pollutants produced. Where input proportions matter,
as they appear to do in many instances, it seems unreasonable always to
postulate the pollutants are produced in strict proportion to output.*
Projecting horizontally and vertically from points on the gross out-
put transformation curve to the pollution production functions, OMX and OMy,
and then projecting once more to determine curve W in the southwest quad-
rant of Figure II. 1, we obtain the combinations of pollutants that deter-
mine the "state of the environment" associated with each combination of
gross and net outputs (for example, points V and V1 correspond to Z and Z',
which in turn correspond to F and F', respectively). Given a "state of
the environment" production function (depicted by curves Qmax, Q-i> Q-2»
etc., which represent sets of alternate pollutant combinations resulting
in successively lower quality environments), the pollutant combinations
along W' may be translated into index numbers measuring the quality of
the environment.
Figure II. 2 summarizes the supply side of the model (sans pollu-
tion controls) in transformation curve UU' by plotting the X-Y-Q results
developed in the northeast and southwest quadrants of Figure II. 1. The
transformation curve UU' is important principally because it is one edge
of the net output/environmental quality transformation surface that we
later argue is required to measure the costs of pollution control pro-
perly. In fact, truly useful benefit-cost evaluations of alternative feas-
ible control policies is impossible without such a transformation surface.
Therefore, we next indicate how the relevant transformation surface can
be derived by generating X-Y-Q transformation curves under alternate pol-
lution control policies. This procedure has four advantages: (1) it is
pedagogically useful in demonstrating the various impacts of pollution
control policies in a general equilibrium setting; (2) it facilitates
explanation of the concept of a "dominant policy"; (3) it indicates why
constraints that reduce feasible policy options, and/or the scale on which
policies may be undertaken, may result in avoidable inefficiencies; and
(4) it demonstrates the major property of the relevant transformation
surface, namely, that an increase in X, Y, or Q, jointly or in pair, is
possible only with a decrease in one or both of the remaining items.**
*
For a critical analysis of the usual view that output and waste pro-
ducts (i.e., pollutants) are produced in fixed proportions, see Richard A.
Tybout, Pricing Pollution and Other Negative Externalities., The Bell Jour-
nal of Economics and Mangement Science, Vol. 3, No. 1 (Spring 1972), pp. 252-
266. Tybout refutes the "neutrality of bribery or compensation" argument;
he does not, however, adopt the general equilibrium approach to pollution
matters suggested in this report.
Combinations of X, Y, and Q on the "relevant transformation surface"
share the property of being technically efficient in the Paretian sense; each
point on the surface describes an X-Y-Q combination such that X,Y, or Q in-
dividually or jointly cannot be increased without some decrease in one or
both of the remainder.
16
-------
Suppose industry x changes its production function (perhaps because
of pollution control legislation), thereby (a) entirely eliminating the
external diseconomy it previously imposed on industry y and (b) partially
reducing PX^ per unit of x. The effects of such a production process
change to reduce pollution are described with the aid of Figures II.3 and
II.4. Since such production process changes are properly viewed as nega-
tive technical progress, gross and net output curves ZZ' and FF' would
shift to ZcZc and FCF^, respectively, while the pollution production
function of the x industry would shift from OMX to OMXC-* By projecting
horizontally and vertically from quadrant to quadrant as before, the new
environmental quality frontier, VCVC, associated with the ZCZC and FCF^
transformation curves is obtained in the southwest quadrant of Figure II.3.
In this instance, the production process change in the x industry improves
the "state of the environment" regardless what composition of output is
ultimately produced. Also note that this would be true even if the pro-
duction process change had not reduced Px^ per unit of x. Thus, we ob-
serve in Figure II.4 that UCUC is above but closer to the vertical axis.
Output would be sacrificed for an improved environment.
In contrast to the impact of production process changes, decisions
to commit resources to pollution abatement activity shift the transforma-
tion curves and environmental quality frontier rather differently. This
is the case because public pollution abatement activity uses resources
and thereby reduces resources available for production (and hence pollution)
without directly affecting the pollutant-generation characteristics of
production.** In Figure II.5, resources devoted to pollution abatement
*
See appendices A and B for derivations of the shifts in the gross
and net output curves shown in Figure II.3. One of the important proposi-
tions demonstrated in appendix B is that production process change in one
industry affects maximum feasible outputs in other industries which employ
the process-changing industry's output as an input. Consequently, the
inward shifts of the gross and net output transformation curves depicted
in Figure II.3 are not incorrect, though at first blush they may appear
unusual to the reader unfamiliar with general equilibrium models that
explicitly allow for inter-industry flows of materials.
** Public pollution abatement activity by itself affects pollutants
generated in private production only indirectly via the changes in factor
proportions it induces in the private sector. Only where aggregate factor
proportions in the private sector are unaffected by public abatement
activity will such indirect (positive or negative as the case may be)
effects be entirely absent. Figure II.5 is constructed on the assumption
that factor proportions are unaffected by pollution abatement activity and,
therefore, it exhibits only the direct effects of such activity as discussed
in the text.
17
-------
FIGURE H. 3
FIGURE IE 4
18
-------
V
Mx
FIGURE E. 5
FIGURE H 6
19
-------
activities (KpK and L L) provide abatement opportunities indicated by
the curve pp', where pollution abatement "outputs", PA^ and PAyl, are
measured in the same units as the pollutants, Pxi and P j^ respectively.
The reduction in available private factor supplies shifts gross and net
output transformation curves inward to Z Z* and FrFr, respectively.
These shifts alone move the environmental quality frontier from W1 to
w1, while efficient pollution abatement shifts the environmental
frontier to VrVr (which is obtained by sliding o"pp' at point o" along
vv1). As in the case of the production process change in the x industry,
however, summary Figure II.6 reveals that output must be foregone to
achieve an improved "state of the environment" by public pollution abate-
ment activity.
The similarity in our findings for these two pollution control
methods does not imply that society would always be indifferent between
the policies. In fact, ruling out political, social and income redis-
tributional questions, it can easily be shown for certain ranges of out-
comes that technical efficiency considerations alone are sufficient to
select one policy over the other, while over other ranges no decision
among policies can be made without knowledge of the decision-maker's
valuations of private goods and environmental quality. We employ
Figures II. 7 and II.8 to demonstrate that this view is correct. These
figures represent the transformation curves and environmental quality
frontiers associated with both policies described earlier; they are
labelled identically to the earlier figures.
Figures II.7 and II.8 reveal that technical efficiency criteria
are sufficient to select between industry x production process change
and public pollution abatement activities only if the desired output
level of Y lies in the range OY»r to OY^. Over that range of outputs on
UrU^. (or F F' ) , outputs of X and Y are larger and the "state of the
environment better than those attainable on UCU^ (or F(,F(1). Thus,
production process change in industry x may be said to be dominated (i.e.,
is clearly inferior to) public pollution abatement activities at the
postulated scale over the range OY™ to OY^. In contrast, with outputs
of Y less than OYjg or greater than OYj environmental quality varies in-
versely with aggregate output. Given the composition of goods output
over these ranges, knowledge of the decision-maker's valuations of pri-
vate goods and environmental quality is required to predict which method
would be selected if these methods were the only ones under consideration.
Note, however, that although UrU in its entirety could conceivably
lie on the X-Y-Q transformation surface associated with technically effi-
cient pollution control methods, only two sections on UCU^ (namely U^G and
EUC) could conceivably be on that transformation surface. Since other
pollution control methods (individually or in varying combinations) could
dominate every section along UrUr and UCUC, one can only conceive—not
predict with certainty—of particular points on UrU^. and U,,!!^ actually
appearing on the net output/environmental quality transformation surface
associated with technically efficient pollution control policies. (We
amplify on this point below.)
20
-------
PA
XI
.0'
FIGURE n. 7
FIGURE E. 8
21
-------
To lllftl point, only I In- <-f!<>rln of I wo pw rl I ( ii I a I po I I n I I on
control methods undertaken at specified scales have been examin<-d.
We have neglected the impact of (1) the full range of control methods,
(2) operated at all feasible scales, and (3) combinations of the vari-
ous control methods operated at their respective efficient scales. We
now turn to these tasks.
Direct regulations of the x industry production function and public
pollution abatement activity constitute only a subset of the vast array
of taxes, subsidies, regulations concerning pollutant emissions, produc-
tion methods, product standards, waste treatment methods and performance
standards, etc. available to reduce if not eliminate pollution and envir-
onmental degradation generated by production activity. Moreover, each
pollution reduction method or policy can be pursued singly or in combina-
tion with others at varying scales and with varying effectiveness.* To
take a particularly obvious case, consider public pollution abatement
activity which conceivably could be undertaken on almost any scale pro-
vided there is public willingness to forego theprivate outputs that would
be sacrificed as factor supplies available for private production decreased,
Referring once more to Figures II.5 and II.6, if public pollution abate-
ment activity were steadily increased without changing factor proportions
in the private sector, gross and net output transformation curves would
necessarily shift toward the origin steadily, as would the environmental
frontier also.** The X-Y-Q combinations associated with these shifts
*
Of course, the framework we have specified here can encompass
any pollution control method or policy which affects (1) relative
product and/or factor prices, (2) factor supplies, (3) production func-
tions, (4) pollution production functions, and/or (5) private and public
pollution abatement facilities or activities. Since little ingenuity is
required to introduce changes of these sorts into the model, or to repre-
sent them in diagrams of the type developed here, we are content merely
to note that the. only pollution reduction policies for which our model
ultimately provides little guidance is where product quality is changed
for pollution control or reduction. Since product quality problems are
generally handled badly in most models, we simply admit our inability to
make a substantive contribution to the study of product quality variation
in a general equilibrium context.
**
To avoid inefficiency and peculiarly shaped private output trans-
formation curves, we do not require that the marginal rate of technical
substitution between factors in public pollution abatement activities
equal the rate of substitution in the private sector. Thus, we assume
that public pollution abatement activities are efficient in a broader
sense than usual and thereby avoid theproblems created by the standard
efficiency rule when applied to the production of public goods. See
appendix C for a discussion of these matters.
22
-------
would generate a transformation surface similar to the UU'U'Um surface
depicted in Figure II.9.* m
This transformation surface is of interest for a variety of reasons.
At the most fundamental and obvious level, (1) UU'U'U shows the net out-
put/environmental quality combinations available in an economy restricted
(for some reason) to public pollution abatement activities as the only
means of controlling environmental quality. The implications of policy
restrictions are discussed in detail below. (2) The external diseconomies
generated by industry x negatively affect industry y output at all points
on the surface, including UUrUm where final output of x is zero. This is
the case because x is a material input employed in the production of y, and
therefore output of x, and hence pollution generated by the production of x,
is positive where final output X is zero. (3) Recalling out earlier dis-
cussion and definition of policy dominance, (2) implies that certain
(possibly all) sections of UU'UmU are dominated by at least one (and
possibly more) other pollution reduction methods. Consequently, only those
sections of UU'U'U that dominate the transformation surfaces associated
with all other pollution reduction methods (individually or in varying com-
binations) become sections on the transformation surface relevant for policy
selection, i.e., the envelope of all transformation surface sections asso-
cieated with absolutely dominant pollution reduction methods.
Except in rare circumstances, however, the Tinbergen-Theil theory
of economic policy (as extended by Brainard) would suggest that the envelope
transformation surface relevant for policy purposes would not include any
sections of the transformation surfaces associated with a single pollution
reduction method.** This conclusion follows as a direct application of the
two major lessons of the theory of economic policy. Under conditions of
certainty, irf general the attainment of n objectives retires n instruments
(pollution reduction methods here), and where more than h instruments are a
available, only n instruments need be selected. Under conditions of uncer-
tainty, however, these conclusions do not hold. Instead, the portfolio of
instruments with the lowest coefficient of variation would be selected, and
this portfolio would generally include the use of all available instruments.
These findings suggest that in an economy where X, Y, and Q (at a minimum)
matter in the selection among pollution reduction methods, single methods
would typically be inferior to combinations of methods. Therefore, we con-
clude that, in general, the envelope transformation surface relevant for
policy purposes may be expected to lie entirely above the transformation
*Although Figure II.9 is drawn on the assumption that public pollution
abatement activities, if undertaken on a sufficiently large scale without
changing aggregate factor proportions in the private sector, would raise
the state of environmental quality to its feasible maximum, such an assump-
tion may be factually incorrect. Certainly such an assumption would be
erroneous for certain other pollution reduction methods, and the transfor-
mation surfaces associated with those methods would approach coincidence
with the Q-axis below Qmax-
**
For an outstanding summary and extension of the Tinbergen-Theil
theory of economic policy, see William C. Brainard, Uncertainty and the
Effcftbiveneets of Policy, American Economic Rev tew, Vol. LXII, No. 2
(May 1967), pp. 411-425.
23
-------
u
m
FIGURE E. 9
FIGURE n.1O
24
-------
surface associated with any single pollution reduction method. Figure 11.10
illustrates this conclusion by showing that the envelope transformation sur-
face relevant for pollution control policy selection, lJU'UgUg, completely
dominates the public-pollution-abatement-activity transformation surface,
UU'UmUm (frot11 Figure II.9), except along UU1. *
Moreover, it follows that political, social, income and/or wealth
distribution, etc., constraints that reduce the set of technologically feasi-
ble policy options, and/or restrict the scale on which particular policies
may be undertaken, would necessarily eliminate some, though perhaps not all,
dominant policy combinations. For this reason, one would expect the polit-
ically, socially, or distributionally constrained transformation surface to
be partly, but rarely entirely, interior to the technically efficient envel-
ope transformation surface described previously, and net output and/or
environmental quality may be sacrificed to attain political, social or dis-
tributional objectives. These possibilities can be illustrated with the
aid of Figure 11.11, where UU'T'U^U is the "unconstrained" envelope transfor-
mation surface from Figure 11.10 and UU'T'U^Ut is the relevant transformation
surface following the elimination of pollution control method that was par-
ticularly effective on pollutants generated by industry y. Note that
X-Y-Q combinations on UTU'T'T lie on both transformation surfaces, while all
combinations on UTT'U^Ug dominate those on UTT'U£Ut. In this instance, by
assumption, we illustrate a situation in which some particular policy cons-
traint may or may not require a sacrifice of net output and/or environmental
quality. Of course, other policy constraints that would definitely neces-
sitate some sacrifice could be depicted in diagrams almost identical to
Figure 11.10.
Of course, the important conclusions to be drawn from this discussion
are rather obvious (though not trivial) and simply stated. Technical effi-
ciency in pollution control and the prevention of environmental degradation
requires a diversified portfolio of pollution control and abatement policies,
and constraints imposed on the selection of control and abatement policies
may require society to sacrifice both output and environmental quality to
achieve desirable political, social, or distributional objectives.
*
Since there are numerous different pollution control and abatement
methods, the relevant transformation surface, UU'U^Ug, :in Figure 11.10 is
drawn smooth and concave with respect to the origin. This seems reasonable,
but where externalities and public goods exist, it is certainly conceivable
that the relevant surface may be neither entirely smooth nor concave.
Second-order conditions often are not satisfied in the presence of external-
ities and public goods. For an insightful discussion of these matters, see
William J. Baumol, Welfare and the State Revisited, a new introduction to
Welfare Economics and the Theory of the State, second edition (Cambridge,
Mass.: Harvard University Press, 1965), pp. 3-8 and 32-36.
25
-------
Q
ro
T
FIGURE H. 11
-------
C. Consumption Activities and Environmental Quality
We commence the specification of the consumption side of our model by
defining the object of individual preferences to be "states of the world."
Individuals are postulated to have the ability to rank alternative states of
the world which differ in terms of (1) the consumption activities produced
and consumed by individuals themselves and (2) the state of the environment,
Q, experienced. In contrast to traditional microeconomic theory where indi-
viduals are assumed to have preferences defined in terras of the amounts of
goods they alone consume, we more realistically assume that individuals'
preferences extend beyond simply consumption activities to "things" which
they experience in some aggregate sense individually and collectively, though
not necessarily equally.
The building blocks of this specification are not original; they are
due entirely to the brilliant insights of Becker and Bradford.* However,
by specifying individual preferences to be an ordered function over alternative
states of the world characterized by a set of consumption activities and some
particular aggregate "thing" actually or potentially experienced by all indivi-
duals (e.g., air quality in the Willamette Valley), we have been able to extend
the literature on the economics of air quality in two important respects. First,
we have been able to derive testable hypotheses concerning which consumption
activities would be negatively affected by deteriorating air quality and thereby
avoid some of the intuitive empiricism that has characterized previous research.
Secondly, we have been able to develop a methodology within which benefit-cost
tests may be applied to evaluate alternative public policies to improve air
quality. The remainder of this section indicates how these results follow from
the above specification of individual preferences.
Put formally, we assume individuals have utility functions of the follow-
ing form:
(10) U » U(Zlf ..... Zm; Q) Q
where the Z's represent consumption activities and the utility function is con
tinuous and quasi-concave. In contrast to traditional neoclassical consumer
theory, the arguments in U are activities and the "state of the environment"
(in some well-defined sense). Consumption activities are regarded as produced
and consumed by individuals, and individuals by themselves control the scale
on which each activity is undertaken. With respect to the state of the
environment, however, individuals are viewed here as being unable to affect it
by individual action alone.
In this framework, the individual is postulated to combine market goods,
X and Y, and time in combinations that are technically defined to "produce"
The two major articles to which we refer are: Gary S. Becker, A Theory
of the Allocation of Time, Economic Journal (September 1965); and David F.
Bradford, Benefit-Cost Analysis and Demand Curves for Public Goods3 Kyklos
(September 1970).
27
-------
particular consumption activities. In other words, there is a production
function of the following sort for each activity
(11) Zj = F (Xj , Y ; T^ ..... Thj ) , j = 1, . . . ,m
where there are h different kinds of time which conceivably could be used to
"produce" the jth consumption activity. Following the lead of Becker and
Sorenson, we assume consumption technologies are linear and therefore rewrite
the general production functions as
(12) X. = a.Z.
Y. = b.Z.
J J J
n = 1, . . . ,h
where a., b., and t ; are fixed input-output coefficients.
Two features of this approach to consumer choice deserve special comment.
First, consumption- activity production functions permit classification of
activities according to their relative time and income intensities. This clas-
sification is a key element in Sorenson1 s derivation of testable hypotheses con-
cerning the effects of changing air quality on consumption activities and expen-
ditures. Second, and more obvious, demands for market goods, X and Y, are
viewed here as derived demands which originate with individuals' selection among
alternative potential consumption activities; and, therefore, demands for goods
shift in response to changes in variables not usually investigated in empirical
studies of consumer demand.
With matters of perspective and definition completed, we now turn to an
analysis of the implications of our view of consumer choice. Once more follow-
ing Sorenson, we postulate that in the selection of consumption activities in-
dividuals treat Q as a parameter and maximize their utility functions subject
to their respective time and income constraints.* With Q viewed as a para-
meter, preferences are appropriately stated over consumption-activity space
only. More precisely, we view the consumer's problem as a non-linear programming
problem, which can be represented as follows:
*
In contrast to our view that changes in environmental variables shift
tastes, it is appropriate to note that Becker has recently argued that the
impact of environmental variables on consumer behavior can be best studied by
introducing such variables into consumption activity production functions in-
stead of through tastes. Becker argues that "this method of handling environ-
mental variables is a powerful tool for greatly expanding the predictive con-
tent of economic theory." We disagree—at least in situations where the Becker
model is applied to the study of air quality variations and consumption activity.
The reason for our dissatisfaction with Becker's suggested is simple: neither
Becker nor we are able to derive explicit testable hypotheses concerning shifts
in consumers' production functions associated with air quality variations, whereas
28
-------
Maximize:
(13) U = U
Subject to:
Q
(14) EEt . Z. + E T = T
nj nJ J s s
Ec.Z. + E w T = G
j J J s s s
Z. >_ 0, T _> 0
where T = time worked in occupations,
"S
T = total time available,
c. = dollar cost per unit of market goods X and Y used in Z.,
w = the wage rate applicable in occupation s,
S
G = "other" nonwage income, and
!L. indicates that the state of the environment is a parameter
of the utility function relevant to the selection of optimal
amounts of consumption activities.
Sorenson has investigated this programming problem in considerable detail. In
particular, he has derived (1) the conditions that must be satisfied for an op-
timal solution (i.e., satisfaction-maximizing set of consumption activities) to
this problem to exist and (2) the impact of parameter changes on optimal activity
levels.
Unquestionably Sorenson's most important results are those obtained where
he assumes that variations in the state of the environment (e.g. air quality)
shift the utility function in a fashion such that
dQ
i.e., deterioration of the environment has negative or zero effects on the mar-
ginal utilities of activities. He concludes the discussion of this particular
parameter change by deriving the testable hypothesis that relatively "income-
intensive" consumption activities will be negatively Affected by deteriorating
environmental quality, where income- and time-intensiveness are defined in terms
of the relative sizes of dollar expenditures per unit of time in the various
consumption activities.
(continued from previous page) we have derived such hypotheses by postu-
lating air quality variations affect consumer tastes. Consequently, although
Becker's suggested method may prove useful in other contexts, we now doubt that
its usefulness for air quality research.
For Becker's views on these matters, see Garry S. Becker, Economic Theory
(New York: Alfred A. Knopf, Inc., 1971), pp. 47-48.
29
-------
The Sorenson hypothesis concerning air quality and consumption activity
suggests that air quality variations may well have considerably more extensive
effects on consumption activity than previous research has revealed by examin-
ing intuitively obvious hypotheses (e.g., studies of soiling, corrosion, clean-
ing and painting expenditures, and so on) and finding rather meager and often
uncertain effects. Although the Sorenson hypothesis proved useful for our
research on seasonal air quality deterioration from open field burning in the
Willamette Valley, perhaps the most useful test of the hypothesis would in-
volve an intensive cross-section study of consumption expenditures, by income
and social groups, across major U.S. metropolitan areas employing 1970 Census
and other data. Until the results of a broad-scale investigation that tests
well-specified hypotheses (e.g., the Sorenson hypothesis) concerning consumer
activity and air quality are known, the real impact of air quality variations
on consumer behavior must be regarded as only partially established by the
existing body of research.*
We next turn to the development of an approach to the valuation of
"the state of the environment" which permits (at least in principle) the
application of benefit-cost tests to the evaluation of alternative public
policies to improve air quality. We continue to postulate that the object
of individual preferences are states of the world which may be characterized
in terms of consumption activity and "the state of the environment."
For expositional simplicity, however, we now assume that there is only one
consumption activity, Z*. This assumption is merely a useful device for
simplifying the derivation of the "total compensated demand function"
which provides the foundation for our methodology of benefit measurement.
Hence, we now write individual utility functions as follows:
(15) U = U(Z*;Q) Q£
where Z* represents some (composite) index for private consumption activity,
Q is defined as previously, and the utility function is again continuous and
quasi-concave .
Our objective now is to show how information concerning individual
preferences may be used to obtain benefit measurements for policies designed
to change the state of the environment or a particular feature (say, air
quality) of it. Since our view of what constitutes benefit-cost analysis
perfectly coincides with that of Bradford:
*The position taken in the text should not be interpreted as suggesting
that most empirical investigations of "intuitively obvious" hypotheses
would not have been undertaken had the Sorenson hypothesis guided research
design. Quite the contrary, many of the "obvious" hypotheses are consistent
with the Sorenson hypothesis. For example, since housing expenditures are
clearly income-intensive, research on housing decisions, land values, etc.,
and variations in air quality would rank high on a research agenda determined
strictly by the Sorenson hypothesis.
30
-------
Benefit-cost analysis can be...described as a technique
to discover whether a proposed change in one variable
of the state description can be feasibly obtained
by changes in other variables such that the new
state is preferred by some citizens, the existing
state preferred by none. More simply but less precisely
put, it is a procedure for testing the Pareto optimality
of the existing state. The general approach is to
seek to determine the maximum amount citizens would
be willing to pay for some change, and the minimum
. amount that must be given up to obtain it. If the
former, the aggregate benefits, exceeds the latter,
the cost of the change, it is the usual view that the
change should probably be undertaken.*
we now suggest a method for measuring the "maximum amount citizens would be
willing to pay for some change." Before doing this, however, it is appropriate
to state explicitly that neither Bradford nor we claim that a benefit-cost
test of the sort described here is sufficient by itself to determine
whether a specified change "should" be undertaken. We agree with Bradford
that answers to this and similar normative questions can only be determined by
introducing a well-defined social welfare function and a specification of
the available alternatives.**
In keeping with traditional consumer behavior theory, we assume that
individuals have the ability to rank all alternative states of the world in
order of preference. It is further assumed that individuals are indifferent
between states identical in terms of their own consumption activity and the
state of the environment, even if the distribution of consumption activity
among other individuals differs in these states. Without this latter assumption,
benefit-cost analysis of the standard sort is quite impossible, particularly
for large changes in the state of the world. The reason for this is clear.
Each individual's willingness to pay for some change would itself change as
other individuals' contributions varied, and therefore benefit estimates
derived from the market or pseudo-market data normally employed in benefit-
cost analysis would be quite inadequate to characterize the range of aggregate
willingness-to-pay figures that are likely to obtain.
*David F. Bradford, op.cit., pp. 787.
**For compelling arguments in support of this position, see David F.
Bradford, Constraints on Public Action and Rules for Social Decision,
American Economic Review (September 1970).
31
-------
Given our assumptions, Figure II. 12 depicts the indifference map
of what we presume to be a typical individual. Each indifference curve
indicates alternative equally-preferred states of the world, and successively
higher curves indicate increasingly preferred states. Note that the indifference
map extends only over the conceivably feasible state-of-the-world space.
If the initial state of the world was such that our typical individual
found himself at point A in Figure II. 12, what would be the maximum amount
of consumption activity that the individual would be willing to forego for
various changes in the state of the environment? The answer to this question
is provided in Figure II. 13, where the curve AOA' is indifference curve
1-2 drawn with new axes to establish a new origin, 0'. Given AO"A', the
new horizontal axis AZ* permits measurement of the maximum amount that the
individual would be willing to forego to secure different environmental states,
given his initial state at point A in Figure II. 12. Since the slope of
AO'A' is the marginal rate of substitution of Z* for Q along !£, AO'A" is
the "total" compensated demand curve associated with the usual compensated
demand curve. Of course, an aggregate "total" compensated demand curve
for different states of the environment would be constructed by the
"vertical" addition of all individual "total" compensated demand curves.
Since the properties of such aggregate "total" compensated demand
functions are unusual, they are worthy of further discussion. First, by
construction, at any point on the aggregate curve all individuals remain on
their initial-state indifference curves; no one's welfare has either increased
or decreased. Second, in contrast to the situations described by the usual
compensated or unccmpensated demand function for a private good, individuals
would not generally have identical marginal rates of substitution between
the numeraire (here, Z*) and the state of the environment at each point along
the aggregate "total" compensated demand function. We do not presume that
individuals have identical tastes for environmental quality. Third, by
construction, the slope at any point of an aggregate "total" compensated
demand function is the sum of individual marginal rates of substitution
given that every individual remains on their initial-state indifference
curve. Consequently, an aggregate "total" compensated demand function
provides a measure of both total and marginal benefits of each environmental
state (given the initial state) in precisely the fashion that an ideal benefit-
cost analysis would require.
On the basis of the preceding analysis, we conclude that there is
little conceptual difficulty in the application of benefit-cost analytical
techniques to the evaluation of projects and policies designed to change the
state of the environment. On the other hand, however, there are obviously
formidable estimation and measurement questions that must be addressed in
any particular application. Some of these matters are discussed in the
following section where the production and consumption sides of our suggested
framework are brought together.
32
-------
Q
QMAX
O
r
FIGURE H12
OMAX
-AZ
FIGURE H. 13
33
-------
D* Benefit-Cost Analysis Of Environmental Quality
Benefit-cost analysis has been described earlier "as a technique to
discover when a proposed change in one variable of the state description can
be feasibly obtained by changes in other variables such that the new state
is preferred by some citizens, the existing state preferred by none."
To put this view slightly differently, benefit-cost analysis is a branch of
applied welfare economics founded on the potential Pareto improvement principle.
A potential Pareto improvement may be defined as an economic rearrangement
in which the gainars could more than compensate the losers, assuming re-
distributions to be costless. However, the fact that some specified economic
change passes a benefit-cost test is not sufficient to determine whether
the change should be made. In general, where one discovers one potential
Pareto improvement, it is rarely difficult to find others as well, and selection
among alternative feasible potential Pareto improvements requires introduction
of a social welfare function. Consequently, at best benefit-cost calculations
provide useful, though hardly conclusive, comparative information for assessing
the desirability of alternative feasible projects, programs, or policies.
As the preceding paragraph indicates, we claim less for benefit-cost
analysis than is often the case. Nevertheless, we believe that assessment of
alternative public policies to change the state of the environment (a public
good) by application of the potential-Pareto-improvement criterion is feasible,
both on the conceptual level and in practice. In this section, we begin our
demonstration of how benefit-cost analysis can usefully be applied where
public and private goods matter to individuals. We do this by bringing
together the production and consumption sides of our framework to specify
the adequacy and comprehensiveness of benefit-cost tests of alternative
pollution control methods and policies. Following completion of this task,
we report the application of the conceptual approach developed here to assess
the benefits and costs of various public policies to control pollution from
open field burning in the Willamette Valley.
To determine whether potential-Pareto-improvements exist (i.e.,
whether any feasible policies have benefit-cost ratios greater than unity),
the analyst must confront the aggregate "total" compensated demand function
with the envelope transformation surface relevant for pollution control
policy selection. This is accomplished here with the aid of Figure II. 14,
where in panel (a) the relevant transformation surface, UU'U^Ujj, is
superimposed on two alternative representations of the aggregate derived
"total" compensated demand surface associated with the assumed initial state
at point G.* Panel (b) of Figure II.1-4 provides a cross-section view of panel (a),
*Since our analytical results would apparently be more easily understood
if obtained in final-goods-environmental-quality space than in consumption-
activity-environmental-quality space, we henceforth refer to aggregate "total"
compensated demand functions or surfaces as relevant derived demand functions
or surfaces. Although this terminology is intended to remind the reader that
we regard consumption activities, not final goods, as arguments in individual
utility functions, it is important to note that the transition from activity-
34
-------
u
u.
V
X
FIGURE E. 14 (a)
MAX
H
FIGURE E 14(b)
D
35
-------
where G serves as the origin for the superimposed axes for two cross-sections,
GH, GH" and GK'KJ, of the two relevant derived demand surfaces and the
relevant transformation surface, respectively.
Based on earlier discussion and appendix A, we postulate that at G
the economy is in full competitive equilibrium with no pollution control
policies. At G, final good outputs are OE units of Y and EG" units of X,
relative final good prices are indicated by the slope of PP" which is
tangent to FF" at G'% and the state of the environment is G"G on an index
with OQjjjax as its maximum value. (Information concerning the income distri-
bution at G car* be derived in the fashion described in appendix A.)
With these; preliminary matters completed, we now indicate how benefit-
cost analysis may contribute to the selection among potential pollution
control policies. If the relevant derived demand surface associated with
the initial state, G, is known to be tangent to the plane ABC in panel (a)
of Figure II. 14, the initial state is technologically efficient (because
the economy is operating on its transformation surface) but potential Pareto
improvements are available for exploitation. Suppose the relevant derived
demand surface at G is known (1) to be convex with respect to the origin
and (2) to trace the elipze GMK as it passes through the relevant transformation
surface, UU^U'l^. In fact, all pollution control policies (individually or
in combination) that would place the economy on or in the space contained
between the relevant demand and transformation surfaces would result in
potential Pareto improvements in the initial state, and all policies falling
within this set would have benefit-cost ratios greater than or equal to unity.
The preceding conclusion and its relation to standard benefit-cost
analysis can be more readily appreciated with the aid of panel (b) to Figure
II. 14. With a given final output of OE units of Y, all policies that
would move the eccnomy to points on or within the space between the cross
sections of the relevant derived demand and transformation curves, GH'
and GJ, respectively, would have measured benefits equal to or exceeding their
costs. For example, adoption of a set of policies placing the
economy at K' would have measured benefits, Gb, measured costs, Ga, and a
benefit-cost ratio, Gb/Ga, clearly exceeding unity. In contrast, the set of
policies moving the economy to K would only have a benefit-cost ratio,
Gc/Gc, equal to one. Consequently, all technologically efficient policies
constrained by the relevant derived demand surface must necessarily have
benefit-cost ratios which exceed or equal one.
*(continued from previous page) to goods-space means that the assumed
quasi-convexity of individual utility functions is not sufficient to establish
the convexity of the relevant derived demand surface. Fortunately,
however, none of our principal analytical results depend on the convexity of
the relevant derived demand surface.
36
-------
Unfortunately, however, the preceding result for technologically
efficient policies appears stronger than in fact it is. To refer to panel
(b) of Figure II. 14 once again, all policies moving the economy to positions
within the space constrained by the relevant derived demand and transformation
curves would also have benefit-cost ratios exceeding unity. For this reason,
benefit-cost ratios greater than one must not be viewed as synonymous with
technologically efficient projects or policies, and the standard benefit-
cost study should always be supplemented with cost-effectiveness investigations
designed to assure selection among technically efficient (i.e., dominant)
policies.
Similar cautionary remarks are also appropriate here concerning
the impropriety of selecting the technologically efficient policy with the
highest benefit-cost ratio (or alternately, the greatest net benefits).
To apply these criteria would mean the selection of policy by maximizing some
index of the potential gains from trading the existing state for some new
state. However, since benefits are defined as the maximum amount individuals
are willing to pay for some specified change in the existing state, while
costs are defined as the minimum amount that must be foregone to obtain the
change, the potential gains reflect individuals own valuations of alternative
states given the existing distribution of income and wealth. Consequently,
selection of the policy with the highest benefit-cost ratio, or greatest net
benefits, is appropriate only if the social welfare function has relative
weights identical to individuals' relative valuations in the existing
non-optimal state. If the social welfare function has any other set of
relative weights, welfare maximization is incompatible with selection of
the policy having the highest benefit-cost ratio, or greatest net benefits,
as these are measured in the standard benefit-cost analysis.
Although the preceding remarks emphasize that benefit-cost analysis
by itself cannot determine the "best" project or policy for society to
undertake, they are not intended to minimize the potential usefulness of
standard benefit-cost analysis. On the contrary, they are intended only to
prevent possible misuse of benefit-cost evaluations by indicating precisely
what a benefit-cost ratio greater than one does or does not reveal. Moreover,
perhaps the most useful feature of benefit-cost evaluations of alternative
policies is their ability to identify policies which would not be Pareto
improvements over the existing situation, thereby assisting the specification
of the set of alternative policies which are properly investigated in great
detail to determine the policy subset that would maximize social welfare.
Thus, negative results often indicate the direction that policy should
change to achieve genuine gains for society.
This position can be illustrated with the aid of Figure II. 14,
panel (a). There, as we have previously observed, if the relevant derived
demand surface associated with the initial state at G traces the elipse
GMK as it passes through the relevant transformation surface VW^\3Z,
all technologically efficient policies that would move the economy from G
to points outside the elipse on UU'U^Ug, would have benefit-cost ratios less
than one. A similar, though perhaps less useful, partition of the set of
technologically efficient policies can also be achieved by benefit-cost
37
-------
analysis without full knowledge oi the relevant derived demand function.
Where the analyst (a) knows individuals' valuations in the initial state and
(b) makes "reasonable" assumptions concerning the properties of the unknown
demand function, benefit-cost calculations can be undertaken and the policy
set partitioned an described above. For example, suppose the analyst
knows that the relevant derived demand function is tangent at G to the plane
ABC in Figure II. 14, panel (a), and assumes that the valuations implicit in
the slopes of the plane ABC at G are "reasonable" for calculation of the
benefits of alternative policies. By this procedure, policies associated
with area GU'G' on the relevant transformation surface would be found to have
benefit-cost ratios less than one, whereas policies associated with area
UGG'U^UZ would have benefit-cost ratios greater than one. Therefore even
without full knowledge, reasonable applications of benefit-cost methodology
would appear to generate useful results for social decision-makers.
Finally, we should note that if the initial state is itself Pareto
optimal, the relevant derived demand surface never lies below the relevant
transformation surface.* In such circumstances, all policies changing the
initial state would have benefit-cost ratios less than one. For this reason,
benefit-cost analysis is properly regarded as a means for testing the Pareto
optimality of the existing (or some initial) state. Of course, where Pareto
optimality obtains, all of the usual necessary marginal conditions also
hold; for example, the sum of individual marginal rates of substitution of
private goods for environmental quality would equal the marginal cost of
environmental quality improvement.
*It should be recalled that the relevant demand curve always refers
to changes in some particular initial state and would be different at each
conceivable state. Furthermore, it should be noted that the proposition
stated in the text does not rule out either corner solutions or multiple
Pareto optimal situations. Such situations cannot be excluded as possibilities
because the relevant derived demand function is not necessarily convex.
38
-------
III. THE COSTS OF ALTERNATIVE OPEN
FIELD BURNING CONTROL POLICIES
A. Introduction
In contrast to the majority of studies concerned with air pollution
control costs, the investigation reported here must focus on the effects of
control regulations on a major segment of a competitive industry. To do this,
we employ statistically estimated supply and demand functions for the major
types of grass seed raised in the Willamette Valley to predict how prices,
outputs, etc. would change following the introduction of, and industry adjust-
ment to, alternative field burning control policies. These results are then
used to calculate the policy-induced changes in consumers'1 surpluses, rents
to specialized factors, etc. in the grass seed industry. Although not all of
the changes we predict and report here are costs in the sense that they mea-
sure the value of goods and services foregone by society as a result of Oregon's
open field burning control policies, each change does measure the impact of
policies on groups potentially important in public policy formation.
Although our methodological approach is quite standard in most re-
spects, our basic model, measurement techniques, and level of disaggregation
distinguish the estimates reported here from those made in other studies with
similar methodological approaches. To be specific, our adoption of standard
partial-equilibrium methodology has not precluded prediction and measurement
of the very different effects of Oregon control policies on seed producers
located inside and outside Oregon or owners of Willamette Valley land present-
ly in seed production. Furthermore, for what we believe is the first time, we
do not measure changes in producers well-being by changes in the areas above
the statistical supply curves employed to predict post-control prices and out-
puts. This separation of prediction and measurement means that our estimates
of changes in producers well-being are probably more realistic than those
sometimes reported in other studies.
The next section presents the analytical framework and measurement
techniques employed here to estimate the costs associated with three alterna-
tive policies regulating open field burning in Oregon. Section C presents our
actual estimates of the changes in consumers' surpluses and producers' rents
associated with the three policies, while section D translates the predicted
reductions in Oregon rents to decreases in the value of Willamette Valley
agricultural land presently in grass seed production.
B. An Analytical Framework for Measuring the Costs of Regulating Open Field
Burning
The concepts of consumers' and producers' surpluses have often been
employed to evaluate the effects of resource misallocation, gains and losses
from international trade, price instability, and public and private investments.*
For an excellent recent survey of the literature on consumers' and
producers' surplus, see J. M. Currie, J. A. Murphy, and A. Schmitz, The Concept
of Economic Surplus and its Use in Eoonomie Analysis3 Economic Journal^, Vol. 81
(December 1971), pp. 741-799.
39
-------
The basic model used to evaluate the welfare effects of investments is depicted
in Figure III.15 which is identical to the figure that Schmitz and Seckler pre-
sent to explain their method of computing the rate of return on investment in
the development of the mechanized tomato harvester.* We repeat the Schmitz-
Seckler figure for two reasons: (1) it conveniently initiates illustration of
the basic concepts used here to measure some effects of regulating open field
burning int he Willamette Valley; and (2) it also illustrates why "obvious"
procedures for measuring producers' surplus may result in unrealistic estimates
clearly inconsistent with basic microeconomic theory. Following this discus-
sion of definitional and methodological matters, we specify the particular
model and estimation procedures employed in this report.
Following Schmitz and Seckler, suppose that some technological improve-
ment reduces,production costs, thereby shifting the supply curve in Figure III.l
from S0 to So following adoption of the technique throughout the industry. The
gain in consumers' surplus is measured as the area E + G + F, while the gain in
producers' surplus is H + I - E. However, if the initial supply curve had been
perfectly elastic, the only gain to society would be the gain in consumers'
surplus E + G + F since producers' surplus does not exist. Schmitz and Seckler
define the sum of producers' and consumers' surpluses to be the gross social
gain from the postulated technological change. They also go one step further
than previous researchers and calculate the net social gain to society. They
define the net social gain to be the gross social gain minus the aggregate wages
that would have been paid to workers displaced by the innovation.
Although each of these measures has been the subject of much criticism,
they do provide rough estimates of magnitudes that are properly regarded as im-
portant in the appraisal of major public and private investments. In any case,
"while it is easy to raise objections to the use of the concept of economic
surplus for providing answers for policy formulation, it is difficult to find
any workable alternatives."**
A. Schmitz and D. Seckler, Mechanized Agriculture and Social Welfare:
The Case of the Tomato Harvester, American Journal of Agricultural Economics,
Vol. 52 (November 1970), pp. 569-580.
**
J. M. Currie, et^al^, op. cit., p. 791. The first section of this
article (pp. 742-765) provides a careful, balanced exposition of the major
theoretical contributions to the development of the concepts of consumers' and
producers' surplus, including full treatment of the objections to the concepts.
40
-------
GAIN IN CONSUMER'S SURPLUS = E + G + F
GAIN IN PRODUCER'S SURPLUS = H * I + E
FIGURE nr.1
41
-------
As indicated by the definition given above, the traditional measure
of producers' surplus is the area above the supply curve and below the price
line. Consequently, what this area actually measures depends on the nature
of the supply curve postulated. By addressing ourselves to this question to
determine how to define producers' surplus in the Oregon grass seed industry,
we discovered a significant error was made in the construction of Figure III.l
by Schmitz and Seckler. (We had previously uncovered the same type of error
had been made in an empirical study of externalities in the farm use of pesti-
cides by Edwards.)* In Figure III.l, both So and So intersect the horizontal
axis, thereby implying that producers' surplus at low levels of output equals
price. Although there has been considerable controversy what supply curve to
employ to measure producers' surplus, economic theorists as a group would
generally agree that short- and long-run marginal production costs, including
or excluding rencs, are zero only in very rare circumstances. Therefore,
Figure III.l must be regarded as constructed in a theoretically unjustifiable
fashion.
This finding led us to reconsider previous definitions and measurements
of producer's surplus. With respect to measurement of producers' surplus, we
conclude that econometric supply functions are unrealistic tools to employ for
two very practical reasons. First, econometrically estimated supply functions
often have either positive horizontal axis intercepts implying zero marginal
costs at low output levels or_ positive vertical-axis intercepts substantially
different reasonable estimates of long-run marginal costs excluding all quasi-
rents to variable factors. Second, even if vertical-axis intercepts were
forced to approximate marginal costs, there is no necessary correspondence be-
tween the slope and degree of curvature of the estimated and desired supply
function. Moreover, with respect to the desired supply function itself, we
found ultimately that we agreed with Mishan that (1) the area above the supply
curve is comparable to consumers' surplus only under special circumstances in
the short-run and (2) the concept of economic rent is "perfectly symmetric"
with the concept of consumers' surplus.** Therefore, to avoid confusion and to
follow Mishan's lead, we henceforth dispense with the term producers' surplus
and instead refer to "aggregate industry rents."
*
See W. F. Edwards, Economic Externalities in the Agricultural Use
of Pesticides and an 'Evaluation of Alternative Policies, unpublished Ph.D.
Thesis, University of Florida, 1969. Edwards employs linear supply-response
functions derived from his econometric study of vegetable and fruit supply
functions in Bade County, Florida to calculate producers' surplus; some of his
supply functions have positive horizontal-axis intercepts. Edwards makes his
error without constructing a theoretically unrealistic figure, however.
In contrast to Edwards, Schmitz and Seckler construct a theoretically
unrealistic figure to define producers' surplus, but they do not calculate a
measure of producers' surplus in their article.
**
E. J. Mishan, What is Producer's Surplus, American Economic Review,
Vol. 58, No. 5 (December 1968), pp. 1269-82, especially pp. 1271-73 and 1278-79,
42
-------
Aggregate industry rents are defined here as aggregate industry
returns to those factors of production typically regarded as perfectly or
highly inelastic in supply to the industry in the short-run. Industry aggre-
gate rent per unit of output equals output price minus median or mean average
variable costs in the industry, and industry aggregate rents are measured as
the product of industry aggregate rent per unit of output and industry output.
The word aggregate appears in our definition because we attempt to measure the
sum of returns to all commonly regarded fixed and/or highly specialized factors
of production. Here, changes in aggregate industry rents replace changes in
producers' surplus as an empirical measure of the welfare (here, income) changes
experienced by owners of specialized factors.
Of course, in absolute terms it is clear that this procedure does not
provide an accurate measure of rents as they are traditionally defined. There
are at least two reasons for this. First, quasi-rents received by variable
factors are excluded from our measure. Second, the non-pecuniary advantages
of particular employments and occupations would only rarely equal the total pay-
ment received by the owners of land, specialized facilities, equipment, entre-
prenuerial ability, etc. In any case, although we are disappointed to work with
an absolutely inaccurate measure of aggregate industry rents, we do not regard
the inaccuracy as fatal given our objective of obtaining a reasonable measure of
the change in rents consequent to the introduction of various pollution control
policies.
Traditional partial equilibrium reasoning underlies the preceding con-
clusion. If some subset of the normally regarded "fixed" factors are truly
specialized to a pollution-policy affected industry, then any decline in aggre-
gate industry rents as defined here would necessarily be entirely borne by
those specialized factors. Moreover, if some "fixed" factors do leave the policy-
affected industry, those which remain because they lack employment opportunities
elsewhere must also bear the entire change in aggregate industry rents as de-
fined here. Consequently, an ability to measure changes in aggregate industry
rents as defined here is equivalent to an ability to measure the change in re-
turns to an industry's "specialized" factors.
In the next section, we present alternative estimates at farm level of
the losses in consumers' surplus and rents to Oregon seed producers, as well as
the gain in rents to non-Oregon seed producers, that would result following im-
plementation of, and industry adjustment to, three alternative policies regulat-
ing open burning of Oregon seed fields. The basic model used to make these
estimates is illustrated in Figure III.2; the table below the figure indicates
the areas measured to compute the losses and gains experienced by the three
groups affected by the policies investigated. As the table shows, alternative
assumptions concerning relevant supply elasticities greatly influence the size
of gains and losses experienced by the affected parties,, In Figure III.2, Oc
measures mean average variable costs for Oregon producers prior to the introduc-
tion of open field burning regulations, while Oc'measures mean average variable
costs after introduction of field burning regulations. As the areas in the
table imply, OP, OQO, and OQj—, represent the pre-regulation price, Oregon output,
and non-Oregon output respectively.
The final section in this part of the report translates the rent losses
to Oregon seed producers into decreases in the value of Willamette Valley agri-
cultural land presently in grass seed production. The basic model used to obtain
43
-------
FIGURE m 2
zs'
o
°k> NO o o
QuQu
Gain or Loss, Alternative Non-Oregon
by Oregon Supply Situation—Supply Situation
Loss in Consumer's Surplus,
Loss in Oregon Rents,
Gain in Non-Oregon Rents,
PabP'
O
Pcdf-Pc'jg
cdec'
PkIP'
O
O
O
Pcdf-Pcih
cdecx
O
O
44
-------
these losses may be illustrated with the aid of Figure III. 3. This model
postulates that policies restricting open field burning would result in a left-
ward shift in the demand for land to raise grass seeds from Ds to Ds and lower
the price of seed land at the margin from OP to OP1.
Since the marginal land presently in seed production has other poten-
tial agricultural and/or non-agricultural uses, we do not postulate a model in
which the supply of land for raising grass seeds is perfectly inelastic. To
emphasize this point, panel (b) of Figure III. 3 was consturcted to represent
the demand curves in panel (a) to demonstrate that demand curve for land in
other uses than seed production, DQJJ, may also be interpreted as the supply
curve of seed land.* For this reason, the demands for land in other uses have
an important influence on the impact of open field burning control policies on
agricultural land values. Explicit attention is given these matters below.
C. Changes in Consumers' Surpluses and Producers' Rents Under Alternative
Open Field Burning Control Policies
This section reports and discusses our estimates of the absolute and
relative changes in consumers' surpluses and producers' (Oregon and non-Oregon)
rents associated with three alternative open field burning control policies.
The effects of each policy are estimated assuming three rather different supply
situations in order to indicate the full range of possible outcomes. This pro-
cedure seemed advisable because Oregon dominates U.S. production of certain seed
crops, and information concerning non-Oregon grass-seed supply functions is very
limited. Referring to Figure III. 2 once again, the three supply situations are:
I. Normal Supply Situation: positively-sloped supply curves,
so and SNO;
II. Polar Supply Situation (1): perfectly inelastic Oregon supply
curve, SO , reflecting limited alternatives available to Willamette
Valley producers, and S^Q; and
III. Polar Supply Situation (2): perfectly elastic non-Oregon supply
curve, S?L, and normal Oregon supply curve, SO (This situation
is unrealistic but allows to establish the maximum potential loss
in Oregon rents.)
A
Panel (b) of Figure III.3 is constructed in a fashion similar to the
one used by Stigler to analyze the price determination of storeable goods with
fixed supplies. See George J. Stigler, The Theory of Prioe, 3rd ed. (New York:
The Macmillan Company, 1966), pp. 96-98.
Of course, the reduction in air pollution associated with the control
policies may increase demands for land in other uses because the attractiveness
of the Valley to potential residents and foot-loose firms increases. Although
we recognize this possibility, we doubt that it is important in the present case.
45
-------
FIGURE EL. 3
panel (a)
*LAND
FIGURE HI. 3
panel (b)
>LAND
46
-------
The three field burning control policies investigated here are:
A. A complete ban on open burning;
B. Open burning permitted once in three years; and
C. Alternate year burning.
These policies would reduce total acres burned per season by 100, 67, and 50
percent, respectively, and each would improve visibility in the Willamette
Valley to a different degree.
Each of the above policies, however, would increase grass seed produc-
tion costs regardless of how individual operators adjust their cultivation,
harvest residue removal, and disease control practices.* Tables 1.3 and 1.4
supply substantial evidence that partial or complete elimination of open field
burning would increase total costs per acre and/or simultaneously reduce yields
per acre. Moreover, in the absence of burning there would probably be increases
in the incidence of seed diseases, decreases in seed purity, and increases in
seed cleaning costs. These last problems would not exist if mobile field sani-
tizers were used to replace open field burning, entirely under policy A and
partly under policies B and C. Fortunately, it now (June 1972) appears that
an effective mobile field sanitizer may be able to operate at a cost as low as
$5 per acre, though $9 per acre is regarded as today's best estimate of its
cost. Since controlled burning with the sanitizer appears to produce the same
beneficial residue disposal and cultivation effects as open burning without pro-
ducing smoke which reduces visibility, the sanitizer must now be viewed as the
low-cost alternative to open field burning. Therefore, we estimate the changes
in consumers' surpluses and producers' rents under our three alternative control
policies assuming Oregon producers adopt the mobile field sanitizer and incur
costs equal alternately to $5, 9 and 13 per acre in the years they do not burn.**
*Here we make the usual assumption that overall efficiency (or, for that
matter, inefficiency) in the Oregon grass seed industry is not affected by the
"pressure" that various pollution control policies would place on the industry.
Since such policies would undoubtedly induce rationalization of industry organiza-
tion, our estimates of declines in producers' rents would be expected to exceed
observed losses.
**
We employ these cost figures at the suggestion of Professor Frank S.
Conklin, who has expended considerable effort attempting to develop realistic
cost estimates for the sanitizer.
We should also note that our estimates assume that the sanitizer has
n£ effect on yields per acre. Professor David Chilcote informs us there is
limited evidence that the sanitizer may conceivably increase yields by producing
more uniform burning than can be accomplished with the open burning technique.
If further research establishes this to be true, our estimates of declines in
producers' rents would necessarily exceed actual losses.
47
-------
Appendix F reports the statistical investigation of grass seed demand
and supply functions that underlie our predictions of the effects of alternative
field burning control policies. There we derive the linear supply and demand
equations employed to compute the estimates of policy-induced changes presented
in Tables III.l and III.2. Those equations, and hence our estimates of surplus
and rent changes, are based on 1965-69 average prices and quantities at the
farm level. Therefore, our surplus and rent estimates attempt to measure what
changes would have occurred if the grass seed industry had adjusted to the open
field burning control policies in the late sixties.*
The estimates presented in Tables III.l and III.2 suggest the following
observations. (1) Except in the case of red fescue, the declines in Oregon rents
under each policy are approximately the same in supply situations I and II. This
result is important because supply situation I assumes "normal" reactions by
Oregon producers to the policy-induced cost increases, whereas supply situation
II assumed that Oregon producers are "boxed in" to seed production because of the
special climactic, soil, and market conditions they face. Interestingly, the
resulting losses for the industry appear to be about the same in either situation.
(2) On the other hand, the declines in Oregon producer rents in supply situations
I and II are only 50-75 percent of the losses we estimate they would suffer in
supply situation III. Consequently, to the extent that non-Oregon supplies are
substantially more responsive to price than we assumed in supply situation I, the
decline in Oregon rents could potentially exceed our "most reasonable" estimates
by perhaps 20 to 50 percent.**
(3) Except in the case of tall fescue, the absolute decrease in con-
sumers' surplus is less than the decrease in Oregon producers' aggregate rents
in supply situation I. This difference is most dramatic in the case of ryegrass.
By the same token, it should be immediately noted that the relative decrease in
consumers' surpluses for ryegrass is slightly greater than the relative in Oregon
producer rents. For all other seeds, however, the relative decrease in consumers'
surpluses is also smaller than the relative decrease in Oregon rents. (4) Of
course, the principal reason for these results is that non-Oregon producers would
expand their outputs as Oregon producers react to introduction of the more costly
production methods associated with alternative open field burning control poli-
cies. It is interesting to observe in Table III.2 that the relative increase in
non-Oregon producer rents exceeds the relative decrease in Oregon rents for tall
fescue, red fescue, chewings fescue, and bentgrass.
(5) Finally, we note that the assumed linearity in our demand and supply
funtions does result in absolute or relative changes in consumers' surpluses and
producers' rents that are proportional to the postulated changes in costs. Al-
though this result is not particularly unusual, it does indicate the potential
errors associated with common extrapolative procedures.
Tables III.l and III.2 report only changes in consumers' surpluses and
producers' rents. See Appendix G for tables which report predicted prices, out-
puts, average variable costs, consumers' surpluses, and producers' rents by seed
type, policy, supply situation, and cost per acre for the mobile field sanitizer
under each policy for the three postulated supply situations.
As previously indicated we regard supply situation III as unrealistic.
Hence, in the text we suggest potential losses might be realistically estimated to
be 20 to 50 percent larger than our "most reasonable" estimates rather than 30 to
100 percent larger as implied by our supply situation III estimates.
48
-------
PAGE NOT
AVAILABLE
DIGITALLY
-------
PAGE NOT
AVAILABLE
DIGITALLY
-------
D. Changes in Agricultural Land Values under Alternative Open Field Burning
Control Policies
We conclude this portion of the report by presenting our estimates of
the decrease in the value of Willamette Valley agricultural land under alternative
open field burning control policies. These estimates are based on data and sta-
tistical results reported in the previous section and appendices E and G. Two
assumptions underlie these estimates: (1) the control policies under investigation
here have no impact on the marginal physical products of land, specialized equip-
ment, special entreprenuerial abilities, etc.; and (2) the policies do not signi-
ficantly alter the contribution of urban influences to Willamette Valley agricul-
tural land values. The first assumption seems reasonable because use of the
mobile field sanitizer would require no fundamental changes in cultivation or pro-
duction practices and, in addition, is postulated here to have no impact on
yields per acre. Hence, the sanitizer would raise costs per unit of output but
not affect marginal products. The second assumption is based on our judgment
that demands for land in alternative uses to seed production would not shift
because the attractiveness of the Valley increases as open field burning is con-
trolled. Consequently, the already existing and projected urban influences on
agricultural land values may reasonably be assumed unchanged by adoption of the
control policies under study here.
Under the assumptions specified above, the percentage decrease in the
difference between price and average variable cost would equal the percentage
decrease expected in rents received by land and other commonly regarded "fixed"
factors in grass seed production. In Table III.3, column (1) presents this
percentage decrease for supply situation II, where all "fixed" factors remain
in seed production after open field burning controls are introduced, and column
(2) presents the percentage decrease for supply situation I, where some "fixed"
factors leave seed production to receive payments at least equal to those they
could receive in seed production. Of course, if current and expected future
earnings in agricultural use were highly correlated, and if current land values
approximately equalled the capitalized value of current earnings in perpetuity,
then these percentage decreases would respectively measure the maximum and
minimum percentage decreases expected in agricultural land values consequent to
each open field burning control policy. Appendix E, however, provides conclu-
sive evidence that neither of these conditions are satisfied in the Willamette
Valley, where distance-to-town is an exceedingly important independent variable
in statistically explaining variations in agricultural land values.
Given that agricultural land values per acre substantially exceed the
capitalized value of estimated rents per acre at interest rates of 8, 9, 10 or
more percent, land values would be anticipated to decline less than the percent-
ages reported in columns (1) and (2) in Table III.3. To predict by how much
land values would decline in the Willamette Valley generally following implemen-
tation of each open field burning control policy, we employ the all-counties •
regression results reported in Table III.4 of appendix E. Since that regression
equation was estimated in log-log form, we obtain columns (3) and (4) of Table
III.3 by multiplying the percentages given in the first two columns by 0.6333,
the regression coefficient for estimated rent per acre (which also served as a
soil quality index in appendix E). The final column of Table III.3 simply aver-
ages our maximum and minimum predicted declines in land values to provide our
"most reasonable" estimate of the declines that would be associated with differ-
ent control policies under different assumption concerning the cost per acre of
the mobile field sanitizer.
53
-------
Table III.3.
Predicted Percentage Decreases in Willamette Valley
Agricultural Land Values Under Alternative Open
Field Burning Control Policies, By Policy and Under
Various Assumptions
Ul
-P-
Mobile
Decrease in Price minus
Average Variable Cost
Predicted Decrease in Agricultural
Land Values
Sanitizer
Cost
Acre
Policy
Ban on Burning
Once in Three Years Burning
Alternate Year Burning
per
$ 5
9
13
5
9
13
5
9
13
Supply
Situation II
CD
7.35%
13.24
19.12
4.90
8.82
12.75
3.68
6.62
9.56
Supply
Situation I
(2)
4.48%
6.61
9.91
2.61
4.62
6.69
1.92
3.47
5.01
High
Estimate3
(3)
4.65%
8.38
12.11
3.10
5.86
8.07
2.33
4.19
6.05
Low
Estimateb
(4)
2.83%
4.19
6.28
1.65
2.93
4.23
1.22
2.20
3.17
Average
Estimate0
(5)
3.74%
6.29
9.20
2.48
4.35
6.15
1.88
3.20
4.61
Source: Tables G.1-G.6
Col. (1) x 0.633
Col. (2) x 0.633
[Col.(3) + Col.(4)]/2
-------
At first blush, this procedure probably appears to "over-interpret"
our regression results. After all, those regression results purport only to
explain variations about mean land values among our sample observations.
Although the rent coefficient could be used to predict the difference in
value between two properties given the rents on all other land within the
Valley, most economists would argue that such regression coefficients cannot
be used to predict changes in the general pattern of values or changes in the
aggregate value of all properties when the pattern or level of rents across
the entire Valley changes.* In most instances we would agree with this objec-
tion to our procedure, but in this case we would not agree for the following
reasons. The regression results in Tables E. 2 and E.4 show rent-per-acre
regression coefficients having magnitudes which imply that declines in rents
would be perfectly capitalized into lower land values at interest rates of
9-11 percent. Since (1) marginal rates of return in the private sector vary
from 8 to 12 percent and (2) agricultural land values in the Valley greatly
exceed capitalized estimated rents at the foregoing interest rates, we regard
the estimates in columns (3)-(5) of Table III.3, which do imply perfect capit-
alization of rent decreases at reasonable interest rates, as probable over-
estimates of the decline in land values realistically expected to follow intro-
duction of each control policy.
In conclusion, we report the estimates in Table III.3 simply to pro-
vide additional perspective on the implications of the three open field burning
control policies under investigation in this report.
*
We readily admit that this would surely be the case following imple-
mentation of open field burning control policies on the scale investigated here.
55
-------
IV. THE BENEFITS OF ALTERNATIVE OPEN
FIELD BURNING CONTROL POLICIES
A. Introduction
As indicated in the introduction to this report, open burning of
agricultural fields takes place between mid-July and late September,
and the resulting smoke constitutes a major seasonal air pollution problem
for Oregon authorities. Most concern has arisen about the visibility re-
ductions associated with this agricultural practice. Since the burning
occurs at the height of the tourist and resident vacation and outdoor-
recreation seasons and the desirability of such activities intuitively
appears to vary directly with visibility, many Willamette Valley residents
argue that tourists and residents select different activities, alter their
expenditure patterns, and experience a rather pervasive decline in the
quality of their environment because of field burning.
In contrast to pollution problems elsewhere, there ±3 little or no
concern about the effect of field burning smoke on health, soiling,
materials damage, land values, or the selection of plant locations and/or
production processes by existing or potential industries in the Willamette
Valley. Undoubtedly the relatively short burning season and the quite
variable pollution levels experienced at different locations combine to
explain this absence of concern. In fact, there is exceptionally little
"hard" evidence on these matters, largely because most strategies for
measuring the effects of pollution require persistent and stable differences
in air quality. Consequently, since smoke from open field burning is
neither persistent nor consistently distributed throughout the Valley for
very long periods, measurement of the benefits associated with alternative
field burning control policies required us to focus on air quality as
(1) a determinant of the "state of the environment" experienced by residents
and tourist and (2) a determinant of household recreation and vacation
activities.
In section B below, we develop measures of the improvement in visibility
that would occur under alternative policies restricting acres burned.
There, we measure changes in the visibility component of the public good
which we term "the state of the environment." Our measurement index is
aggregate-people-miles-of-minimum-visibility, and we follow standard
benefit measurement practice by measuring the policy-induced changes in
that index from a base representing optimal field burning practice. No
attempt to value these changes in a "public good" is made in this section.
Instead, valuation questions are addressed in Part V of the report.
In section C, attention turns to Willamette Valley resident outdoor
recreational activity under alternative field burning control policies.
Estimates of the value of increases in resident recreation activity are
57
-------
developed by employing the results of empirical investigations reported in
appendixes D and H, estimates of visibility changes from section B, and
unit-recreation-activity values suggested by the Water Resources Council.
Three additional observation are appropriate here. First, no
estimates of the impact of changes in air quality on tourist-behavior in
Oregon are reported here. In fact, the multi-variate statistical analysis
of tourist survey data presented in appendix H discovered no support for
the hypothesis that tourist behavior in Oregon in August 1971 was affected
by air quality deterioration due to smoke from open field burning. Con-
sequently, at least for now, we must conclude that tourist behavior,
once committed to an Oregon vacation, is largely unaffected by pollution
conditions which they discover in the Willamette Valley. Tourists typically
pass through the Valley on route to the coastal, plain, or mountain attrac-
tions which brought them to Oregon. They are therefore exposed to the ef-
fects of open field burning for relatively short periods when compared to
residents, and hence it is not too surprising that tourist behavior is
not statistically significantly affected by air quality variation in the
Valley.
This brings us to our second observation. The absence of evidence
that visibility restrictions due to open field burning smoke do affect
short-run tourist decisions once they have embarked on an Oregon vacation
does not establish whether or not significant longer-run effects exist.
For example, the rate at which tourists return to Oregon during their
future vacations conceivably may be affected marginally by their exposure
to the effects of open field burning. Unfortunately no evidence is now
available either to support or to refute this hypothesis. To generate
"hard" empirical evidence on this and most other conceivable effects of
open field burning would appear to us to require a misallocation of research
effort. In our view, the major benefits of alternative field burning
control policies are received by Willamette Valley residents in the form
of an improved "state of the environment," and the magnitudes of these
benefits are reasonably well-estimated in sections B and C below.
Finally, the measurement index developed here is not intended to
imply that individuals' attitudes toward, and presumably valuations of,
improvements in visibility do not exhibit considerable variation. Avail-
able evidence for Eugene and Salem reveals that such an assumption would
grossly misrepresent the facts.*
*See Robert G. Mason, The Effect of Air Pollution on Public
Attitudes and Knowledge, a report prepared for APCO,EPA under the same
contract with Oregon State University as this report (Corvallis, Oregon:
June 1972).
58
-------
B. Visibility Under Alternative Field Burning Control Policies
In appendix D of this report and in our second annual report, we have
presented strong evidence that (1) variations in acres-burned statistically
explain a major fraction of observed variations in minimum daylight visi-
bility in Eugene and Salem and (2) DEQ policies regulating acres-burned
according to present and predicted meteorological conditions have reduced
smokiness and increased minimum visibility in Eugene while producing the
opposite effects for Salem. We have also employed 1971 burning season
data and our multiple-regression findings to calculate equations which
relate minimum visibility in Eugene and Salem to relevant acres-burned
(see the final section of appendix D). Since increases in south Willamette
Valley acres-burned under northerly wind conditions reduce visibility
in Eugene, while increases in south acres burned under southerly wind con-
ditions reduce minimum visibility in Salem, we now employ our calculated
pollution production function equations to determine how alternative burning
programs with respect to south Valley acres in 1971 would have affected
visibility in Eugene and Salem. Put slightly differently, we seek to
determine the optimal burning program for south Willamette Valley acreages.
Only after the specification of the optimal burning program with no
in total acres burned can one measure the genuine gains in visibility
attributable to reductions in acres burned.
In 1971, 132,000 acres were burned in the southern half of the
Willamette Valley on the 25 days with meteorological conditions compatible
with those postulated in our estimation of pollution production functions
for Eugene and Salem. Since only 170,000 acres were burned in the south
Valley during the entire 1971 burning season, we do not believe that it
is unreasonable to characterize an optimal program based on just 25 days
for our present purposes, though we would urge cautious, additional research
before actual implementation of the optimal program presented here.
Finally, in making the required computations, we postulate that northerly
winds occur on 16 days and southerly winds on 9 days — those conditions
exactly matching what in fact occurred in 1971.
Granted the value judgment that miles of minimum visibility in Salem
and Eugene are equally valuable per affected individual,* the optimal
acreage to burn under southerly wind conditions, X , may be obtained by
*Although other value judgments are possible, none appear more
reasonable than this one. Moreover, none of our conclusions would change
if we postulate that miles of minimum visibility in either Salem or
Eugene were 2,3,4, or even 5 times more valuable at the margin than in
the other city. Evidence to support this judgment is provided below.
59
-------
maximizing the following welfare function:
(16) W = W16N_V_ + 9N V
SL E S S
where w = welfare weight for miles of minimum visibility per affected
person in Eugene,
Np = number of persons in Eugene urban area,
N = number of persons in Salem urban area,
V_, - 11.3 - 0.00418 /132,000-X\ + 7.127 x 10~6 f 132,OOP -
E r~i6—) a . „ (—16—
Vc = 14.0 - 0.0006/X\ + 1.792 x 10""
or V_ = 14.0 - 0.00036/XV
{9l
X = south Willamette acres-burned under southerly wind conditions.*
One obtains the value of X which would maximize this function by setting
its partial derivative with respect to X equal to zero and solving for
X . Interestingly, the same solution is obtained with non-linear or linear
pollution production functions: X°>132,000 with w = 1. Therefore, the
optimal burning program with respect to south Willamette Valley acres
requires that all acres be burned under southerly wind conditions, none
under northerly wind conditions. Moreover, this result is not sensitive
to changes in total acres burned (at least up to 170,000) or moderate
changes in postulated minimum visibilities in the absence of field burning.
The three major reasons for this result are easily stated. First,
a given increase in daily acres-burned under northerly wind conditions re-
duces minimum visibility in Eugene approximately 7 to 11 times more than
the same increase under southerly wind conditions would reduce visibility
in Salem. Visibility in Eugene is clearly more sensitive to smoke from
open field burning than Salem. Second, other things equal, Salem minimum
visibilities exceed those in Eugene by a substantial margin. Thus, changes
in Salem visibility are necessarily absolutely and relatively smaller in
response to field burning than those in Eugene. Third, the Eugene urban
area is more populous than the Salem area (139,319 v. 92,866 according to
the 1970 U.S. Census), thereby weighting miles of minimum visibility in
Eugene more heavily.
*The pollution production function equations are taken from
appendix D. These functions are based on 1971 data which, of course,
is quite appropriate in the present context.
60
-------
Two further interesting conclusions can be reached with the aid of
our proposed welfare function. With w = 1, one can compute W under
(a) the actual 1971 burning program where X = 117,000 acres and (b) the
optimal burning program where X = 132,000. Under the actual 1971 program
W equals 25,628,000 people-miles-of-visibility (pmv), while under the
optimal program W equals 32,757,000 pmv with non-linear Salem pollution
production function, or 32,477,000 pmv with the linear Salem function.
Although such numbers must be interpreted with care, they do afford an
opportunity to develop indexes to measure improvements in air quality.
This is done below. Second, one can calculate the value of w implicit
in the 1971 burning program by setting X equal to 117,000 and finding
the value of w which would result in the partial derivative of (16)
with respect to X equal to zero. When this was done, we found w equalled
0.0555. In words, the marginal value of improvements in Eugene minimum
visibility implicit in the 1971 burning program appear to have been only
about 6 percent of the marginal value of improvements ia Salem. Of course,
this finding must be interpreted with great caution. We do not wish to imply
that the DEQ burning program has been consciously designed to generate larger
benefits for Salem than Eugene; we know that this is definitely not the case
because with no regulation of field burning Salem visibility would be greater
and Eugene visibility lower; and we know that the DEQ burning program has pur-
posefully been designed to improve Eugene visibility to the disadvantage of
Salem. Instead we wish to stress that, jLf_ miles of visibility in the Salem
and Eugene urban areas are treated as equally valuable per person and our
pollution production functions are accurate, future DEQ burning programs
should increase south Valley acres burned under southerly wind conditions.
Despite the predictable deterioration in Salem visibility associated with
this type of program, it would be consistent with maximization of at least one
one "reasonable" welfare function.*
Although all other burning programs associated with no reduction in
total acres burned in the south Valley result in fewer people-miles-of
minimum-visibility than the optimal policy of burning only under southerly
wind conditions, we would be remiss not to report our findings concerning
at least three of these policies. The three policies examined have the
following objectives: (I) absolute equality of mininum daylight visi-
bility in Eugene and Salem; (2) equality of the absolute declines in
minimum daylight visibility in the two cities; and (3) equality in the
relative declines in minimum daylight visibility in each city. To
*The adoption of the optimal burning program would not appreciably
increase the implicit marginal value of improvements in Eugene minimum
visibility. Under the optimal policy, w = 0.0574 or just 3.6 percent more
than under the 1971 burning program. Therefore, as noted in an earlier
footnote, none of our conclusions would be changed if we postulated miles
of minimum visibility in Salem to be considerably more valuable at the
margin than in Eugene. In a very important sense, in this instance
alternative specifications of the parameters and forms of the social welfare
function would have no impact (except in extreme cases) on our conclusions.
61
-------
determine the appropriate acreage to be burned under southerly wind conditions,
we employ our calculated pollution production function equations presented
above and solve the following three equations for X:
VE = Vg (Absolute Equality)
11.3 - V£ = 14.0 - V (Equal Absolute Declines)
/11.3 - VE\ = /14.0 - V \ (Equal Relative Declines)
11.3 14.0
Our solutions to these equations indicated that approximately 122,000
South Valley acres should be burned under southerly wind conditions to
assure that Salem has the same visibility as Eugene when 10,000 acres are
burned under northerly wind conditions. In contrast, to achieve equal
absolute and relative declines in Salem and Eugene minimum visibilities,
about 112,000 and 110,000 acres, respectively, should be burned under
southerly wind conditions, the remaining 20,000 and 22,000 acres being
burned under northerly wind conditions.
Undoubtedly the most remarkable feature of these results is that they
neatly bracket the south Willamette Valley acreage actually burned under
the 1971 DEQ burning program. Although these policies fail to consider the
numbers of persons experiencing visibility restrictions due to open field
burning, none of the policies can be regarded as particularly outlandish.
Therefore, our findings here provide additional evidence that the present
DEQ burning program is basically sound; our single recommendation for
improvement would be to increase acres burned under southerly wind conditions.
We now turn to perhaps our most interesting and important task, the
estimation of the net change in total people-miles-of-minimum-visibility
(pmv) under three alternative policies regulating open field burning.
Our procedure is rather straight-forward and begins with the estimation
of total people-miles-of-minimum-visibility in the Willamette Valley on
the 27 days during 1971 for which our statistical investigations (see
appendix D) suggest minimum daylight visibilities in Salem and Eugene are
predictably related to acres burned in different zones of the Valley.
Table IV. 1 reports these estimates for the optimal burning program with
no reduction in acres burned and three alternative policies regulating
burning, as well as the particular estimates of population and expected
miles of minimum visibility which underlie the total pmv estimates.*
*Although our best estimates of minimum daylight visibility
in Salem under southerly wind conditions are based on the non-linear
pollution production function equation presented earlier in this section
in the notes to Table IV. 1 we also report estimates based on the linear
equation V = 14.0 - 0.00036 (X/9). We do this because such differences
in functional form have considerable effect on our total pmv estimates
for Salem under the optimal burning program and thereby influence our
estimates of the net change in total pmv for the Valley under different
policies.
62
-------
Table IV.1. Estimates of Total People-Miles-of-Minimum-Daylight-Visibility Under Alternative
Open Field Burning Policies for 27 Days in the 1971 Burning Season
North Winds, 18 Days
North Valley population
North acres burned
Salem minimum visibility
Total North Valley pmv
South Valley population
South acres burned
Eugene minimum visibility
Total South Valley pmv
South Winds, 9 Days
North-Valley-plus population
South acres burned
Salem minimum visibility
Total North-Valley-plus pmv
Optimal Ban on
Open Open
Burning Burning
340,000 340,000
74,1166
10.67 13.80
65,300,000 84,456,000
340,000 340,000
15,000
11.30 11.30
69,156,000 69,156,000
450,000 450,000
132,000
9.06a 14.0
36,673,000d 56,700,000
North-Valley-minus population 110,000 110,000
North acres burned
Salem minimum visibility
Total North-Valley-minus
pmv
Eugene area population
Eugene minimum visibility
Eugene area pmv
Willamette Valley Totals, 27
Total acres burned
Total Valley pmv
10,346
10.67 14.00
10,563,000 13,860,000
200,000 200,000
15.1 15.1
27,180,000 27,180,000
Days
181 ,41?
208.872.0008 251,352,000
Alternate
Year Open
Burning
340,000
12,033
12.17
74,480,000
340,000
ii;so
69,156,000
450,000
66,000
10.56
42,768,000
110,000
5,173
12,34
12,216,700
200,000
15.1
27,180,000
83,206
225,801,000h
One in Three
Years Open
Burning
340,000
8,022
12.70
77,724,000
340,000
11.30
69,156,000
450,000
44,000
11.92°
48,276,000f
110,000
3,449
12.89
12,761,000
200,000
15.1
27,180,000
55,471.
235,097,000i
Note: See the text for sources and computations procedures.
a Linear function estimate -
b Linear function estimate =
c Linear function estimate =
d Linear function estimate =
e Linear function estimate •
f Linear function estimate =
g Linear function estimate «
h Linear function estimate =
i Linear function estimate »
8.72
11.36
12.24
35,316,000
46,008,000
49,572,000
207,516,000
229,041,000
236,393,000
-------
Three comments concerning these estimates are in order. First, al-
though our estimates of the populations in various zones of the Willamette
Valley affected by open field burning smoke under specified wind conditions
are not presented with great confidence, we do believe that they are of the
"right order of magnitude." Second, we employ minimum visibility estimates
for Eugene and Salem as estimates of visibility for large zones of the
Valley with much reluctance. However, there is no alternative because
visibilities are not regularly reported at other locations in the Valley.
Third, we employ our calculated pollution production function for Salem
under northerly wind conditions to estimate minimum visibility in the
northern portion of the Valley under southerly wind conditions when north
Valley acres are burned; again, in the absence of other evidence, we adopted
this procedure to develop total pmv estimates that "appear to be of the
right order of magnitude."
For these reasons we must admit that there are sources of unreli-
ability both in the data and in the concepts employed to obtain the total
pmv estimates for each zone of the Valley under different wind conditions.
Moreover, since we have no independent estimates of the zone totals, the
total pmv estimates for the full 27 days have all the unreliabilities that
their components in combination contribute. Nonetheless, we doubt that
the total pmv estimates are more than 15 percent either way from the
"true" figures.
Of course, the expected net increase in total pmv attributable to
each alternative open field burning policy for the 27-day portion of the
1971 burning season is obtained by substracting the total pmv attainable
under the optimal burning program (i.e., burn all south Valley acres
under southerly wind conditions) from the total pmv associated with each
policy. The first column of Table IV.2 reports these figures. It should
be recalled that this procedure purposely excludes increases in pmv
attainable with no reduction in total acres burned from the increases in
pmv attributable to policies that reduce total acres burned. To do other-
wise would overstate the genuine net increase in total pmv attributable
to each policy.
Since we have been unable to estimate satisfactory models to predict
how minimum daylight visibility would vary with changes in acres burned
under all, meteorological conditions, we must employ an extrapolative pro-
cedure to obtain estimates of the increase in total pmv for the entire 1971
burning season^ Following careful examination of the alternatives, we de-
cided that it would be best to use an extrapolative technique based on (a)
the estimated increase in total pmv attainable in the 27-day period and (b)
the percentage of total Valley acreage burned during the remainder of the
burning season. In 1971, 181,412 acres of the total 233,379 acres were
burned during the 27-day for which we have estimated total pmv in the Valley;
the remaining 41,967 acres, or 22.27 percent of the total, were burned during
the other 50 days in the season.*
*State of Oregon, Department of Environmental Quality, Air
Quality Control Division, Field Burning in the Willamette Valley, 19713
a report date April 30, 1972, Table II, pp. A4-7.
64
-------
Table IV.2.
Estimates of the Expected Increase in People-Miles-of-
Minimum-Daylight-Visibility Under Alternative Open Field
Burning Policies for the 1971 Burning Season.
Assumed Pollution Production
Functions and Open Field
Burning Policies
Willamette Valley Residents
27 Days'
(1)
Season
(2)
Tourists
Season
(3)
Non-linear Functions
Ban on open burning 42,480,000
Alternate year open burning 16,929,000
Once in three years open burning 26,225,000
53,100,000
21,161,000
32,781,000
5,310,000
2,116,000
3,278,000
Linear Function for Salem
Ban on open burning
Alternate year open burning
Once in three years burning
a Derived from Table IV. 1 as
b Col. (1) x 1.25
c Col. (2) x 0.10
43,836,000
21,525,000
28,877,000
explained in text.
54,795,000
26,906,000
36,096,000
5,480,000
2,691,000
3,610,000
65
-------
Given that the remaining 52,000 acres would be burned over a period almost
twice as long as that period in which 181,000 acres were burned, it ap-
peared reasonable to increase our 27-day estimates by 25 percent to obtain
estimates of the net increase in total pmv for the full burning season in
1971. The second column of Table IV.2 indicates our basic estimates of
the total increase in people-miles-of-minimum-visibility that Willamette
Valley residents would have experienced during the 1971 burning season
under each of the three policies restricting open field burning.
Next we turn to estimation of the total days spent in the Willamette
Valley during the 1971 field burning season by out-of-state tourists. Per-
haps 75 percent of the almost 6 million out-of-state summer tourists
visited Oregon during the field burning season. Well-informed Oregon State
Highway Division personnel suggest that possibly 80 percent of these
visitors spend ore day in the Willamette Valley during their average 3.9
days in Oregon. Consequently, we estimate that out-of-state tourists
spent approximately 3.6 million days in the Willamette Valley during the
1971 burning season.*
Since the 680,000 Willamette Valley residents spent about 40 million
days in the Valley during the 1971 field burning season, out-of-state
tourists would constitute slightly less than 10 percent of the population
potentially affected by field burning. If one assumes that residents and
tourists are similarly exposed to visibility reductions from open field
burning, out-of-state tourists would experience increases in total pmv
equal to perhaps 10 percent of the increase experienced by Valley residents.
On these assumptions, the third column of Table IV.2 presents our "most
reasonable" estimate of the total increase in people-miles-of-minimum-
visibility that out-of-state tourists in the Willamette Valley would
have experienced under each of the three policies restricting open field
burning.
Finally, Table IV.3 puts our estimates of the increases in visibility
under each of the three policies in slightly different perspective by
expressing the increases as a percentage of the totals attainable under
the optimal burning program. Of course, how substantial one judges these
increases to be depends on how much one values improved minimum daylight
visibility. In the concluding part of this report, the estimates presented
in Table IV.2 are used to determine the minimum marginal value of
visibility improvements required for each of the three policies restricting
open field burning to have benefit-cost ratios equal to one.
*The estimates made in this paragraph were developed from data
reported in 1971 Qut-of-State Tourist Revenue Study, a report by the
Economics Unit, Planning Section, Oregon State Highway Division, dated
February 1972, following consultation with Highway Division economists.
66
-------
Table IV.3. Percentage Increases in People-Miles-of-Minimum-Daylight-
Visibility Under Alternative Open Field Burring Policies
for the 1971 Burning Season
Salem Pollution Production
Policy Function Under Southerly Winds
Non-linear Linear
Ban on open burning 20.34% 21.03%
Alternate year open burning 8.10 10.37
Once in three years open burning 12.56 13.92
Note: Percentages calculated with Willamette Valley totals, Table IV.1, and
increases in totals, column (1) of Table IV.2.
67
-------
C. Resident Outdoor Recreational Activity Under Alternative Field
Burning Control Policies
In appendix H of this report, Sorenson has presented the results of
his empirical investigations of attendance or use at certain public or
private facilities in Eugene and Salem. That investigation employed multiple
regression analysis to examine the determinants of swimming pool and golf
course use, as well as the number of visitors at the Oregon State Capitol
Building in Salem and overnight campers at Armitage State Park near Eugene.
Attendance or use was treated as the dependent variable, while the inde-
pendent variables include minimum daylight visibility, high temperature,
and day of the week. The emprical results obtained must be regarded as
mixed, particularly with respect to the statistically suggested zero
relationships between rounds of golf, indoor swimming pool attendance in
Eugene, and minimum visibility. On the other hand, the estimated relation-
ships between minimum visibility and attendance at outdoor swimming pools
in both Eugene and Salem were positive and statistically significantly
different from zero at the 95 or 99 percent confidence level. Overall,
Sorenson concluded that his statistical results provided substantial
presumptive and empirical support for the hypothesis that air quality
affects the extect to which individuals undertake outdoor recreational
activities. He further suggests that the evidence supports the hypothesis
that consumers in fact do engage in activity substitution under variation
of air quality.
Here, however, we accept Sorenson's findings and inquire into the
magnitudes and values of recreational activity changes induced by variations
in air quality. We begin this exercise by presenting the arc-elasticities
of swimming pool attendance in Salem and Eugene with respect to minimum
daylight visibility:
Eugene, outdoor public pools 0.18
Eugene, indoor public pools 0.10
Eugene, all public pools 0.16
Salem, outdoor public pools 0.20
These elasticities were calculated from Sorenson's study over the range
of minimum visibilities in Eugene and Salem associated with the optimal
burning program and the three policies restricting open field burning
under investigation. Only the indoor public pool elasticity in Eugene.is
calculated with a statistically insignificant regression coefficient.
The above elasticities assume major importance here because (1) a
large fraction of the Eugene and Salem area populations has relatively
easy access to the pools under study, (2) swimming is clearly the number
one outdoor recreational activity tor Willamette Valley residents during
the field burning season, and (3) a high proportion (perhaps 50-75
percent) of all swimming activity undertaken by residents occurs at
locations where visibility is commonly affected by smoke from open field
68
-------
burning.* Given our estimate that Willamette Valley residents swam on
1.5 to 2.3 million occasions in zones affected by field burning smoke
during the typical recent season, the above elasticities imply that the
20 percent improvement in visibility associated with a ban on open field
burning (see Table IV.3) would increase the days on which residents swim by
60,000 to 90,000. Although the implied 4 percent increase in swimming
activity appears small, it amounts to the not inconsiderable sum of
$30,000 to $45,000 when valued at the admission fee of $0.50 charged
adults at Salem and Eugene public pools.
Moreover, at a minimum, these calculations based on relatively "hard"
numbers suggest that the value of outdoor recreational activity affected
by visibility restrictions due to open field burning may be quite sub-
stantial. For example, if other outdoor recreational activity were both
affected by field burning smoke and valued similarly to swimming, which
constitutes about 16 percent of total resident recreational activity, then
a ban on open field burning might result in an increase in recreational
activity valued between $180,000 and $270,000. However, the theory and
regression results presented in appendix H suggests that these figures
probably inaccurately value outdoor recreational experiences foregone by
Valley residents because of visibility restrictions due to open field
burning for the following reasons. First, there is no evidence to support
the hypothesis that all outdoor recreational activities engaged in by
Willamette Valley residents are affected similarly by variations in
visibility-** In fact, appendix H reports evidence refuting this hypothesis
and suggesting that some outdoor activities may hardly be affected at all.
Second, individuals very probably substitute one recreational activity for
another within the field burning season or in some other season, there-
by reducing the net impact of variations in visibility due to open field
burning. Third, not all resident outdoor recreational activity is under-
taken within the Willamette Valley. On the contrary, Valley residents
are peripatetic outdoor recreators who regularly travel to coastal and
mountainous areas to engage in what we classify as recreation activities
here.
*An as-yet unpublished, Oregon State Highway Division telephone
survey of recreational activities engaged in by Oregon residents, by
county, provides data suggesting that swimming activity constitutes
16.7 percent of total Willamette Valley resident recreational activity
during the field burning season. Pleasure driving, bicycling, picnicking,
and outdoor games are ranked 2 through 5 by Valley residents for the
burning season. All numbers of persons participating in outdoor recre-
ational activities cited in the text are derived from this Highway Division
survey.
**The types of outdoor recreation activity to which we have
reference here include: fishing, boating and water skiing, swimming,
camping, hunting, bicycling, horseback riding, playing outdoor games
and sports, picnicking, walking for pleasure, hiking, nature study,
rockhounding, etc., pleasure driving, sightseeing, attending outdoor
sporting or cultural events, playing golf, and others feasible
during the summer months in the Willamette Valley. Each of these activities
was surveyed in the Highway Division study referred to earlier.
69
-------
For the above reasons, we consider it unreasonable to assume that
the elasticity of aggregate resident recreational activity with respect
to minimum daylight visibility is as high as 0.2, and we therefore assume
that it equals 0.1 to calculate the increases in aggregate resident
recreational activity presented in Table IV.4. There, our high, medium,
and low estimated increases assume that 90,70, and 50 percent of residential
recreational activities are undertaken within the Willamette Valley and
potentially affected by restrictions in visibility due to open field
burning smoke. Our medium estimate of 70 percent is probably the most
realistic figure for each policy.
The final three columns of Table IV.4 present our estimates of the value
of outdoor recreational activities foregone by Willamette Valley
residents because of visibility restrictions due to open field burning.
The estimates are based on the range of values suggested for the evaluation
of outdoor recreation days by the Water Resources Council.* Since these
values were developed as single unit values to be "assigned per recreation
day regardless of whether the individual engages in one activity or several,"
we regard the estimates at each value in Table IV.4 to be relatively gen-
erous because the basic data we used treated each activity separately
although it is certain that residents engaged in more than one activity
per day on numberous occasions. In any event, given individual abilities
to substitute other valuable activities for outdoor recreational activities,
our best estimates of the value of increases in Willamette Valley resident
outdoor recreation activity consequent to the three policies under in-
vestigation would be as follows: (1) $249,000 for a ban on open burning,
(2) $111,000 for a policy allowing alternate year open burning, and
(3) $160,000 for a policy allowing open burning once in three years.
*See Water Resources Council, Evaluation Standards for Primary
Outdoor Recreation Benefits," Supplement No. 1 to U.S. Senate Document
No. 97 (87th Congress, 1964), pp.4.
70
-------
Table IV.4. Number and Value of Increases in Willamette Valley Aggregate
Outdoor Residential Recreational Experiences Under Alternative
Open Field Burning Policies for the 1971 Burning Season.
Percent of
Activities Increase in Value of Increased Resident
Undertaken Resident Recreational Experiences
nej.a curning roncy Within
Valley
Ban in Open Burning 90
70
50
Recreational
Experiences3 $0.50
320,000
249,000
178,000
$160,000
124,500
89,000
$1.00
$320,000
249,000
178,000
$1.50
$480,000
373,500
267,000
Alternate Year Burning
Once in Three Years Burning
90
70
50
90
70
50
142,000
111,000
79,000
205,000
160,000
114,000
71,000
55,500
39,500
102,500
80,000
57,000
142,000
111,000
79,000
205,000
160,000
114,000
213,000
166,500
118,500
307,500
240,000
171,000
Increase = (0,1)(v)(f)(17,130,000), where 0.1 = assumed elasticity
of aggregate resident recreational activity with respect to minimum visibility,
v = average percent change in visibility under each policy (from Table IV.3),
f = percent of activities undertaken within the Willamette Valley, and
17,130,000 is our estimate of aggregate 1971 resident outdoor recreation experiences
(derived from Oregon State Highway Division special survey of recreational
activity in Oregon).
71
-------
V. A CONTRIBUTION TO THE EVALUATION OF ALTERNATIVE
OPEN FIELD BURNING CONTROL POLICIES
To conclude this investigation of the benefits and costs of alternative
open field burning control policies, we draw together our best estimates
of the benefits and costs of each policy. We do not report benefit-cost
ratios, net present values or rates of return for each policy, however.
We do not calculate such evaluative indexes because we have not heen able to
devise any practical technique for measuring the marginal value of increases
in the public good "improved visibility." On the other hand, we do attempt
to contribute to the evaluation of the policies by providing estimates of
their respective implicit total, average, and marginal costs. Unquestion-
ably our most interesting finding is that improvements in Willamette Valley
visibility are produced under conditions of decreasing average costs.
Table V.I presents the "most reasonable" benefit and cost estimates
developed in the two preceding parts of this report. Column (2) indicates
the decrease in aggregate Oregon producer rents in supply situation I, by
field burning control policy and by alternate assumed costs per acre for
the mobile field sanitizer. In contrast, column (3) reports the sum of our
predicted changes in Oregon and non-Oregon producer rents and consumers'
surpluses (or what is often termed "total social loss"), again assuming
supply situation I. As we indicated in part III and appendix F, although
we would not claim our postulated supply and demand functions perfectly
reflect industry conditions as they were during the late sixties, we do
believe the aggregate estimates in columns (2) and (3) are tolerably accu-
rate.
In contrast, we have much less confidence in our valuations and pre-
dictions of the increase in Willamette Valley resident outdoor recreation
activity. For this reason, column (4) presents our highest, lowest, and
median estimate under each policy. We prefer to make our uncertainty in
this respect obvious because these benefits affect the net measurable cost
estimates reported in columns (5) and (6) so very much. We provide esti-
mates of net measurable costs from both an Oregon and national viewpoint
because either could conceivably be considered relevant under various cir-
cumstances.
Several observations concerning the estimates in Table V.I are appro-
priate here. First, the decrease in Oregon producers' aggregate rents under
each policy is quite substantial in both absolute and relative terms. Our
estimates range from a minimum of $0.536 million to $2.599 million per year
after industry adjustments in price and output. Moreover, the Oregon rent
decrease equal 60-70 percent of the sum of changes in producers' rents
(Oregon and non-Oregon) and consumers' surpluses. Second, our highest esti-
mates of Willamette Valley resident recreation benefits never exceeds 50
percent of our lowest estimate for the decrease in Oregon producer rents.
When compared to the sum of all policy-induced changes in rents and surpluses,
recreation-benefits never exceed 40 percent. In consequence, as clearly indi-
cated by columns (5) and (6), even if resident outdoor recreationists were
required to pay according to the full value of the benefits they receive
-------
Table V.I. Summary of Benefits and Costs of Alternative Open Field
Burning Control Policies Under Various Assumptions and
From Oregon and National Viewpoints (thousands of dollars)
Mobile Sanitizer
Cost per Acre
(1)
Complete Ban
$ 5
9
13
Once in Three Years
5
9
13
Decrease in
Oregon Rents
(2)
$1,064
1,882
2,599
Burning
715
1,271
1,788
Decrease in
"Sum"
(3)
$1*610
2,835
4,002
1,085
1,922
2,706
Resident Outdoor
Recreation Benefits
(4)
$480
249
89
480
249
89
480
249
89
308
160
57
308
160
57
308
160
57
Implicit
Oregon Viewpoint
Col.(2)-Col.(4)
(5)
$ 584
815
975
1,402
1,633
1,793
2,119
2,350
2,510
407
555
658
963
1,111
1,214
1,480
1 ,628
1,731
Total Costs
National Viewpoint
Col.(3)-Col.(4)
(6)
$1,130
1,361
1,521
2,355
2,586
2,755
3,522
3,753
3,913
777
925
1,028
1,614
1,762
1,865
2,398
2,546
2,649
Alternate Year Burning
5
9
13
536
960
1,365
816
1,453
2,276
213
111
40
213
111
40
213
111
40
323
425
496
747
849
920
1,152
1,254
1,325
603
705
776
1,240
1,342
1,413
2,063
2,165
2,236
Sources: Tables III.l and IV.4
-------
from control of open field burning, that amount could only partially com-
pensate producers and/or consumers for their respective decreases in well-
being.
The implication of these findings is obvious. For all field burning
control policies investigated here, none can conceivably pass a benefit-
cost test without a positive value being given the public good "improved
visibility." Columns (5) and (6) of Table V.I indicate our estimates of
the net total costs of achieving the various improvements in Willamette
Valley visibility associated with each field burning control policy. Of
course, for a policy to pass the benefit-cost test, one must claim that
the benefits of improved visibility associated with that policy equal or
exceed the net total cost figures given in columns (5) and (6). Note that
by adopting an insular Oregon viewpoint the benefits which must be claimed
for improvements in visibility to pass a benefit-cost test are only 50-65
percent as large as those required from a national viewpoint. Consequently,
one can easily conceive of a situation where from Oregon's viewpoint one
or more policies pass benefit-cost tests which they fail if a national view-
point is adopted. Moreover, this could occur even when improvements in air
quality are valued more highly nationally than they are in Oregon.
In our opinion, however, the implicit total cost figures presented
in columns (5) and (6) of Table V.I do not place the implicit costs of each
policy in particularly useful perspective. This is the case because those
costs are not expressed relative to the population benefited or the benefits
actually associated with each policy. Therefore, we now express the implicit
costs of the visibility improvements associated with each policy per Willamette
Valley resident and per mile of increased visibility per person. Such figures
will indicate the minimum value that the benefits of improved visibility must
assume per person or per person-mile for each policy to pass the standard
benefit-cost test. Table V.2 reports these values in dollars per Willamette
Valley resident per annun, while Tables V.3 and V.4 express the same values
(in cents) as average and marginal costs of additional miles of minimum visi-
bility per person-mile.*
Although it is a matter of opinion whether the values reported in
Table V.2 are too high for any particular policy to pass a benefit-cost test,
we are impressed with the considerable variation in the minimum values that
the benefits of improved visibility must assume for policies to pass a benefit-
cost test. On the other hand, it must be observed that the variation revealed
in Table V.2 becomes unimportant for public policy formation if_ the minimum
value placed on improved visibility is sufficiently high. For example, if the
minimum value for improved visibility equals or exceed $4 per Willamette Valley
resident, all policies except one would pass a benefit-cost test; only a ban
on burning with mobile sanitizer costs at $13 per acre would fail.
*Values in Table V.2 were obtained by dividing the implicit total costs
of each policy (from columns (5) and (6) in Table V.I) by 680,000, the esti-
mated population of the Willamette Valley. Implicit average costs in Tables
V.3 and V.4 were calculated by dividing the implicit total costs of each
policy by the total increases in pmv (from Table IV.2), while implicit mar-
ginal costs were found by dividing the increment in implicit total costs by
the incremental change in pmv under each policy.
75
-------
Table V.2. Implicit Total Costs of Improvements in Willamette Valley Minimum
Daylight Visibility, by Viewpoint and Policy Under Various
Assumptions (dollars per Willamette Valley resident)
Mobile
Sanitizer
Cost per
Acre
$ 5
9
13
Outdoor
Recreation
Benefits
High
Median
Low
High
Median
Low
High
Median
Low
Strictly
Alternate
Yr . Burning
$0.48
0.63
0.73
1.10
1.25
1.35
1.69
1.84
1.95
Oregon Viewpoint
Burning Once
in Three Yrs.
$0.60
0.82
0.97
1.42
1.64
1.79
2.18
2.39
2.55
Ban on
Burning
$0.86
1.20
1.43
2.06
2.40
2.64
3.12
3.46
3.69
National Viewpoint
Alternate
Yr . Burning
$0.89
1.04
1.14
1.82
1.97
2.08
3.. 03
3.18
3.29
Burning Once
in Three Yrs.
$1.14
1.36
1.51
2.37
2.59
2.74
3.53
3.74
3.90
Ban on
Burning
$1.66
2.00
2.24
3.46
3.80
4.05
5.18
5.52
5.75
Source: Table V.I.
-------
Table V.3. Implicit Average and Marginal Costs of Improvements in
Willamette Valley Minimum Daylight Visibility, by Policy
Under Various Assumptions and From a Strictly Oregon
Viewpoint (cents per person-mile).
Mobile
Sanitizer
Cost per
Acre
$ 5
9
13
Outdoor
Recreation
Benefits
High
Median
Low
High
Median
Low
High
Median
Low
Alternate Year
Burning
Average
Cost
1.34
1.77
2.06
3.11
3.53
3.83
4.79
5.22
5.51
Marginal
Cost
1.34
1.77
2.06
3.11
3.53
3.83
4.79
5.22
5.51
Burning
Three
Average
Cost
1.18
1.61
1.91
2.80
3.23
3.53
4.30
4,73
5.03
Once in
Years
Marginal
Cost
0.81
1.25
1.56
2.08
2.52
2.84
3.15
3.59
3.90
Ban on
Burning
Average Marginal
Cost Cost
1.08
1.51
1.81
2.60
3.03
3.32
3.93
4.36
4.65
0.91
1.33
1.62
2.25
2.68
2.97
3.28
3.70
3.99
Source: Tables IV.2 and V.I.
-------
00
Table V.4. Implicit Average and Marginal Costs of Improvements in
Willamette Valley Minimum Daylight Visibility, by Policy
Under Various Assumptions and From a National Viewpoint
(cents per person-mile).
Mobile
Sanitizer
Cost per
Acre
$ 5
9
13
Outdoor
Recreation
Benefits
High
Median
Low
High
Median
Low
High
Median
Low
Alternate Year
Burning
Average
Cost
2.28
2.67
2.94
4.69
5.08
5.34
7.80
8.19
8.49
Marginal
Cost
2.28
2.67
2.94
4.69
5.08
5.34
7.80
8.19
8.49
Burning
Three
Average
Cost
2.05
2.44
2.71
4.26
4.65
4.92
6.33
6.72
6.99
Once in
Years
Marginal
Cost
1.29
1.64
1.87
2.78
3.12
3.36
2.49
2.83
3.07
Ban on
Burning
Average Marginal
Cost Cost
1.90
2.29
2.56
5.97
4.36
4.64
5.94
6.32
6.59
1.64
2.03
2.30
3.45
3.84
4.15
5.24
5.61
5.89
Source: Table IV.2 and V.I.
-------
Although the perspective provided by Table V.2 is informative, to
an economist Tables V.3 and V.4 reveal the results of this study in a more
useful fashion. There, we report the implicit average and marginal costs
for improvements in Willamette Valley visibility per person-mile, by mobile
sanitizer cost per acre, amount of recreation benefits, and field burning
control policy.* In Table V.3 where we report implicit costs from a strictly
Oregon viewpoint, person refers to Willamette Valley residents only, In con-
trast, in Table V.4 where we adopt a national viewpoint, the relevant popu-
lation includes Willamette Valley residents and tourists.
Tables V.3 and V.4 are particularly valuable for two reasons. First,
the implicit costs of field burning control policies are there expressed
relative to the objective of such policies—namely, improvements in Willamette
Valley visibility. Presumably rational policy evaluation is less difficult
and less subject to error where the costs of alternative policies are expressed
relative to the units of the objective desired. Again, one cannot objectively
judge whether these costs are sufficiently high (low) that particular policies
fail (pass) the standard benefit-cost test. Nevertheless, these cost estimates
should at least put the values of visibility improvements required for policies
to pass a benefit-cost test into proper perspective.
Second, Tables V.3 and V.4 reveal that improvements in Willamette Valley
visibility are produced under conditions of decreasing average costs. The
implications of this finding deserve special comment.
As is well known, for the attainment of an optimal allocation of re-
sources, the marginal value (price) of outputs must everywhere equal marginal
cost. If marginal value diverges from marginal cost, economic theory pre-
dicts that a reallocation of resources and some change in the output mix may
increase economic well-being. (For present purposes, we may safely and con-
veniently ignore the myriad qualifications that welfare economists have found
to apply to these propositions.) Application of the marginal cost pricing
rule creates a dilemma under conditions of decreasing costs, however. By
setting price equal to marginal cost, economic efficiency is achieved but
the operation would run at a deficit. By setting price equal to average cost
and thereby avoiding a deficit, economic welfare would not be maximized
because price would exceed marginal cost. Proposed solutions to this dilemma
include (a) setting price equal to marginal cost and financing the deficit
out of general tax revenues, (b) setting the marginal unit's price equal to
marginal cost and charging discriminatory non-uniform infra-marginal prices,
(c) setting price equal to marginal cost and charging a once-only (or possi-
bly annual) lump sum amount for the right to purchase at the marginal-cost-
determined price, as well as variations on these three. Note that each of
these solutions involve special non-market arrangements; under decreasing
cost conditions, additional taxation, price discrimination, creation of ex-
clusive groups, and other devices are required to achievs a welfare-maximizing
solution.
*Since the alternate-year burning policy improves visibility least among
the three policies, average and marginal costs for that policy must be assumed
equal.
79
-------
The relevance of the preceding remarks for the evaluation and selec-
tion of the optimal open field burning control policy in the Willamette
Valley is as follows. It is commonly believed that if air quality (here
measured by minimum daylight visibility) was not a public good and, hence,
could be valued and sold in some fashion comparable to private goods, or
±f_ transactions and bargaining costs plus institutional arrangements did
not preclude negptiations between air polluters (here seed growers) and
pollutees (here Valley residents and tourists), no public policy regulating
polluting activities (here open field burning) would be necessary. More-
over , public decisions which generate benefits and losses to different
groups (here seed growers, consumers, Valley residents, and tourists) could
be avoided. Although satisfaction of the stated conditions would undoubtedly
eliminate these public policy problems in numerous situations, satisfaction
of neither condition would obviate the need for a public policy with respect
to open field burning. Since improvements in visibility are produced under
conditions of decreasing costs, the optimal level of visibility would neces-
sarily be associated with revenues too small to compensate growers and con-
sumers if visibility "sold" at a uniform infra-marginal price. In contrast,
at non-optimal levels of visibility, there would be inefficiency and (unless
price equalled average costs) a deficit or surplus to be "managed." In
either hypothetical case, decreasing costs would be sufficient to recreate
a public policy problem with respect to open field burning, and special
arrangements involving government coercion and discrimination among various
groups would be required to achieve a welfares-maximizing solution. Conse-
quently, the problems associated with open field burning are of public con-
cern for more reasons than are generally appreciated.
In conclusion, we remind the reader of a very simple point. Although
the marginal value of improved visibility in the Willamette Valley is un-
known, reasonable estimates can and have been made of the implicit values
that such improvements must assume for alternative open field burning con-
trol policies to pass standard benefit-cost tests. (see Tables V.2, V.3,
and V.4). Given such information, a welfare-maximizing solution to the
field buraing problem in the Willamette Valley reduces itself to specifi-
cations of (a) a reasonable minimum value for improved visibility and (b)
the relative desirability of alternative methods (including doing nothing)
to distribute equitably the gains and losses under the selected policy.
Following tradition, we suggest that these specifications are appropriately
made in the political arena.
80
-------
APPENDIX A
NOTES ON TWO-SECTOR GENERAL
EQUILIBRIUM MODELS OF PRODUCTION
AND DISTRIBUTION
This appendix develops materials that underlie the general equili-
brium model presented in the main text of this report. Some of the
materials are standard fare in economic textbooks, while others are new.
In either case, however, numerous propositions and relationships asserted
in the main text are derived and/or explained here.
The first section of the appendix provides brief derivations of
possible production and distribution situations in two-sector general
equilibrium models, initially for an economy with no intermediate goods
and then for an economy with inter-industry flows of materials. The de-
finitions and assumptions on which our later analysis relies are stated
in this first section. In the second section techniques are developed to
analyze the impact of changes in factor supply and technology on produc-
tion possibilities. This section is fundamental because pollution con-
trol policies often involve such changes. The third section concludes
the appendix by specifying the impact of technological external disecono-
mies on production possibilities. The derivation reported there is new.
A.I Fundamental Production and Income Distribution Possibilities
This section presents production and income distribution possibili-
ties in a two-sector general equilibrium model where factor supplies and
production technology are fixed.* Figure A.I shows the usual Edgeworth-
Bowley production box diagram. The vertical sides of the box (00X = 0*0 )
measures the economy's fixed supply of capital, while the horizontal
sides of the box (OOy = O'Ox^ measure its supply of labor. Isoquants
for X are drawn with reference to 0^, and the isoquants for Y are drawn
with reference to Oy The tangency points of the two sets of isoquants
trace out the efficiency locus OxQOy. The efficiency locus indicates the
alternative factor allocations between X and Y which maximize the output
of either good given the output of the other good.
Assuming both production functions are homogeneous of the first
degree (i.e., constant returns to scale), the outputs of X and Y can be
measured for any point in the box by the distances cut off by the appro-
priate isoquants along any ray from theorigins Ojj and 0-^. The diagonal
*The following derivation of the production possibility curve in
the Edgeworth-Bowley production box diagram uses the technique developed
by K. M. Savosnick, "The Box Diagram and the Production Possibility Curve,"
Ekonomisk Tidskrift. Vol, 60, No. 3 (September 1958), pp. 183-97.
A-l
-------
INPUT of LABOR
O'
ro
OUTPUT Of X
FIGURE A.1
-------
of the box, O^Oy, provides a convenient common ray through both iso-
quant maps. Therefore, the output of X at point Q on the efficiency
locus is measured here by the distance 0XS cutoff on the diagonal by
the isoquant xx, while the output of Y at Q is measured by OyR cut off
on the diagonal by isoquant yy. Furthermore, given constant returns to
scale, the ratios O^S/O^Oy and GyR/O^Oy indicate the respective propor-
tions of the economy's maximum possible outputs of X and Y produced
at Q.
By projecting vertically from S to the horizontal axis OOy, the
output of X at Q would be OXp measured in units such that the economy's
maximum possible production of X is OOy. Similarly, by projecting hori-
zontally from R to the vertical axis OOx, the output of Y at Q would be
OYp measured in units such that the economy's maximum possible output
of Y is OOjj. Consequently, point P is the point on the economy's pro-
duction-possibility, or transformation, curve corresponding to point Q
on the production efficiency locus. Of course, by plotting in similar
fashion the output combinations associated with all points on the effi-
ciency locus O^QOy, one would obtain the complete production possibility,
or transformation, curve OjjPOy with reference to the origin 0.
Thus, there is a one-to-one corresondence between the production-
efficiency locus and the production-possibility curve. For every effi-
cient input combination there is an output combination on the production-
possibility curve, and vice versa. This fact will prove useful later in
our analysis.
We next develop the income distribution relationship in this model.
Assuming perfectly competitive factor and product markets, the factor-
price ratio at point Q in Figure A.2 would equal the identical marginal
rates of substitution between capital and labor in both industries, i.e.,
the slope of the isoquants at point Q. Therefore, the value of capital
measured in units of labor can be obtained by constructing the line
OyM, with the same slope as the isoquants at Q, to intersect 0^0' ex-
tended to M. The distance OxM indicates national income at P measured
in units of labor, 0^0' representing the income of labor in terms of
labor and O'M representing the income of capital in terms of labor.
By projecting vertically from M to N on OOy extended, the income distri-
bution relationship is transferred to the horizontal axis. In turn, by
drawing OP, PN, and WOy (parallel to PN), OP is divided into two segments
OW and WP (=01) representing the shares of output at P being received by
labor and capital, respectively.
Points W and I are reference, as opposed to actual consumption,
points. Labor's (capital's) income is a budget line through W(I), with
reference to origin 0, with slope equal to the tangent to the transfor-
mation curve at P. Under competitive conditions, of course, the slope
of a tangent to the production-possibility curve indicates relative
commodity prices.
A-3
-------
>
O
X!
INPUT of LABOR
a
OUTPUT of X
FIGURE A. 2
N
-------
By repetition of the above procedure one obtains the income
distribution curves for labor and capital, LWL1 and KIK' , repectively.
These curves establish the location of the budget lines for labor and
capital as the intersection of rays from the origin 0 to the relevant
point on theproduction-possibility curve. Since X is the capital-
intensive commodity, capital's (labor's) share of output increases
(decreases) as the economy increases the output of X.
Next consider an economy with two outputs, X and Y, each of
which uses the two primary factors, capital and labor, and the output
of the other industry as inputs.* The productive capacity of the eco-
nomy is defined by
X = xd^.K^) - al • y (Ly,Ky)
Y = -a2 • x (LX,KX) + y(Ly,Ky)
Ko = Kx + Ky
where X and Y are final goods and a]_, a£ are fixed coefficients repre-
senting respectively the requirements of X for producing Y and of Y for
producing X.
If there were no inter-industry flows, i.e., a, and a- were assumed
equal to zero, and if the production functions for X and Y were homogen-
eous of degree one, then the usual production-possibility curve between
X and Y could be derived from a production box diagram as in the previous
section. Suppose such a derivation would result in the curve C^POy in
Figure A. 3. However, if a^ and a.^ do not equal zero, then OxPOy does not
apply. Instead, with each output also serving as an input for the other
output, the true production possibilities for X and Y are within the area
This curve may be obtained in the following fashion. Construct a
set of new axes PD and PH originating at P. Construct lines PE and PG
such that their slopes with respect to PD and PH express the fixed re-
quirement of one output per unit of the other, i.e., a^ and a£ respectively.
The intersections PE and PG with the original axes indicate the coordinates
of point P', corresponding to P, and express the maximum outputs of X and
Y available for consumption (and/or trade) in an economy where inter-indus-
try flows described by a^ and a 2 hold. Of course, repetition of this pro-
cedure for all points on the ZPZ' segment of C^POy permits construction
*This model is due to Vanek. See J. Vanek, Variable Factor Proportions
and Inter-Industry Flows in the Theory of International Trade, Quarterly
Journal of Economics, February 1963, pp. 129-142.
A-5
-------
o>
o
x
INPUT of LABOR
OUTPUT of X
FIGURE A.3
-------
of the net output for consumption without trade curve FP'F'. In the
absence of trade, segments 0XZ and OyZ' on the gross output curve repre-
sent unattainable outputs because the required inputs are not available.
For example, at Z the total output of X is absorbed in the production of
Y (OZ is parallel to GB), and any attempt to increase the output of Y
would simultaneously increase the intermediate demand for X and reduce
the quantity of X, and hence Y, produced. (A similar argument holds at
the other end of C^POy-) However, with trade in intermediate goods
possible, whenever the world terms of trade are smaller (larger) in ab-
solute value than slope of the tangent to FP'F' at F(F'), the attainable
level of final demand for Y(X) lies somewhere beyond F on the Y-axis
(F1 on the X-axis) and depends on world terms of trade.*
Points on FP'F1 (such as P') indicate net outputs for final demand,
while their corresponding points on ZPZ' (such as P) indicate gross out-
puts in the economy. Consequently, where inter-industry flows exist, the
income distribution relationship is appropriately expressed with refer-
ence to the net output production-possibility curve FP'F'. The procedure
followed in deriving the income distribution curves for labor and capital,
LWL' and KIK', respectively, in Figure A.4 differs from the procedure
presented in the previous section in only one respect: lines were drawn
northwest from N (and from Oy parallel to lines from N) to points on the
net (or opposed to the gross) production possibility curve. Rather than
reproduce our earlier derivation verbatim, we merely label Figure A.4
identically to Figure A.3 so that our previous derivation applies equally
well to the situation reported in Figure A.4. (Of course, with trade
capital's and labor's maximum (minimum) shares of output could exceed
those indicated by the endpoints of KIK' and LWL'. The precise maximum
would depend on world terms of trade.)
*For a full discussion of these matters, see Stephen E. Guisinger,
Negative Value Added and the Theory of Effective Protection, Quarterly
Journal of Economics, LXXXIII (August 1969), pp. 415-433.
A-7
-------
>
00
NPUT of LABOR
o'
OUTPUT of X
- KG LH
F'
FIGURE A.4
O
Y
9
N
-------
A.2 Changes in Factor Supply and Technology*
Consider first the case of a reduction in the supply labor. Its
effect would be to reduce the size of the production box diagram and
shift the efficiency and production-possibility loci toward the origin.
For example, Figure A.5 reproduces the production box diagram 00x0'Oy
and production possibility curve OxPOy presented in Figure A.I. With a
reduction in labor supply, the size of the box decreases to OOxO*OY* and
the efficiency locus becomes OxQ*Oy*. To determine the new production-
possibility curve, we once again measure outputs by distances cut off
by isoquants along rays projected from the isoquant map origins, now,
QX and Oy*- Outputs are no longer measured by distances on the produc-
tion box diagonal, however. Instead, outputs are measured along rays
with identical slopes from origins QX and Oy*. To do otherwise would
alter the units in which outputs of X and Y were originally measured.
In Figure A.5, rays OxS*M* and Oy*R*N* have identical slopes equal to
the slope of OxOy amd serve the purpose originally fulfilled by the
diagonal OxOy. (The diagonal of the new box serves no useful purpose
and, therefore, is not indicated in Figure A.5.)
If the economy devoted its entire resource endowment to the pro-
duction of X, it would operate on isoquant x«* at 0*. Therefore,
maximum attainable output of X would be measured by the distance OxM*
cut off by x»«* on the ray OxOy, while the output of Y would be zero.
By projecting vertically from M* to the horizontal axis, the output of
X at Oy* would be OX^* measured in units such that the economy's pre-
vious maximum possible output of X is OOy. Similarly, if the economy
maximized its output of Y, it would operate on isoquant y-^* at QX and
the maximum output of Y is obtained by projecting horizontally from N*
to the vertical oxis. By employing ray Oy*N* to measure output of Y, the
maximum possible output of Y, OYjj*, is measured in the same units before
and after the reduction in labor supply.
Finally, consider the outputs of X and Y associated with the Q*
allocation of factor supplies. The output of X at point Q* is measured
by the distance OxS* on ray OxM*, while the output of Y is measured as
Oy*R* on ray Oy*N*. Projecting vertically and horizontally as before,
the outputs of X and Y at Q* would be OX* and OY*, respectively, and P*
is the point on the economy's new production-possibility curve corresponding
*For expositional simplicity only, in this and the following section
we derive shifts in production possibility curves for economies with no
inter-industry flows of materials. Since the procedures developed in the
previous section to deal with inter-industry relations can easily and
obviously be applied in situations analyzed in this and the following sec-
tion, there would appear little reason to complicate the presentation of
new analytical results needlessly.
A-9
-------
o
INPUT of LABOR
o
OUTPUT of X
FIGURE A. 5
-------
to point Q* on its new production efficiency locus. Moreover, by plotting
in a similar manner the outputs associated with all points on the efficiency
locus OXQ*OY*, the complete new production-possibility curve YM*P*XM* would
be obtained.
Of course, the above technique can be applied to all conceivable changes
in factor supply in our simple two-factor-two-commodity general equilibrium
model. To support this view, Figure A.6 is labelled identically to Figure
A.5 but shows the shift in the production-possibility curve following a simul-
taneous decrease in labor supply and increase in capital supply. The deriv-
ation of the new production-possibility curve YN P*XM* would be obtained.
Of course, the above technique can be applied to all conceivable changes
in factor supply in our simple two-factor-two-commodity general equilibrium
model. To support this view, Figure A.6 is labelled identically to Figure
A. 5 but shows the shift in the production-possibility curve following a
simultaneous decrease in labor supply and increase in capital supply. The
derivation of the new production-possibility curve Y«*P*X^* would follow
the procedure developed in the preceding paragraphs.
Continuing to assume production functions are homogeneous of the
first degree and ignoring consideration of production technologies requiring
intermediate goods, we now specify the impact of production process changes
to reduce pollution on an economy's production-efficiency locus and produc-
tion-possibility curve. The technique presented here is new. Nowhere in
the literature is the impact of technical progress (positive or negative) on
production-possibilities specified in a fashion comparable to the technique
developed by Savosnick to derive the production-possibility curve from the
production-efficiency locus. Although the following analysis concentrates on
the "negative technical progress" associated with production process changes
to reduce pollution, the techniques presented are obviously and easily
applicable in situations where there is positive technical progress. Similar-
ly, although the analysis deals only with changes in the production function
for X, our results are equally applicable to changes in the production
function for Y.
Production process changes to reduce pollution can be classified in
a fashion similar to that employed by Hicks to classify (positive) technical
progress.* Production process changes to reduce pollution would consist of
changes in production functions such that more input(s) are required to
produce any given level of output, i.e., negative technical progress. Such
changes are classified here as neutral, capital-using, or labor-using
according to their effect on the marginal productivities of capital and
labor. A neutral change would be one which reduces the marginal productivi-
ties of both factors in the same proportion. A capital-using change would
*j. R. Hicks, The Theory of Wages, McMillan & Co. (London), 1932,
131ff.
A-ll
-------
o*
NPUT of LABOR
Q*
>
Fo
N
O
OUTPUT of X
O
Y
FIGURE A. 6
-------
reduce the marginal productivity of capital less than the marginal produc-
tivity of labor. A labor-using process change would reduce the marginal
productivity of capital more than the marginal productivity of labor. Each
of these changes is depicted in Figure A. 7, where isoquants xx and x'x'
indicate alternative combinations of capital and labor required to produce
some given level of X before and after a production process change to re-
duce pollution. With given relative factor prices (P-.P-, i.e. parallel to
P P ), a neutral production process changes would not cnange the capital-
labor ratio employed to produce X, whereas a capital-using (labor-using)
process change would increase (decrease) the capital-labor ratio.
With definitional matters completed, the analysis of production process
changes may begin. Consider first the case of a neutral production process
change in the X industry. Its effect is to shift X isoquants uniformly
proportionately away from their origin, 0 , thereby reducing the outputs
associated with the set of original isoquants proportionately, say m per-
cent. For this reason, a neutral change produces no change in the produc-
tion-efficiency locus in Figure A.8, although it does produce an m percent
decline in the output of X associated with each possible efficient factor
allocation between X and Y. Consequently, outputs of X are no longer
appropriately measured along the diagonal 00 but rather along the ray
0 M, where OM = (1-m) (00 ) , or the maximum possible output of X following
tne postulated neutral production process change to reduce pollution.
Measuring along 0 M, the output of X at Q on the efficiency locus is now
measured by the distance OS', where S' lies on 0 M at the intersection
of a horizontal projection from S. Projecting vertically from S', the out-
put of X at Q would now be OXp1, which equals (1-m)(OXp). Consequently,
point P1 is the point on the economy's production-possibility curve corres-
ponding to point Q on the production-efficiency locus after the postulated
neutral production-process change, and the new production-possibility curve
would be 0XP'M. (Note that X and Y continue to be measured in the same units
and with reference to the origin 0.)
Next, the impact of a labor-using production process change in X to
reduce pollution is specified with the aid of Figure A.9. As before, 0 QO
and 0 POY represent the original production-efficiency locus and production-
possibility curve, respectively, and yy, xx, and x x are typical isoquants
of production functions with homogeneity of degreemone. A shift to a labor-
using production process in X would shift the production-efficiency locus
from OyQOy to °XQ'0Y, and the maximum possible output of X would decline
from 00 {that output associated with isoquant x x ) to OM (that output
associated with isoquant x'x^). With a labor-using, linearly homogeneous
production function, outputs of X can again be measured along the ray OJ1
because there would necessarily be equal proportionate reductions in the
output obtained from any factor combination. Thus, the output of X at Q'
on the new production-efficiency locus would be measured by the distance
OS", where S' lies at the intersection of a horizontal projection from T
(where x'x' cuts the diagonal). Consequently, projecting vertically from
S', the output of X at Q' is determined to be OXp' and point P' lies on the
economy's new production possibility curve 0 P'M.
A.
A-13
-------
R
LABOR
A. NEUTRAL CHANGE
51
5
"X,
O
•*• LABOR
B. CAPITAL USING CHANGE
5
-LABOR
O
C. LABOR USING CHANGE
FIGURE A. 7 PRODUCTION PROCESS
CHANGES TO REDUCE POLLUTION
A-14
-------
INPUT of LABOR
>
o
OUTPUT of X
FIGURE A.8 SHIFT IN PRODUCTION - POSSIBILITY CURVE
WITH NEUTRAL PRODUCTION PROCESS CHANGE
IN X TO REDUCE POLLUTION
-------
>
o
X
INPUT of LABOR
YP
o
OUTPUT of X
FIGURE A.9
p
X
SHIFT IN PRODUCTION-POSSIBILITY
CURVE WITH LABOR-USING
PRODUCTION PROCESS CHANGE
IN X TO REDUCE POLLUTION
-------
Finally, Figure A. 10 specifies the new production possibility curve
0 P'M that our model economy would face following a capital-using prod-
uction process change in X that shifts the production-efficiency locus
toward the origin 0. Figure A.10 is labelled identically to Figure A.9
and the derivation of the new production-possibility curve C^P'M would
precisely parallel the derivation given in the preceeding paragraph.
A-17
-------
>
00
o
x
INPUT of LABOR
o
OUTPUT of X
FIGURE A.10 SHIFT IN PRODUCTION-POSSIBILITY
CURVE WITH CAPITAL-USING
PRODUCTION PROCESS CHANGE
IN X TO REDUCE POLLUTION
-------
A.3 Technological External Diseconomies and Production Possibilities
Firms in a competitive industry may impose Pareto-relevant or techno-
logical external diseconomies on (1) other firms in the same industry and/
or (2) firms in other industries through their influence on technological
production relationships, as well as on (3) consumers, and thereby them-
selves and other firms, through their influence on the "state of the envi-
ronment." The effects of productive activity on the state of the environ-
ment is discussed in detail in the main text of this report. Here we
specify the direct impact of technological external diseconomies on prod-
uction possibilities. Although the technique presented here is new, it
applies our extension of Chiptnan's concept of "parametric externalities"
and the technique developed in the previous section to examine negative
technical progress.* Consideration is given to both output- and input-
generated external diseconomies.
Though other specifications of technologically relevant external dis-
economies are possible, our extension of Chipman's concept of parametric
external economies of scale to cover external diseconomies is particularly
convenient because a production possibility curve can be derived rather
easilv despite the comolexities created in simole economic models bv exter-
nal diseconomies. Our concept of external diseconomies is as follows. Each
entreprenuer in a competitive industry is assumed to act as if his firm has
constant returns to scale, and departures from this constant factor-output
relationship are regarded by him as the result of perturbations in his unit-
homogeneous production function (even where such disturbances are partly
caused by changes in his own firm's output and/or factor usage). Such shifts
in production functions are assumed to be related to the levels of aggregate
industry output and/or factor employment. From this viewpoint, parametric
external diseconomies generate negative technical progress within an indus-
try, or for some other industry (or industries), that is directly related to
its output and/or factor employment. Where previously we examined once-only
negative changes in technology, we now consider successively increasing neg-
ative technical changes directly related to its own or some other industry's
output and/or factor usage.
Consider first neutral parametric external diseconomies in the X indus-
try that depend on the output of X. Its effect would be to shift X isoquants
uniformly and proportionately farther and farther from their origin 0 in
Figure A.11 as the output of X increases. Since the externality is neutral,
the production-efficiency locus does not shift but the rays from 0 along
which outputs of X are measured do shift. In the case of a parametric ex-
ternal diseconomy, the appropriate measurement rays from 0 become success-
ively more steep as the output of X increases, e.g., from the original diag-
onal 00 to 0 M , 0 ML, and 0^1 successively. (Of course, outputs can be
measured along such rays in the usual fashion because a parametric external
diseconomy means that the typical firm's production function, and hence the
industry aggregate production function, exhibits constant returns to scale
at each given level of industry output.) Accordingly, in Figure A.11 factor
* John S. Chipman, External tieonomies of Scale and Competitive Equili-
brium, Quarterly Journal of Economics, Vol. LXXXIV, No. 3 (August 1970),
DD. 347-385.
pp
A-19
-------
ro
o
o
o
OUTPUT of X
X
-. X3X2 P
FIGURE A, 11
M3X3 M2 M1
O
Y
-------
combinations which would result in outputs OX , OX , and OX if industry
output were assumed to be zero (and output could be measured as usual along
the diagonal 00) would provide outputs OX' (measured along 0 M ), OX'
(measured along 0 M ), and OX' (measured along 0 M ) in the presence or a
neutral parametric external diseconomy. Consequently, our model economy's
production-possibility curve would be 0 P'P'P'P' in the presence of a para-
metric external diseconomy (where 0 P P P 0 would be the production-possi-
X -L 2. j Y
bility curve if production of X by the technology represented by isoquants
x1, x , and x did not generate external diseconomies for the X industry).
Consider next a situation where the production of X generates a neutral
parametric external diseconomy for industry Y that depends on the output of
X. In this instance, the external diseconomy shifts Y isoquants uniformly
and proportionately farther and farther from their origin 0 in Figure A.12
as the output of X increases but does not shift the production-efficiency
locus. Now, however, the appropriate measurement rays? from 0 become suc-
cessively less steep as the output of X increases, e.g., from the original
diagonal 00 to 0,31.. , 0 M , 0 M , and 0 1YL successively as the output of X
increases from zero to Ox , OX~, OX-, and OX,, respectively. Factor combi-
nations which would result in outputs OY , OY , OY , and OY. ±f_ X industry
output were zero (and output of Y could Be measured as usual along the
production box diagonal) would provide outputs OY' (measured along 0 M ),
OY' (measured along 0 M ), OY' (measured along 0 A ), and OY' (measures
along 0 M,) in the presence or a neutral parametric external diseconomy
generated by industry X. Therefore, 0 P'P'P'P'0 represents our model
economy's production possibility curve in the presence of an external dis-
economy produced by industry X, while 0 P P P P.O would be the production-
possibility curve ±f_ the production of A aia not generate external disecon-
omies in the production of Y by the technology represented by isoquants y ,
y2, y3, and y^.
Two important observations must be made before concluding this discus-
sion of a situation where the production of X creates an external diseconomy
in the production of Y.
(1) Changes in the production functions for X and Y, individually or jointly,
may moderate (if not entirely eliminate) the impact of the external disecon-
omy associated with present production technologies. Though industry X im-
poses diseconomies on industry Y, the appropriate control method is not
obvious. This point is treated more elaborately in the main text of this
report.
(2) Where industry X generates diseconomies for industry Y, such diseconomies
can (though not necessarily) result in production-possibility curves that
are not entirely concave with respect to the output space origin, e.g.,
Figure A.12 provides an illustration of a situation where an external dis-
economy generated by industry X leads to a convex production-possibility
curve at relatively low outputs of Y. Therefore, phenomena typically associ-
ated with increasing returns to scale may obtain in situations where external
diseconomies exist.*
* It should also be noted that if industries X and Y were both postulated
to generate parametric external diseconomies affecting one another, the produc-
tion possibility curve could conceivably exhibit reversals in its usual curva-
ture in the neighborhood of both ends.
A-21
-------
o
ro
Q.
O
OUTPUT of X
A.12
-------
To this point attention has centered on output-generated parametric
external diseconomies. The expositional method has relied on the device
of shifting measurement rays as X industry output changed. It seems ob-
vious that one could also adapt this same device to deal with input-gener-
ated external diseconomies: measurement rays would then be shifted as X
industry employment of capital, labor, or Y (also an intermediate good in
our complete model) changed. Since there would appear to be little advan-
tage to a considerable repitition of our previous discussion, we content
ourselves with the observation that the relationship between input employ-
ment (or, for that matter, output) and the magnitude of its associated
diseconomies, and therefore its associated output reductions, can conceiv-
ably be quite complex. Under various "reasonable" assumptions, there would
appear to be little difficulty in obtaining production-possibility curves
with alternate concave and convex ranges when the magnitude of external
diseconomies varies with factor employment levels and combinations.
Furthermore, to this point we have also assumed that external disecon-
omies have negative impacts at all levels of output and input usage. Of
course, in many instances a certain "threshold" output or input use must be
exceeded before an externality's effects have a discernible impact. Under
such circumstances, once again only a minor adjustment in the technique
developed previously is required to derive the production possibility curve.
Where industry X produces an external diseconomy affecting itself, but only
once some minimum output is exceeded, the measurement ray for X would be
the production-box 'diagonal at outputs equal or less than the relevant
"threshold" output, and the point on the diagonal above the threshold out-
put would serve as the origin of the new measurement rays associated with
output levels greater than the threshold output. In contrast, where indus-
try X generates £.n external diseconomy for industry Y after the output of
X exceeds some threshold level, there would be no change in the origin of
the new measurement rays, and new rays would replace the diagonal for mea-
surement purposes only after the output of X exceeded its threshold level.
Consequently, the existence of thresholds poses no substantive problems in
the derivation of production possibility curves.
Similarly, we do not here present the derivation of production-possibil-
ity curves for either capital-using or labor-using parametric external dis-
economies. Such diseconomies would require us to construct diagrams much
as before but for situations where both the production-efficiency locus and
appropriate associated measurement rays would simultaneously shift as output
and/or factor employment changed. Again, there would appear to be many
"reasonable" cases, some of which would result in production-possibility
curves indistinguishable from those previously derived and some of which
would have alternate concave and convex ranges. Diminishing returns set in
almost immediately in this sort of situation, and therefore we do not pursue
the derivation of production-possibility curves under conditions of parametric
external diseconomies any further here.
A-23
-------
APPENDIX B
A GENERAL EQUILIBRIUM ANALYSIS OF
PRODUCTION PROCESS REGULATION
Regional general equilibrium models that explicitly allow for inter-
industry flows of materials often provide the most appropriate conceptual
framework within which to investigate the economic consequences of alter-
native methods of controlling pollution. This is the case because major
polluting industries are often regionally concentrated and provide a sub-
stantial fraction of the region's income by exporting some intermediate
and/or final good to the "rest-of-the-world." Of course, no originality
can be claimed for this observation. In fact, the Air Pollution Control
Office (APCO) of the Environmental Protection Agency (EPA) recognized the
need to focus on regional economies a number of years ago, and it commis-
sioned CONSAD Research Corporation, under subcontract to TRW Systems, Inc.,
for the development and demonstration of a regional general equilibrium
model as an operational analytical tool for evaluation of alternative pol-
lution abatement policies. Although further refinements and new (now.un-
available) data are required to make the CONSAD model system as operational
as APCO presumably desires, the potential benefits of the CONSAD regional
economic model system have been demonstrated—at least in part. At this
time, however, despite the investigation of the effects of three air pol-
lution control strategies in the CONSAD model, the regional and general
equilibrium effects of air pollution control policies do not appear to be
fully understood.
This appendix explores the positive economics of production process
regulation to achieve pollution-control in simple models,of closed and open
regional economies with interindustry flows of materials. Viewed most
broadly, the analysis is intended to acquaint the reader with the uncer-
tainties associated with pollution abatement in the real world. Viewed
slightly more narrowly, the results of the analysis for closed economies
are relevant to the evaluation of national pollution-control regulations
for particular industries. The results obtained for regional economies
are important for the design of regulations affecting a region's export
industries. In both cases, general equilibrium analysis allowing for inter-
industry flows of materials demonstrates that, a priori, many of the "obvi-
ous" effects of industry-wide pollution-control regulations are quite un-
certain. Therefore, empirical investigations of specific regulations for
particular industries in certain region(s) are required to determine the
qualitative, as well as the quantitative, effects of various pollution-
control policies.
The analysis is intended (a) to extend the theoretical literature on
pollution-control to cover industry-wide regulations (as opposed to those
affecting individual firms or small segments of an industry) and (b) to
demonstrate hitherto-neglected aspects of pollution control. Throughout,
the analysis assumes perfectly competitive product and factor markets and
ignores the impact of regulation-induced factor price and income distribu-
tion changes on product demand.
B-l
-------
B.I Analysis in Model of Closed Economy
Consider the simplest possible general equilibrium model with inter-
industry flows of materials: a model with two outputs, XL and X2, each of
which employs labor and the output of the other industry as inputs. Formally,
Xl ' fl (X21'V
X2 * f2 (X12'V
Lo = Ll + L2
where X.. indicates the quantity of the ith commodity used by the jth indus-
try, L.^the amount of labor employed by the jth industry., and LQ is some
constant. As usual, f, and f^ are assumed to be continuous, concave func-
tions, which have continous first partial derivatives and are homogeneous
of the first degree. These relations and assumptions are sufficient to
allow derivation of the customary transformation, or production possibility,
curve.
In figure B.I outputs are measured in a positive direction and inputs
are measured in a negative direction. The functions f^* and f2* represent
f with L, = L and f2 with L2=Lo' resPectively- They are total product
curves where production at any point uses all available labor. Consequently,
if a point on f2* were selected, there is no production of commodity 1 except
as an input in the production of commodity 2. Point B is a possible prod-
uction point only with the production and trade of OG of commodity 1 as an
input of commodity 2.
If A and F are the maximums of f * and f-,*, respectively, then all prod-
uction points on ABEF, where BE is drawn tangent to f,* and fo*9 are possible
and efficient with trade allowed. However, for this model of a closed econ-
omy, the production possibility curve is CD. All efficient production points
can be achieved with production activities given by the- slope of OB and OE,
despite the availability of many other production coefficients. Figure B.I
is therefore a geometric picture of Samuelson's Substitution Theorem.
We are now ready to employ Figures B.2, B.3, and B.4 to illustrate
several important, but previously neglected, aspects of production-process
regulation to obtain pollution reduction.
(1) Suppose, for example, that the Pollution Control Authority bans the
procuction process f2 and the next most efficient process is labor using,
i.e., for any X less X2 can be produced. In Figure B.2, this shifts f2*
downward, result's in a new lower tangency line, and shifts the production
possibility curve CD downward. Regulation of production processes in in-
dustry 2 reduces the maximum possible output of both industries because
industry 1 uses the output of industry 2 as an input. Moreover, in this
example the new production possibility curve slopes less steeply than CD,
B-2
-------
FIGURE B.I
B
, **
FIGURE B.3
21
FIGURE B.4
21
B-3
-------
and the decrease in the maximum possible output of industry 2 is relatively
greater than the decrease in industry 1. Consequently, the relative price
of commodity 2 must increase.
(2) This is not the case, however, where (a) the production of commodity 1
requires no labor (see h,* in Figure B.3) or (b) the asymptote to its total-
product curve (see the broken line asymptotic to g.^* in Figure B.3) inter-
sects the total-product curve for X before and after the introduction of
regulation. In such cases, regulation shifts the production possibility
curve from CD to C'D', reduces the maximum possible output of both industries
the same percentage, leaves the slope of the production possibility curve
unchanged, and relative prices remain the same.
(3) Finally, consider a regulated production process change such that the
same output of commodity must be produced with more labor and less commodity
1. Two possibilities exist. Either the total product curve f2* shifts to
a new tangency on BD at B', or it shifts downward and rightward with results
identical to those already discussed. Of course, with the new production
process OB' in Figure B.4, no decrease in either output is necessary, and
the production possibility curve CD remains unchanged. Thus, production
process regulation to achieve pollution reduction does not necessarily de-
crease the productive capacity of a fully-employed economy.
B.2 • Analysis in Regional Model
Consider next the simplest general equilibrium, regional model with
trade in intermediate products: a model with two regions, A and B, producing
two outputs, X-^ and X2, each of which requires labor and the other commodity
as inputs. Without trade or inter-regional labor migration, the productive
capacity of each region is defined by
XA = g (X ,LA)
1 2 12 2
TA _ TA . TA
o " Ll + L2
and
(X21'L1>
XB = e (X ,LB)
2 2 12 2
LB - LB + LB
Lo ~ Ll + L2
B-4
-------
where the relations among variables and the assumptions are the same as
those postulated previously.
With trade and no labor migration, each region would have a set of
possible and efficient production points similar to ABEF in Figure B.I.
Of course, the entire production possibility set of each region would be
given by a geometric figure similar to OABEF, say, 0 ABEF for region A
and 0 KLMN for region B (neither shown here). Given these conditions,
there are four types of production possibility curves that are possible
for the nation. These four types of national curves are obtained by
"sliding" 0 ABEFat point 0 along KLMN on 0 KLMN, terminating the sliding
where (i) tne slope of AB would become less than the slope of LM and (ii)
the slope of EF would become greater than the slope of LM, and they are
presented in Figure B.5. Regional differences in technologies, labor
skills, and/or labor endowments generate the different shapes and sizes
for 0 ABEF and 0 KLMN that result in such diverse production possibility
curves for the nation.
Three observations are appropriate before investigating the effects
of regulation in this model. First, and perhaps theoretically most inter-
esting, Samuelson's Substitution Theorem does not generally apply to a
nation with trade in intermediate and final products among regions despite
the fact the theorem holds for each region without trade. The reason for
this result is that trade permits regions to produce outputs impossible
without trade. Second, the production possibility curve for the nation is
not obtained by simply adding the loci of its respective regions; negative
outputs of the same commodity are economically impossible for all regions.
Third, ruling out corner solutions, substantial differences in regional
endowments can mean that Samuelson's Substitution Theorem does hold for the
nation. In such circumstances, relative prices and production coefficients
are determined by a single region—the "dominant" region. Figure B.5.depicts
such a situation.
Turning now to production process regulation to obtain pollution re-
duction, we initially ignore regulation-induced labor migration to discover
how trade alters the conclusions obtained for closed economies.
(1) Suppose, as before, the Pollution Control Authority of Region A bans
the production process g? and the next most efficient, permissible process
is labor-using. .If A is the dominant region before and after introduction
of regulation, trade in intermediate and final product among regions alters
none of the conclusions obtained previously; they also hold for the nation
as a whole. The relative price of commodity 1 may remain unchanged or de-
crease, and productive capacity may decrease uniformly, differentially or
not at all.
(2) Of course, if Region A were dominant and the Pollution Control Authority
of Region B banned production process e_, then relative prices would remain
unchanged but the national production possibility curve, PP1, would shift
toward 0 with no change in its slope.
B-5
-------
'B (a)
Vl "B (b)
(c)
FIGURE B. 5
RGURE B.6
FIGURE B.7
B-6
-------
(3) Where the Substitution Theorem does not hold for the nation, however,
the effects of regional regulation are less certain because consumer demands,
regional production possibilities, and regional size jointly determine the
outcome. One example will suffice to demonstrate this. In Figure B.6,
production possibilities are PP' before regulation, and the nation operates
at point Q. Suppose that regulation shifts production possibilities from
PP' to RR', i.e., the Substitution-Theorem holds for the nation after regu-
lation is introduced. Ruling out the indeterminate corner cases on RRr ,
the relative price of commodity 1 would increase: the slope of RR1 is less
than the slope of PP' at Q. Consequently, with trade and no labor migration,
production process regulation in one region can result in an increase, de-
crease, or no change in the relative price of the regulated commodity. Of
course as always, regulation would decrease the production possibilities of
the "regulating" region and the nation. Without migration, total and per
capita national income and income received in the regulating region decline.
In a genuine regional model, however, factors of production as well as
final products are mobile. Therefore, we now relax the assumption that there
is no labor migration in response to regulation-induced income differentials.
Interestingly, labor mobility alters previous results only slightly. Provided
the production function that replaces the one banned by the regional Pollu-
tion Control Authority is continuous, concave and homogeneous of the first
degree, migration would eliminate regional per capita income differences with-
out all labor in the regulating region migrating to the unregulated region.
Diminishing returns assure this result. Consequently, labor migration fol-
lowing regulation would reduce the regulating region's production possibili-
ties, expand those of other regions, and disperses the income reduction of
regional pollution-regulation throughout the nation.
Figure B.7 neatly depicts the impact of different assumptions concerning
labor mobility. As before, regulation in one region would shift national
production possibilities from PP' to RR1 with no inter-regional movement of
labor. However, if labor is mobile and responds to wage differences, national
production possibilities would shift from PP' to MM' rather than RR'. Other-
wise, none of the earlier predictions are altered. Factor mobility merely
moderates the income-reducing effects of pollution controls; it does not
alter the a_ priori indeterminacy of the impact of pollution controls on
relative commodity prices.
B-7
-------
APPENDIX C
PRIVATE PRODUCTION POSSIBILITIES AND EFFICIENT
GOVERNMENT PRODUCTION OF PUBLIC GOODS
This appendix examines some of the implications of the suboptimization
technique frequently recommended by management consultants, cost-effective
specialists, and benefit-cost analysts in their reports to public agencies.
One of their most common recommendations involves the application of tradi-
tional economic efficiency criteria to the design and operation of govern-
ment facilities to produce public-good service outputs in amounts which may
or may not approximate an unknown optimum. To be precise, they suggest
that the marginal rate of technical substitution between factors in public
production be the same as that obtaining in private production. (Of course,
this suggestion is most often expressed in less technical language.) The
implications of this efficiency rule are typically either regarded as ob-
bious or the reader is referred to a standard microeconomic theory textbook
where efficiency in private good production is demonstrated to require
satisfaction of this rule. However, where the amount of a good to be pro-
duced is determined by, say, legislative or executive action instead of
consumer demand and, therefore, may or may not approximate some unknown
optimum, an additional constraint has been introduced, and the general theory
of the second best would suggest that application of the recommended effi-
ciency rule may have unsuspected and unintended results. For this reason,
this appendix reports a general equilibrium analysis of the impact on pri-
vate production-possibilities of the recommended efficiency rule for govern-
ment production of a public good. The analysis shows that "efficient"
production of some given finite public good output implies a petal-shaped
transformation curve for private outputs rather than the usual convex, con-
tinuously negatively-sloped transformation curve typically postulated in
the literature. The appendix concludes by discussing some implications of
this finding.
C.I. Mathematical Analysis
Assume three Cobb-Douglas production functions
y - T R 1~a m
A - L K (L)
A A
Y = LY R^"6 (2)
G = LG RQL~y (3)
where L. equals labor employed and R. is the capital-labor ratio in the
relevant sector. Factor supplies are assumed to be fixed
C-l
-------
Lv Rv + LY Rv + Lr Rr = K
XX II >j vi
where K and L are the fixed endowments of capital and labor. These equations
can be rewritten as
gRG + (l-g)R = R (6>
hR]( + U-h)Ry = R (7)
where g is the proportion of the total labor endowment employed by the govern-
ment and h is the proportion of privately employed labor allocated to the
production of X.
Following Hary G. Johnson (Econometrica, July 1966), with no government
L , R , and g equal zero, while Rr> equals R. Assuming competitive conditions
in private markets, long-run equilibrium implies that the marginal rates of
technical substitution between factors are the same in industry X and Y, i.e.,
<* j, = S R (8)
and, therefore, using (7)
R = bR (9)
X a + (b-a)h
(10)
V /K NU
Y a + (b-a)h
where a = -^— and b = T-^T-. Taking advantage of (9) and (10), (1) and (2) can
i . . 1—Ot -L—P
be rewritten as
X = (h)(L) I bR
bR
a + (b-a)h
f aR
1-a
1-B
(11)
(12)
a + (b-a)h
Differentially (11) and (12) with respect to h, Johnson obtained
dX
dX _ dh _ /a-a , nxB-l, , ,. <*-& a + h(b-a)
(aR) (a
dY - - a + (b-a
dh
Thus, the normal transformation curve is negatively sloped.
C-2
-------
Efficient government production of public goods would appear to most
economists to require that the marginal rate of technical substitution
between factors in public production be the same as that obtaining in
private production, or
a R = b R = c Rp (14)
A l (j
Y
where c = —-J—. Postulating efficient government production of some finite
public good output G in precisely the sense of (14), we now seek to deter-
mine whether the transformation curve for private outputs X and Y is con-
tinuously negatively sloped.
Given that (14) means
J RG and RY = £ RG,
production functions (11) and (12) can be rewritten as
- - Yl c R )1~a (15)
X = h (L - GR Y ) (- V
o 3.
Y = (1-h) (L - GRGY ~1) (£ Rg)1'3 (16)
— -
where privately available and employed labor is L - GR . Differentiating
(15) and (16) with respect to R and performing some minor algebraic manipu-
lation, we obtain
dX
^ - -^ - f-JU fc/-° (c/-1 R 3-a
dY dY *• l-hj (a} S; G
dRG
(l-a)L-(Y -c
(lH3)L-(Y-e)GR Y -1
(17)
Since the sign of the bracketed term in (17) varies with R if ot^P, tlie slope
of transformation curve between X and Y given efficient production of G as
defined by (14) also varies with R . Consequently, the transformation curve
for private outputs under the assumed conditions is petal-shaped rather than
continuously negatively-sloped.
C.2. Discussion and Implications
The reason for the foregoing, previously unsuspected result is that ef-
ficient production of a given public good output introduces three additional
constraints into the standard two-good-two-factor general equilibrium model.
These additional constraints are (1) the given level of public good output,
(2) the production function for the public good, and (3) the "efficiency
rule" requiring that the marginal rate of technical substitution between
factors in public good production equal the marginal rate of substitution
in private production. Of course, once the additional constraints are ex-
plicitly stated and recognized, the general theory of the second best would
immediately suggest to most economists that relationships which held prior
C-3
-------
to the introduction of the constraints may be subject to serious qualifica
tion. Thus, in an important sense, our result cannot be regarded as partic-
ularly surprising.
On the other hand, this finding does have important implications for
public policy. Some of these implications may be illustrated most clearly
with the aid of Figure C.I, which presents two transformation curves.
OMZ*NO represents the petal-shaped private good transformation curve that
would obtain if the government followed the usually recommended efficiency
rule in producing some given amount of public good output, G, while P*Z*P*
represents the maximum attainable combinations of private good outputs X
and Y such that public good output equals G. (The marginal rate of techni-
cal substitution between factors in public production would equal that in
private production at all points on OMZ*NO but only at point Z* on P*Z*P*.)
Figure C.I.
P*
Z*
P*
X
Following the often-recommended efficiency rule would minimize the
private output reductions associated with public good production if and only
if demands for private goods would result in the production of the Z* output
combination. If private demands would lead to the production of any output
combination other than Z* on OMZ*NO, then the government would only appear
to minimize the cost of public good output by equating its marginal rate of
technical substitution between factors with that rate of substitution ob-
taining in the private sector. In fact, by following the usual efficiency
rule the government would impose a substantial unnecessary constraint on
the private sector: the output space enclosed by OMZ*NO is considerably
less than that enclosed by OP*Z*P*0.
C-4
-------
Our major conclusion is obvious. Genuine suboptimization where some
given output of pure public goods is produced will typically require delib-
erate violation of economists' usual cost minimization rule. Determination
of the appropriate factor combination for truly efficient production of non-
market-determined public good outputs is considerably more complex than has
previously been recognized.
C-5
-------
APPENDIX D
THE POLLUTION PRODUCTION FUNCTION
D.I. The Importance of the Pollution Production Function
An important fraction of our continuing research effort has been de-
voted to the specification and estimation of the pollution production
functions associated with open burning of agricultural fields in the Wil-
lamette Valley. Open field burning produces smoke that, depending upon
atmospheric conditions, is transported to various points in the Valley or
out of the Valley. To economists, the problem of accounting for the impact
of this burning process on visibility appeared analogous to the problem in
economics of estimating what is called a 'production function1. A produc-
tion function is an analytical concept that asserts the existence of a set
of fixed relationships between an output or set of outputs on the one hand
and a set of inputs on the other. For economists, the problem of estimating
production functions has been confined to providing adequate specification
of the reasonable independent variables in the process and ascertaining, a_
priori, what mathematical form this production process might reasonably
take. Economists have had some success in the empirical estimation of par-
ticular production functions.*
In our research, a pollution production function has been regarded as
an analytical concept that postulates the existence of a set of stable
relationships between daylight visibility in particular locations in the
Valley under varying present and recently past meteorological conditions
and recently past emission rates throughout the Valley. Meteorological
conditions determine how the atmosphere transports emissions from field
burning from one location to another in the Valley, and acres burned de-
termine emission rates and atmospheric loading in different zones of the
Valley. Meteorologists have described the Willamette Valley as a box with
a movable lid and a set of ventilating windows-** Alternative specifica-
tions of relevant meteorological variables consistent with that model have
been explored in our research. Empirical estimation has concentrated on
functions with minimum daylight visibility at Eugene and Salem as the
dependent variable and acres-burned among the independent variables.
There are two reasons why our research initially and continuously has
focused quite intensively on estimation of the pollution production function.
First, research success promised to provide immediately valuable information
for the Oregon Department of Environmental Quality in the development of its
field burning control policy to minimize the harmful effects associated with
* For example, see Marc Nerlove, Estimation and Identification of
Cobb-Douglas Production Functions, Chicago: Rand McNaliy, 1965.
** Olsson, Lars E. and Wesley L. Tuft, A Study of the Natural Ven-
tilation of the Columbia-Willamette Valleys, Technical Report No. 70-6,
Department of Atmospheric Sciences, Oregon State University, Corvallis,
Oregon, June 1970, pp. 133-137.
D-l
-------
smoke produced by open burning of agricultural fields. For precisely this
reason, Bruce Snyder, the meteorologist with the DEQ, and faculty members
of the Department of Atmospheric Sciences at Oregon State University have
generously supplied data and advice throughout our research. Second, and
most important given our research objectives, measurement of the benefits
to be obtained from various field burning control policies is impossible
without the pollution production function. This is the case because the
pollution production function is a major determinant of the marginal valua-
tion schedule of air quality improvements from field burning control.
Assuming known pollution production functions and a known (constant
or declining) marginal value schedule for successive improvements in air
quality, the above point may be established with the aid of Figure D.I,
where an (unspecified) index of air quality is measured along the vertical
axis, while acres-burned and the marginal value of successive air quality
improvements are measured horizontally in panels (a) and (b), respectively.
In panel (a), curves MR and MU illustrate hypothetical relationships
derived from pollution production functions estimated under conditions
where burning is and is not regulated according to prevailing and forecast
air quality and meteorological, conditions. In panel (b), the marginal
valuations of successive improvements in air quality are plotted with and
without well-regulated burning, and mr and mu are associated with MR and MU,
respectively.
Air
Quality
Figure D.I.
Air
Quality
u
Acres
Burned
(a)
(b)
D-2
-------
Following standard benefit measurement techniques, the gross benefits
of field burning regulated according to prevailing and forecast air quality
and meteorological conditions would be measured by the area rcud with no
reduction in total acres burned. Furthermore, since the benefits rcud
are obtainable with no reduction in total acres burned, the maximum benefits
from a complete-ban of field burning would be measured by the area mar,
whereas regulated burning coupled with an acres-burned reduction equal to
PQ would result in benefits measured by ber. Consequently, correct esti-
mation of the benefits derived from all field burning control programs would
be quite impossible without the pollution production function.
D.2. The Model
The basic hypothesis of our model is that daylight visibility is a
function of prevailing meteorological conditions and the load of contam-
inants placed in the atmosphere by polluters. However, since atmospheric
load at any given location and time is itself a function of present and
recently past contaminant emission rates at that and other locations, a
more complete and accurate statement of the hypothesis would be as follows.
Daylight visibility at a given location is a function of (1) the present
and recently past air-contaminant emission rates at that and other loca-
tions and (2) the present and recently past capacity of the atmosphere to
dilute and transport its contaminants from one location to another.
Operational specification of this hypothesis is no simple task.
Contaminants are emitted at non-uniform rates from numerous sources for
which data are limited or not available while atmospheric dispersion and
transportation processes are complex and meteorological data above the
surface are obtained only twice per day for the Willamette Valley^
Consequently, for statistical testing, daily variations in minimum daily
visibility are hypothesized to be "explained" by variations in different
sets of independent variables reflecting air-contaminant emission rates
and the atmospheric transportation process. Alternative specifications
of the present and past emission rates (via various proxy variables) and
the capacity of the atmosphere to disperse and transport contaminants
require discussion, however, prior to the presentation of our results.
Although atmospheric load is the result of emissions from numerous
sources, the primary concern here is with the atmospheric loading effect
of field burning on visibility. Therefore, the research is designed to
test a strong hypothesis about the visibility-field burning relationship.
This is accomplished by employing (a) a theoretically reasonable data
configuration, (b) an explicit hypothesis concerning emissions from other
than field burning sources for which data are not now systematically
collected, and (c) proxy variables for daily field burning emission rates
and background load before fields are burned on any given day. This pro-
cedure means that field burning is treated as an important marginal con-
tributor to atmospheric loading. Variations in field burning emission
rates together with variations in the background load and emissions from
sources other than field burning are regarded as jointly determining
variations in atmospheric loading.
D-3
-------
Regarding the selected data configuration, data are excluded for
days on which field burning was less than 100 acres in total. Although
the decision to confine consideration to 100-plus acre burning days is
admittedly arbitrary, it does not appear unreasonable.
Unfortunately, despite the importance of automobiles, pulp mills,
etc. as sources of air contaminants, no daily data are systematically
collected on emissions from these sources. Without these data, it is
hypothesized that daily emissions from these sources are regular and stable
and that, therefore, they can be treated analytically as constant.
Regarding field-burning emission rates, the number of acres burned
throughout, or in particular zones of, the Willamette Valley serve as a
proxy for field-burning emissions. Data are available on the acreage of
total field burning permits issued daily by fire district in 1969 and 1970,
and these data can be aggregated to reflect the burning in various zones
of the Valley or in the Valley as a whole. In contrast, however, there is
no need to identify or measure the sources of the varying daily initial
background load existing prior to each day's field burning. This is be-
cause background load is appropriately treated as a predetermined variable
(i.e., determined by emissions from many sources on the previous day(s) as
well as past meteorological conditions). Here, visibility at 7 a.m. at
the location under study provides the proxy for background load.
Turning next to the capacity of the atmosphere to dilute and trans-
port its load, the variables that seemed most appropriate were mid-
afternoon inversion height, atmospheric stability, wind speed, and wind
direction. Although we had access to the full range of meteorological
data, we deliberately did not seek to use it all. In our view, whatever
model we hypothesized and tested, it was essential to minimize colinearity
among the independent vaiables. For this reason other variables generally
associated with atmospheric stability (such as surface temperature, rela-
tive humidity counts, etc.) were excluded.
Regarding the empirical data, several problems deserve note. The
data on inversion height were measured in millibars of pressure and used
in this form. Hence, the relationship between inversion height and mini-
mum daylight visibility was predicted to be negative: the higher the
measured inversion pressure, other things equal, the lower would be visi-
bility. We recognize one shortcoming of the analysis in that no donsid-
eration was given to the thickness of the inversion layer.
Utilization of a stability measure that was theoretically plausible
and empirically meaningful was difficult. Daily data .are available on
air temperature changes between the surface and 900 millibars, 800 and 900
millibars, and 700 and 800 millibars. These air temperature change data
are measures of stability. For purposes of model specification, the prob-
lem was to select a particular band, a combination, or use all three bands
as stability proxies. The appropriate measure or measures of stability
should reasonably be confined to the mixing layer. After considering the
direction of the prevailing winds and the height of the Cascade Range, it
was decided to use the air temperature change data in the surf ace,, to 900
millibar range as the measure of stability.
D-4
-------
The character of the physical problem being considered suggested that
wind direction must play a prominent role in the analysis. The problem was
how to treat it. The method of measuring wind direction makes the data
difficult to use if one attempts to employ continuous variables. Two
schemes were considered which relied on arbitrary designation of winds from
the north, south, east, and west. One of these schemes employed a weight-
ing arrangement for simultaneously taking account of both wind speed and
direction. Given the measurement and data problems and given the predicted
strong relationship between wind direction and low visibility in Eugene and
Salem, it was decided to take wind direction into account by a data sorting
technique. That is, winds from the north could potentially contribute to
deterioration of the visibility at Eugene and Salem. On the other hand,
given the locations of Eugene and Salem, winds from the south should improve
visibility. Therefore, it was decided to investigate (a) visibility in
Eugene only on days with winds from the north and (b) visibility at Salem
first, on days with north winds and, second, on days with southerly winds.
Reference has been made to the "relevant wind". Wind measurements
are available at both the surface and at 850 millibars. Meaningful analy-
sis required that a decision be made as to which measure to use. The rele-
vant data are clearly those that apply to the mixing layer. If the inver-
sion height was lower than 850 millibars (i.e., more than 850 millibars of
pressure) the only relevant wind data on speed and direction was provided
by the surface measurements. On the other hand, if the inversion pressure
was above 850 millibars it seemed reasonable that measurements of wind
direction and speed at 850 millibars would be utilized in the analysis.
We adopted this as a selection rule for determining which measurements of
wind speed and direction were considered for each day. Furthermore, in
the model variants where wind speed was treated as an independent variable,
the data were expressed in knots per hour.
Specification of the empirical data which could reasonably serve as
proxies for the theoretical determinants of minimum daylight visibility
was but one portion of the model specification problem. Completion of this
problem required specification of a theoretically plausible mathematical
form. Adopting the principle of "Occam's Razor" as a guide, first, we
tested a linear version of the model. Next we examined a form in which the
independent variables were treated as multiplicative and the function as
exponential. This form was plausible on physical grounds as well as having
achieved noteworthy success when adopted for estimating economic production
functions. For statistical purposes, the multiplicative-exponential form
meant estimation of equations linear in the logarithms of all variables.
D-5
-------
V
V
V
V
V
- Vz (A, V?A)
= VII (A' V7A' V
= VIII (A' V7A' X4P' WS
' VIV (A' V7A' Z4P' S)
= VV (A' V7A' X4P> WS'
)
S)
D.3. Statistical Results*
Five variants of our basic model have been tested for Eugene and Salem
on 1969, 1970, and 1971 data for days on which at least 100 acres burned.
These variants differ from one another according to their respective speci-
fications of the independent variables reflecting atmospheric capacity to
dilute and transport load. The model variants, expressed in general form,
are as follows:
i
I.
II-
HI.
IV.
V.
where V = minimum afternoon daylight visibility at the Eugene or Salem
airport, A = total acres burned in the Willamette Valley or total acres burned
north and/or south of the Salem airport according to the relevant wind
direction at 4 p.m., Vy^ = visibility at 7 a.m. at the Eugene or Salem
airport, I^p = inversion height at 4 p.m., Wg = relevant wind speed at
4 p.m., and S = atmospheric stability at 4 p.m.
As mentioned above, the mathematical forms assumed for purposes of
analysis were the linear and lot - log forms. Interestingly, the estima-
ted relationships and the statistical significance of the regression
coefficients did not vary dramatically between functional forms. However,
for all models the fits in terms of both Rr and the F-statistic were clearly
better when the log - log form was employed. Consequently, this appendix
reports regression results for models in log - log form only.
< The multiple regression results obtained for each model variant, as
well as the beta coefficients of each regression coefficient, are reported
in Tables D,l through D. 15. (Since the size of estimated regression
coefficients varies with the units in which each independent variable is
stated, regression coefficients do not indicate the importance of the
independent variables in statistically explaining variations in minimum
* The results reported in this section extend and superside those
presented in our second annual report (dated June 1971) and our paper
"Field Burning and Visibility Reduction in the Willamette Valley: A
Pollution Production Function", which was given at the Eighth Annual
Meeting, Pacific Northwest International Section, Air Pollution Control
Association, Spokane, Washington, November 18, 1970.
D-6
-------
afternoon daylight visibility in Eugene or Salem. Such information can be
deduced from beta coefficients, however. Beta coefficients convert regres-
sion coefficients into standardized units. Here, the beta coefficients
measure the effect on minimum daylight visibility of typical changes in the
independent variables in terms of sample standard deviations.) Tables D.I
through D.10 present regression results for Eugene and Salem with winds from
the north in 1969-70, 1971 and 1969-70-71, while Tables D.ll through D.15
report our regression findings for Salem under southerly wind conditions
during 1969-70, 1970-71, and 1969-70-71.
Viewed most broadly, the regression results reported in Tables D.I
through D.15 are noteworthy on three counts. First, the results for 1969-
70 and 1969-70-71 support the predictions of our simple atmospheric models
reasonably well. Second, the rather poor performance of the same models
in 1971 directly derives from the burning regulations applied by the Oregon
Department of Environmental Quality, Air Quality Control Division. Third,
despite significant changes in burning regulations over the 1969-70-71
period, the regression coefficients obtained for the acres burned variables
are remarkable stable in size. Since acres burned are subject to direct
control, this finding has considerable importance for the design of smoke
management and field burning control policies. The evidence to support
these judgments follows.
Consider first our results for Eugene and Salem under northerly wind
conditions in 1969-70 and 1969-70-71. For these periods, all statistically
significant and insignificant variables have their theoretically anticipated
signs and therefore support the simple atmospheric models postulated here.
Minimum daylight visibility varies inversely with relevant (i.e., north and/
or south) acres burned and atmospheric stability and directly with visibility
before burning, mid-afternoon inversion height (recall that inversion heights
are measured here in millibars of pressure), and relevant wind speed. More-
over, only the statistically insignificant regression coefficient for atmos-
pheric stability changes substantially in size and relative importance when
one examines alternative specifications of the atmospheric load transpor-
tation and dilution mechanism for the same time period or between the two
time periods. Otherwise, alternative specifications of our models hardly
change the regression coefficients for relevant acres burned, visibility
before burning, inversion height, or wind speed. This stability in the size
of regression coefficients is particularly important because the specifica-
tion of the log - log functional form means that the estimated coefficient
provides a direct measure of the elasticity of minimum visibility with respect
to the various independent variables.* For example, the estimated regression
* The elasticity of a dependent variable with respect to some indepen-
dent variable indicates the responsiveness of the former to changes in the
latter and is measured by the percentage change in the dependent variable
divided by the percentage change in the independent variable, all other
independent variables being held constant.
D-7
-------
coefficient for South-Valley-acres-burned in Table I).2 indicates that a JO
percent change (above or below the 1969-70-71 average) would have meant an
opposite 3-3.5 percent change in Eugene minimum visibility in 1969-70-71
period, holding all other variables constant.
Furthermore, with few exceptions, it is encouraging to note that the
most important independent variables in terms of their respective beta
coefficients are the statistically significant acres burned, visibility
before burning, and inversion height variables. Since statistically sig-
nificant regression coefficients are not always associated with relatively
large beta coefficients, this evidence suggests that model variants II and
III are superior to variants I, IV, and V in terms of statistical explana-
tory power as well as in terms of goodness of fit as indicated by their
respective R^ and F-statistics. Consequently, our regression results pro-
vide support for the hypothesis that daylight visibility in the Willamette
Valley is a function of the load of contaminants placed in the atmosphere
by polluters and the prevailing meteorological conditions. This conclusion
holds with equal force for Eugene and Salem.
For Eugene, when winds were from the north, a statistically signifi-
cant relationship obtained (see Table D. 1, 2, 3, 4, and 5) between varia-
tions in minimum daylight visibility and relevant acres burned, as well as
other independent variables reflecting air-contaminant emission rates and
the atmospheric transportation process. These findings clearly establish
the impact of field burning smoke on visibility at Eugene. Furthermore,
these findings support the fundamental correctness of the 1970 and 1971
Department of Environmental Quality field burning program and schedule to
reduce deterioration of visibility in Eugene by curtailing acres burned
under north winds. Also, Tables D.I through D.5 corroborate the DEQ find-
ing that smoke from North Valley burning was an important determinant of
visibility in Eugene in 1969 and 1970.*
In the case of Salem under northerly wind conditions, however, consider-
ably less robust results were obtained for each of our five models. A
comparison of Tables D.6-D.10 with Tables D.1-D.5 indicates that the good-
ness-of-fit as measured by R^ and F-statistics is less in all time periods
for every model for Salem than for Eugene. Although all Salem regression
coefficients have their theoretically anticipated signs in the 1969-70 and
1969-70-71 periods, a comparison of beta coefficients indicates that the
relative importance of visibility at 7 a.m. (an index of background load
before burning) is greater for Salem than Eugene, while mid-afternoon in-
version height is less important at Salem. Nonetheless, the statistical
significance and importance of acres-burned in the northern half of the
Valley is clear; variations in acres-burned contribute to variations in
minimum visibility.
* Department of Environmental Quality, Air Quality Control Division,
"Field Burning in the Willamette Valley, 1970," staff report dated April
8, 1971, p. 14.
D-8
-------
Finally, for Salem under southerly wind conditions (see Tables D.ll,
12, 13, 14, and 15), a statistically important relationship obtained over
the 1969-70-71 period between minimum daylight visibility and acres burned
in the South Willamette Valley, visibility before burning in Salem, and the
mid-afternoon inversion height at Salem. Other indexes of atmospheric load
before burning (e.g., minimum visibility at Eugene) produced weaker statis-
tical results than those reported here. The theoretically unanticipated
signs for the wind speed and atmospheric stability variables are fortunately
associated with statistically insignificant and relatively unimportant (in
terms of the size of their beta coefficients) variables.
Of course, these findings for Salem and Eugene demonstrate the conflict
inherent in the 1970 and 1971 DEQ field burning program to minimize visibil-
ity losses in Eugene by reducing South Valley acres burned under north winds
while allowing large South Valley acreages to be burned with south winds.
Burning South Valley acreage under northerly wind conditions reduces visibil-
ity in Eugene, while burning such acreage with south winds lowers visibility
in Salem. Such conflicts are necessarily involved in smoke management prog-
rams which attempt to minimize the harmful effects of field-burning smoke
by attempting to control the distribution rather than the volume of the smoke.
These conflicts are discussed extensively in the main text of this report.
We next turn to the exceptionally poor performance of our five models
when fit to 1971 data for Eugene and Salem. Tables D.I through D.10 indi-
cate that every model had an R equal to zero and numerous statistically
insignificant regression coefficients with theoretically unanticipated signs
when fit to 1971 data. In the case of Eugene (see Tables D.1-D.5), the only
variable with a regression coefficient statistically significant at the 20
or 30 percent level is acres burned in the southern half of the Willamette
Valley. For Salem (see Tables D.6-D.10), the only coefficients statistically
significant at the 20 or 30 percent level are those for acres burned in the
northern portion of the Willamette Valley. Most of the other variables were
statistically insignificant and had signs contrary to our theoretical
expectations.
Needless to say, these poor results for 1971 suggested that further
analysis was required. Two types of additional analysis were undertaken.
First, an analysis of variance was performed to test whether the set of
coefficients in our estimated models for 1971 was significantly different
from the set of coefficients for the same models fit to 1969-70 data, i.e.,
do the models fitted separately to 1969-70 and 1971 periods statistically
explain significantly more variation in minimum daylight visibility than
the same models for the 1969-70-71 period as a whole.* Second, a detailed
analysis of the 1969-70 and 1971 data was performed to determine whether the
observed changes in our regression results are related, at least in part,
to the apparent increasing effectiveness of the DEQ regulated burning policy
over the 1969-70-71 period. The results of these analyses follow.
* See Gregory C. Chow, "Tests of Equality between Sets of Coefficients
in Two Linear Regressions," Econometrica, Vol. 28, No. 3 (July 1960), pp. 591-
605 for the appropriate testing procedure and definition of the test statistic,
D-9
-------
Despite the very substantial decrease in the goodness-of-fit of our
models between the 1969-70 and 1971 subperiods, an analysis of variance
indicated that we could not reject at the 10 percent level the hypothesis
that for Eugene the two sets of regression coefficients for all fLve
.models came from the same population. The apparent explanation for this
finding lies in the rather great similarity in the size of most regression
coefficients for the two subperiods, especially the similarity in the size
of the most important (in terms of its beta coefficient) South-Willamette-
Valley-acres-burned variable.
In the case of Salem, however, an analysis of variance indicated that
we could reject at the 5 percent level the hypothesis that the 1969-70 and
1971 sets of regression coefficients come from the same population for model
II and at the 10 percent level for models III, IV, and V. These findings
suggest only a rather limited shift in the underlying atmospheric mechanism
to dilute and transport load between the 1969-70 and 1971 periods.* Since
meteorological conditions were basically the same over the 1969-71 period,
these findings in turn suggested the possibility that the effectiveness of
DEQ burning regulations changed over the period, thereby contributing to
the poor performance of the models when fit to 1971 data.
Field burning regulations, schedules, and actual acres burned under
different meteorological conditions were redesigned each year over the
1969-71 period in an attempt to keep smoke away from the Eugene-Spring
field area while having a minimal effect on Salem. With controls of varying
stringency, the DEQ simultaneously redistributed and reduced daily acreage
burned under northerly wind conditions and increased acreages burned under
persistent southwesterly winds. Since the daily decision concerning which
acreages to be burned was made early each morning on the basis of the then
existing air quality conditions plus predicted meteorological conditions
for later in the day, significant and (perhaps) systematic differences in
our regression results would be expected following implementation of a
positive burning policy. Consequently, with controlled variations in acres
burned according to visibility-before-burning and predicted afternoon meteor-
ological conditions, one would expect to observe a deterioration in the
performance of our models over the 1969-71 period.
A careful study of Table D.16 reveals how DEQ burning regulation policies
and improved prediction of afternoon meteorological conditions have jointly
caused our models to perform uniformly poorly in 1971 when compared with
earlier years. Table D.16 reports the simple correlation coefficients be-
tween acres burned in the northern and southern sections of the Willamette
Valley, visibility-before-burning, and mid-afternoon inversion height in
1969, 1970, 1969-70, and 1971 at Eugene and Salem. Since positive field
* Although all other possible shifts in structure were also examined
for Eugene and Salem (1969 v. 1970, 1969 v. 1971, 1969 v. 1970-71, and
1970 v. 1971), none proved significant at the 10 percent level.
D-10
-------
burning control policies aimed at keeping smoke away from Eugene and Salem
would imply positive correlations between acres burned, visibility at 7 a.m.,
and mid-afternoon inversion height (recall that inversion height is measured
here in millibars of pressure), it is encouraging to note that 5 of 6 cor-
relation coefficients for 1971 have signs consistent with such positive
control policy whereas only 2 of 12 signs in 1969 and 1970 are consistent
with a positive policy. Unquestionably the most dramatic policy reversal
occurred with respect to acres burned in the South Willamette Valley under
northerly wind conditions. In 1969-70 the simple correlation coefficient
between South acres-burned and visibility before burning equalled -0.394
while in 1971 it equalled 0.404, and a perverse policy in 1969-70 became
a positive policy in 1971. Consequently, in the case of both Eugene and
Salem, an effectively implemented DEQ burning control policy appears to
contribute to explaining the rather poor performance of our simple models
when fit to 1971 data. We therefore conclude that the observed difference
in our regression results for 1969-70 and 1971 derive from policy and not
from an incorrect specification of fundamental relationships.
D.4. Implications and Tentative Conclusions
The objective of a smoke management policy is the minimization of the
harmful effects associated with the smoke produced by the open burning of
a given total acreage of agricultural fields. Operationally, smoke manage-
ment as practiced by the DEQ attempts to minimize visibility losses attri-
butable to field burning smoke in Eugene and Salem. Smoke management thus
minimizes rather than eliminates visibility losses in certain areas, while
undoubtedly increasing visibility losses in other areas. Viewed most broadly
the foregoing results clearly establish that smoke management can success-
fully reduce visibility losses from field burning smoke in Eugene and Salem.
Moreover, our results suggest that reductions in acres burned, especially
acres burned between Salem and Eugene would greatly facilitate successful
smoke management.
In our second annual report we employed regression results for Eugene
in 1969 and 1970 to study the effectiveness of the 1970 DEQ burning program.
There we concluded that probably not more than half of the 1969-70 improve-
ment in Eugene visibility was attributable to the reduction in acres burned
in the Valley under northerly wind conditions. Moreover, perhaps as much
as two-thirds of the 1969-70 improvement in Eugene visibility resulted from
the controlled 1970 geographic distribution of acres-burned according to
predicted late morning and afternoon meteorological conditions. All of
which suggested that perhaps most of the success of the 1970 DEQ burning
program and schedule derived from (a) accurate meteorological forecasts and
(b) control over the geographic distribution of acres-burned rather than
the reduction in acres-burned.
Here, in contrast, we conclude by reporting equations which represent
pollution production functions for Eugene and Salem. Such functions are
required to predict the impact of alternative burning control policies.
The desired functions were obtained by solving the following equations for
a and a :
D-ll
-------
V - V*
3A71 al + 2a2A71
A71 A71
where V?1 = mean 1971 minimum daylight visibility
V* = estimated mean minimum daylight visibility with no relevant
_ acres burned
A = mean 1971 relevant acres burned
1 = estimated elasticity of minimum daylight visibility with
respect to carefully regulated relevant acres burned.
This procedure explicitly assumes that there exists a continuous, smooth,
inverse relationship between_minimum_daylight visibility and relevant acres
burned which passes through V and A?- and has a curvature consistent with
the log - log functional form employed in our multiple regression statis-
tical analysis. We expected and did obtain equations where a < 0 and
a2 > 0.
The upper half of Table D.17 presents our estimated equations
for the pollution production functions associated with carefully regulated
field burning for Eugene and Salem under northerly and southerly wind con-
ditions. (The_ lower half of Table D.17 presents the data which,
together with V*, were employed to calculate a., and a_.) V* was estimated
for Eugene and Salem under northerly wind conditions By calculating mean
minimum daylight visibility on days in August (the heart of the burning
season) 1967 when the Oregon State Fire Marshall prohibited all burning.
In the case of Salem under southerly wind conditions, however, careful
study suggested that 14.0 miles would be a "reasonable" estimate of minimum
daylight visibility with no burning. Otherwise the data employed to calcu-
late a and a« were obtained from our 1971 and 1970-71 (or 1969-70-71) re-
gression results for estimates of N and mean observed 1971 minimum daylight
visibility and relevant acres burned. Since the 1971 DEQ burning program
was in fact reasonably successful in minimizing the visibility losses asso-
ciated with open field burning, our procedure for calculating the equations
presented in Table D.17 implies that the reported equations specify pollu-
tion production functions that approximate the functions truly relevant for
D-12
-------
estimating the benefits of reducing acres burned in the Willamette Valley
under different wind conditions.* Consequently, these equations underlie
the benefit estimates reported in the main text of this report.
* Additional evidence supporting the reasonableness of the functions
reported in Table D.17 was obtained when we compared our direct regression
estimates of visibility elasticities with respect to acres burned in 1970
with those calculated at mean 1970 acres-burned with the functions in Table
D.17. The calculated elasticity estimates were in no instance as much as
5 percent different from the direct elasticity estimates. We therefore
conclude that our pollution production function equations appear "reasonable".
D-13
-------
Table D.I
Model I Regression Results For Eugene
Under Northerly Wind Conditions
Regression Coefficients, Etc.
Constant
North acres burned
South acres burned
Visibility at 7 a.m.
Number of observations
R2
1969-70
2.458
-0.304b
(2.050)
-0.332a
(3.970)
0.269C
(1.912)
40
0.520
1971
0.976
0.287X
(0.751)
-0.352d
(1.512)
0.060
(0.173)
16
-
1969-70-71
2.003
-0.140
(0.962)
-0.341a
(4.277)
0.262b
(2.064)
56
0.368
F-statistic
Beta Coefficients
North acres burned
South acres burned
Visibility at 7 a.m.
15.860C
0.806
0.531 0.288X
1.028 0.573
0.495 0.065
12.418C
0.284
1.261
0.608
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
Significant at the 30 percent level.
D-14
-------
Table D.2
Model II Regression Results For Eugene
Under Northerly Wind Conditions
Regression Coefficients, Etc.
Constant
North acres burned
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Number of observations
R2
1969-70
13.967
-0.312b
(2.267)
-0.314a
(4.018)
0.244C
(1.862)
-3.928b
(2.604)
40
0.588
1971
9.059
0.235X
(0.603)
-0.322d
(1.353)
-0.138X
(0.329)
-2.647
(0.869)
16
_
1969-70-71
10.556
-0.1546
(1.094)
-0.341a
(4.429)
0.195d
(1.542)
-2.877b
(2.183)
56
0.414
F-statistic
Beta Coefficients
15.502'
0.781
11.180C
North acres burned
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
0.545
0.966
0.447
0.626
0.231X
0.518
0.126X
0.333
0.311
1.261
0.439
0.621
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
Significant at the 30 percent level.
D-15
-------
Table D.3
Model III Regression Results For Eugene
Under Northerly Wind Conditions
1969-70
1971
1969-70-71
Regression Coefficients, Etc.
Constant
North acres burned
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
Number of observations
o
F-statistic
Beta Coefficients
14.608
9.334
0.589
12.767'
0.572
10.549
-0.312
(2.272)
-0.296a
(3.727)
0.224C
(1.709)
-4.229a
(2.773)
0.2226
(1.140)
40
0.261X
(0.568)
-0.3326
(1.264)
-0.131X
(0.297)
-2.735
(0.837)
-0.075X
(0.123)
16
-0.176
(1.237)
-0.325a
(4.154)
0.166e
(1.287)
-2.926b
(2.224)
0.211e
(1.101)
56
0.416
9.223C
North acres burned
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
0.540
0.892
0.409
0.664
0.273
0.228X
0.507
0.119X
0.336
0.049X
0.351
1.180
0.366
0.632
0.313
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
Significant at the 30 percent level.
D-16
-------
Table D.4
Model IV Regression Results For Eugene
Under Northerly Wind Conditions
Regression Coefficients, Etc.
Constant
North acres burned
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Atmospheric stability at 4 p.m.
Number of observations
R2
F-statistic
Beta Coefficients
North acres burned
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Atmospheric stability at 4 p.m.
1969-70
\
13.697
-0.313b
(2.241)
-0.315a
(3.979)
0.244C
(1.845)
-3.886b
(2.536)
0.151
(0.349)
40
0.575
12.115a
0.545
0.968
0.449
0.617
0.085
1971
8.880
0.246X
(0.541)
-0.325s
(1.272)
-0.150X
(0.305)
-2,647
(0.829)
0.164
(0.157)
16
-
0.569
0.217X
0.511
0.122X
0.333
0.023
1969-70-71
10.515
-0.154e
(1.080)
-0.3423
(4.371)
0.195d
(1.522)
-2.871b
(2.150)
0.025
(0.050)
56
0.402
8.769a
0.310
1.256
0.438
0.618
0.014
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
a
Significant at the 1 percent level.
b
Significant at the 5 percent level.
£
Significant at the 10 percent level.
Significant at the 20 percent level.
e
Significant at the 30 percent level.
D-17
-------
Table D.5
Model V Regression Results For Eugene
Under Northerly Wind Conditions
Regression Results, Etc.
Constant
North acres burned
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
Atmospheric stability at 4
Number of observations
R2
F-statistic
Beta Coefficients
North acres burned
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
Atmospheric stability at 4
1969-70
14.361
-0.312b
(2.244)
-0.297
(3.688)
0.255
(1.693:)
-4.189
(2.701)
0.220
(1.113)
p.m. 0.135
(0.312)
40
0.578
10.3733
0.544
0.894
0.411
0.644
0.270
P-m- 0.076
1971
9.119
0.279X
(0.513)
-0.337
(1.179)
-0.148X
(0.284)
-2.742
(0.795)
-0.082X
(0.126)
0.219
(0.071)
16
-
0.430
0.217X
0.499
0.120X
0.337
0.053X
0.030
1969-70-71
10.531
-0.1756
(1.221)
-0.325
(4.094)
0.166
(1.272.)
-2.924b
(2.192)
0.211
(1.089)
0.011
(0.022)
56
0.404
7.532a
0.350
1.175
0.365
0.629
0.312
0.006
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
d
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
6 Significant at the 30 percent level.
D-18
-------
Table D.6
Model I Regression Results For Salem
Under Northerly Wind Conditions
Regression Coefficients, Etc.
Constant
North acres burned
Visibility at 7 a.m.
Number of observations
I2
F-statistic
Beta Coefficients
North acres burned
Visibility at 7 a.m.
1969-70
0.795
-0.172d
Cl. 451)
0.5993
(3.953)
46
0.255
9.435a
0.383
1.044
1971
2.124
-0.305d
(1.519)
-0.124X
(0.456)
18
-
1.285
0.376
0.113X
1969-70-71
1.097
-0.201°
(1.935)
0.436
(3.214)
64
0.156
7.463a
0.517
0.859
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
Significant at the 30 percent level.
e
D-19
-------
Table D.7
Model II Regression Results For Salem
Under Northerly Wind Conditions
Regression Results, Etc.
Constant
North acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Number of observations
f\
1969-70
10.596
-0.1386
(1.215)
0.4783
(3.120)
-3.333b
(2.330)
46
0.324
1971
0.455
-0.289d
(1.349)
-0.090X
(0.298)
0.537X
(0.298)
18
1969-70-71
4.573
-0.198C
(1.913)
0.385b
(2.666)
-1.168
(1.020)
64
0.1S7
F-statistic
Beta Coefficients
8.747
0.834
5.326
North acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
0.306
0.785
0.586
0.345
0.076X
0.076X
0.511
0.712
0.273
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Q
, Significant at the 1 percent level.
b
Significant at the 5 percent level.
Q
Significant at the 10 percent level.
d
Significant at the 20 percent level.
e
Significant at the 30 percent level.
D-20
-------
Table D.8
Model III Regression Results For Salem
Under Northerly Wind Conditions
1969-70
1971
1969-70-71
Regression Results, Etc.
Constant
North acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
Number of observations
—2
R
F-statistic
Beta Coefficients
North acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
11.522
-0.119
(1.044)
0.383b
(2.227)
-3.708°
(2.544)
0.243e
(1.197)
46
0.331
6.986
0.261
0.557
0.637
0.300
0.929
-0.275e
(1.194)
-0.104X
(0.326)
0.394X
(0.199)
-0.078X
(0.223)
18
—
0.595
0.316
0.086X
0.053X
0.059X
4.611
-0.201°
(1.935)
0.344°
(2.270)
-1.213e
(1.058)
0.156
(0.933)
64
0.155
4.203a
0.518
0.607
0.283
0.249
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
a
Significant at the 1 percent level.
b
Significant at the 5 percent level.
c
Significant at the 10 percent level.
d
Significant at the 20 percent level.
e
Significant at the 30 percent level.
D-21
-------
Table D.9
Model IV Regression Results For Salem
Under N rtherly Wind Conditions
Regression Results, Etc.
Constant
North acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Atmospheric stability at 4 p.m.
Number of observations
I2
1969-70
10.589
-0.1388
(1.202)
0.477a
(3.035)
-3.340b
(2.302)
0.030
(0.070)
46
0.308
1971
0.931
-0.2726
(1.218)
-0.097X
(0.309)
0.771X
(0.400)
0.655
(0.463)
18
—
1969-70-71
4.518
-0.199°
(1.903)
0.382b
(2.599)
-1.171
(1.014)
0.069
(0.161)
64
0.143
F-statistic
Beta Coefficients
6.406C
0.644
3.936s
North acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Atmospheric stability at 4 p.m.
0.306
0.773
0.586
0.018
0.320
0.081X
0.105X
0.122
0.513
0.700
0.273
0.043
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
Significant at the 30 percent level.
D-22
-------
Table D.10
Model V Regression Results For Salem
Under Northerly Wind Conditions
Regression Results, Etc.
Constant
North acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
Atmospheric stability at 4 p.m.
Number of observations
1969-70
11.514
-0.120
(1.034)
0.381b
(2.169)
-3.715b
(2.513)
0.2436
(1.183)
0.031
(0.072)
46
1971
0.507
-0.262e
(1.093)
-0.107X
(0.323)
0.651X
(0.305)
-0.060X
(0.164)
0.626
(0.423)
18
1969-70-71
4.557
-0.202C
(1.924)
0.341b
(2.213)
-1.216e
(1.051)
0.156
(0.925)
0.068
(0.158)
64
F-statistic
Beta Coefficients
0.314
5.454C
0.482
0.140
3.312
North acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
Atmospheric stability at 4. p.m.
0.262
0.549
0.637
0.300
0.018
0.299
0.088X
0.083X
0.045X
0.116
0.519
0.597
0.283
0.249
0.043
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
Significant at the 30 percent level.
D-23
-------
Table D.ll
Model I Regression Results For Salem Under
Southerly Wind Conditions
1969-70
1970-71
1969-70-71
Regression Coefficients, Etc.
Constant
South acres burned
Visibility at 7a.m.
Number of observations
o
F-statistic
Beta Coefficients
South acres burned
Visibility at 7a.m.
-0.842
-0.113e
(1.219)
0.433
(1.764)
17
0.107
1.965
0.284
0.425
0.740
-0.057
(0.643)
0-374C
(1.790)
18
0.022
1.809
0.172
0.478
0.829
-0.0836
(1.243)
0.373C
(2. 005)
26
0.079
2.690
0.350
0.564
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
o
Significant at the 1 percent level.
Significant at the 5 percent level.
Q
Significant at the 10 percent level.
Significant at the 20 percent level.
e
Significant at the 30 percent level.
D-24
-------
Table D.12
Model II Regression Results For Salem Under
Southerly Wind Conditions
Regression Coefficients, Etc.
Constant
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Number of observations
R2
F-statistic
Beta Coefficients
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
1969-70
22.647
-0.273a
(3.971)
0.187e
(1.128)
-7.218a
(4.631)
17
0.663
10.6163
0.687
0.187
0.849
1970-71
13.082
-0.182°
(1.990)
0.285d
(1.556)
-4.049b
(2.490)
18
0.270
3.691b
0.458
0.358
0.573
1969-70-71
15.182
-0.190a
(2.976)
0.284C
(1.565)
-4.768a
(3.394)
26
0.367
6.454a
0.694
0.423
0.791
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
Significant at the 30 percent level.
D-25
-------
Table D.13
Model III Regression Results For Salem Under
Southerly Wind Conditions
Regression Coefficients, Etc.
Constant
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
Number of observations
-2
R
F-statistic
Beta Coefficients
South acres burned
Visibility at 7
Inversion height at 4 p.m.
Wind speed at 4 p.m.
1969-70
21.979
-0.2743
(3.837)
0.179
(1.028)
-7.018
(4.137)
0.091
(0.389)
17
0.640
7.437a
0.689
0.176
0.825
0.067
1970-71
14.499
-0.160d
(1.620)
0.281
(1.506,)
-4.498
(2.533)
-0.182X
(0.701)
18
0.238
2.791
0.380
0.353
0.594
0.164X
1969-70-71
16.442
-0.1823
(2.786)
0.277
(1.751)
-5.156
(3.407)
-0.143X
(0.739)
26
0.352
4.877
0.657
0.413
0.803
0.1 74X
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
g,
Significant at the 1 percent level.
t)
Significant at the 5 percent level.
c
Significant at the 10 percent level.
Significant at the 20 percent level.
e
Significant at the 30 percent level.
D-26
-------
Table D.14
Model IV Regression Results For Salem
Under Southerly Wind Conditions
Regression Coefficients, Etc.
Constant
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Atmospheric stability at 4 p.m.
Number of observations
1969-70
27.131
-0.2663
(4.055)
0.165
(1.035)
-8.546
(4.934)
-0.321X
(1.035)
17
1970-71
12.386
-0.190°
(1.957)
0.283d
(1.496)
-3.829b
(2.138)
0.089
(0.353)
18
1969-70-71
15.179
-0.190a
(2.876)
0.284°
(1.768)
-4.767a
(3.087)
0.000
(0.002)
26
F-statistic
Beta Coefficients
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Atmospheric stability at 4 p.m.
0.696
9.326a
0.669
0.162
1.005
0.266X
0.217
2.627
0.465
0.355
0.508
0.355
0.335
4.621£
0.686
0.422
0.737
0.000
Note: Figures in parentheses are t-statistics, and X indicates that the
coefficient has a theoretically unanticipated sign.
Significant at the 1 percent level.
Significant at the 5 percent level.
Significant at the 10 percent level.
Significant at the 20 percent level.
Significant at the 30 percent level.
D-27
-------
Table D.15
Model V Regression Results For Salem
Under Southerly Wind Conditions
Regression Coefficients, Etc.
Constant
South acres burned
Visibility at 7 a.m.
Inversion height at 4 p.m.
Wind speed at 4 p.m.
Atmospheric stability at 4 p.m.
Number of observations
-2
1969-70
28.000
-0.265a
(3.856)
0.169
(1.010)
-8.805a
(4.243)
-0.064X
(0.255)
-0.064X
(1.390)
17
/-* /" ~t -t
1970-71
14.877
-0.155d
(1.3311
0.282d
(1.448)
-4.617C
(2.028)
-0.200X
(0.586)
-0.0 30X
(0.090)
18
r\ t r r\
1969-70-71
17.396
-0.177b
(2.576)
0.279°
(1.725)
-5.451a
(3.065)
-0.174X
(0.797)
-0.087X
(0.335)
26
r\ i
-------
Table D.16
Simple Correlation Coefficients Between Acres Burned in the Northern
and Southern Sections of the Willamette Valley, Inversion Height at
4 p.m., and Visibility at 7 a.m., Eugene and Salem, 1969, 1970, 1969-
70, and 1971.
Eugene
1969
1970
1969-70
1971
1969
1970
1969-70
1971
Salem
1969
1970
1969-70
1971
Acres Burned in Willamette
North Valley South
Visibility at 7 a.m.
-0.316
0.142
-0.090
0.249Z
Inversion Height at 4 p
0.136
-0.126Z
0.024
-0.210Z
Visibility at 7 a.m.
-0.196
0.010Z
-0.087
-0.037
Valley
Valley
-0.106
-0.620
-0.394
0.404Z
.m.
0.005
0.280
0.130
-0.164Z
Inversion Height at 4 p.m.
1969
1970
1969-70
1971
0.204
0.079
0.149
-0.248Z
Note: Z indicates that the sign of the correlation coefficient is consistent
with a positive field burning control policy.
D-29
-------
Table D.17
Equations For Open-Field-Burning Pollution
Production Functions In The
Willamette Valley
V*
3T
Eugene ,
North Wind
Conditions
11.30
-4.18 X 10~3
Salem,
North Wind
Conditions
13.80
-2.53 X 10~3
Salem,
South Wind
Conditions
14.00
-2.33 X 10~3
7.13 X 10
-7
1.42 X 10
-7
2.80 X 10
-7
V
71
A
71
8.04
926 South Acres
-0.333
10.67
1337 North Acres
-0.27b
7.75
4710 South Acres
-0.19°
Note: V*, a , and a. obtained as described in text.
3 Tables D.2 and D.3
Tables D.8 and D.9
° Table D.12
D-30
-------
APPENDIX E
THE DETERMINANTS OF AGRICULTURAL LAND VALUES
IN THE WILLAMETTE VALLEY
This appendix reports the results of a statistical investigation of
the determinants of agricultural land values in the Willamette Valley.
Primary emphasis in the study was given to the determinants of the intra-
and inter-county structure of land values in the eight Willamette Valley
counties where substantial quantities of grass seed are raised. The eight
counties purposely selected for inter-county study are Benton, Clackamas,
Lane, Linn, Marion, Polk, Washington, and Yamhill. In contrast, the par-
ticular 72 grass seed operations included in our land value study consti-
tute a subset of a larger sample of seed operations from which we could
determine the ownership of 161 parcels (at different locations) in seed
production during 1970 and thereby secure county assessment records for
each parcel.* This subset of operations permitted intensive intra- and
inter-county study of land values only for Benton, Linn, and Marion Coun-
ties. Fortunately, this appears quite appropriate because .Linn and Benton
Counties are major producers of ryegrass, while Marion County leads all
other Oregon counties in the production of bentgrass, red fescue, and
chewing fescue. These seed varieties are among those that Oregon dominates
the nation in production.** The statistical results reported therefore
encompass the geographical area of concern quite well.
* The large sample was selected by means of two-stage sampling tech-
nique for a detailed industrial-organization-type study of the Willamette
Valley seed industry. Within each county seed operations were selected
at random. The analysis of data obtained from the large sample provides
much of the information concerning resource organization, cost structures,
levels of resource utilization, farm sizes, and managerial objectives in
the Willamette Valley seed industry presented elsewhere in this report.
For additional details concerning the industry, see Douglas E. Fisher,
An Economic Analysis of Farms Producing Grass Seed in the Willamette Valley3
With Special Attention to the Cultural Practice of Field Burning (unpub-
lished Ph.D. dissertation, Oregon State University, 1972).
** Middlemiss, W. E. and R. 0. Coppedge, Oregon's Grass and Legume Seed
Industry in Economic Perspective, Special Report 284, Cooperative Extension
Service, Oregon State University, Corvallis, Oregon. April 1970.
E-l
-------
Here, least-squares multiple-regression analysis provides the analyti
cal tool for our empirical investigation of the determinants of bo_th agri-
cultural land market values and agricultural land value differentials.
Land value differentials are defined as the difference between the market
and farm use values, where farm use value equals the capitalized value of
the assessor's estimate of the rent at which the land could presently be
leased. The determinants of land value differentials, as well as market
values, are investigated here because such differentials are large, posi-
tive, exhibit greater variance than market values per acre, and necessarily
reflect the higher future earnings anticipated from Willamette Valley land
now in agricultural use.
E.I The Model
The basic hypothesis of our model is that the market value of any par-
cel of agricultural land depends on the capitalized value of the parcel's
current and expected future net earnings. Of course, since current and
future uses of any given parcel of land are not likely to be the same in a
rapidly urbanizing area, current land values are unlikely to be closely
related to current net earnings. Such land would generally be anticipated
to be converted to new agricultural or non-agricultural uses in the future,
and a parcel's expected earnings in those new uses would be the major deter-
minants of the parcel's current market value. Therefore, our basic hypo-
thesis requires reformulation for purposes of testing.
Recalling that land now in agricultural use may remain in agricultural
use or be converted to non-agricultural use at some point in the future, at
least two independent variables (or indices) would appear required "to cap-
ture" the dual influence of these possibilities on current land values. To
be precise, we hypothesize that current and expected future earnings in
agricultural uses are directly related to a consistent and well-defined
index of soil quality, while expected future earnings in non-agricultural
uses are directly related to the relative accessibility of land parcels to
nearby urban centers. Although neither of these hypotheses is particularly
unusual when compared to those made in other studies of agricultural land
values, the particular soil quality index selected and the absence of data
aggregation of the usual sort jointly distinguish this research effort from
those undertaken previously.
Dealing first with data aggregation and future conversion to non-agri-
cultural uses, we follow previous studies by measuring the accessibility
of land parcels in terms of highway distance to the nearest town with a
population over 1,000 but do so for each of the 161 different parcels of
land in our sample of 72 farms. Previous investigations, however, have
treated farms, even farms composed of non-contiguous land parcels, as the
relevant units for study. Consequently, in contrast to these other agri-
cultural land value studies, this one avoids the bias in accessibility
measurements introduced when non-contiguous land parcels are treated analy-
tically a£ ±f_ such parcels were contiguous.
E-2
-------
Turning now to our hypothesis concerning expected future earnings in
agricultural uses, we suggest that future productivity and earnings are
more closely and directly related to informed judgment concerning a parcel's
soil quality than to the parcel's current net earnings. This seems reason-
able because current earnings necessarily reflect basic soil quality, the
entreprenuerial effort of the current operator, and current land use. How-
ever, given the existence of well-functioning agricultural land and commo-
dity markets, better quality soils would generate higher than average earn-
ings per acre, and vice versa. On such assumptions, therefore, a consistent
and well-defined index of soil quality could serve as a proxy for expected
future earnings in agricultural uses.
To devise a continuous index of soil quality suitable for use in a
multiple-regression study of land values is no simple task. In fact, soil
quality is multi-dimensional and would reflect, in combination, soil tex-
ture, depth of surface soil, nature of subsoil, fertility, etc. Fortunately
for present purposes, agricultural land in the Willamette Valley is typically
assessed in terms of its "farm use value," which equals the capitalized value
of the county assessor's estimate (for assessment purposes only) of the rent
at which it would presently be leased. Given the relatively high level of
professional skill demanded of Oregon assessors, as well as the requirement
that he treat properties uniformly, this rent estimate provides a consistent
and continuous index of soil quality based on informed judgment. Moreover,
the assessor's judgment necessarily reflects, albeit imperfectly, the weights
the market presently establishes on the various dimensions of soil quality.
Finally, it should be noted that the degree of association between the
accessibility (or "urban influence") variables and market values are expected
to vary among counties according to the county's present urbanization and
the importance attached to future conversion of agricultural land to urban
uses by different assessors. For this reason, some experimentation with
dummy variables reflecting county urbanization seemed appropriate and is
reported here.
E.2 Statistical Results
Three variants of our basic model have been tested against agricultural
land market value and differential value data for Benton, Linn, and Marion
Counties individually and jointly, as well as for the eight Willamette
Valley counties together. These variants differ from one another in terms
of their respective specifications of the independent variables and the
assumed functional forms assumed for analysis. The model variants are as
follows:
I. V = V (SQ, DT)
II. LogV = V (Log SQ, Log DT)
III. V = VIII(SQ,DT,D],D2)
E-3
-------
where V = assessed market value per acre or assessed farm use value, SQ
current rent per acre, i.e., our index of soil quality, DT = distance to
nearest town with population over 1,000, D = dummy variable equal to I
if population of the nearest town is greater than 1,000 but less than
10,000 and zero otherwise, and D = dummy variable equal to 1 if the pop-
ulation of the nearest town is greater than 10,000 but less than 20,000
and zero otherwise.
Tables E.I through E.6 report our regression results for model vari-
ants with agricultural land market values per acre as the dependent vari-
able, while Tables E.7 through E.ll present the results obtained with land
value differentials per acre as the dependent variable.
With regard to our market value results, it is important to note that
all estimated regression coefficients reported in Tables E.I - E.6 have the
expected signs and support the hypotheses of our basic model.* Moreover,
Tables E.3 and E.4 indicate that model variant II performs exceedingly well
when judged by the usual statistical tests, particularly when one compares
its performance with that of variants I and III when tested on combined
three and eight county data. In terms of goodness of fit as measured by
R and the F-statistic, model variant II performs better than either vari-
ant I or III, and both estimated regression coefficients are statistically
significant at the one percent level and properly signed.
Two features of the regression results deserve special comment. First,
Marion County's relatively large constant terms and distance-to-town t statis-
tics are consistent with the hypothesis that urbanization has had a greater
influence on agricultural land values there than in Benton and Linn Counties.
In fact, if urbanization is measured by aggregate county population, popula-
tion of the.largest city in the county, or total retail sales, Marion County
Would be classified as more urbanized than either Benton or Linn Counties.
Second, for model variant II, the size and statistical significance of our
soil quality regression coefficients is relatively greater and the distance-
to-town coefficient relatively smaller for Linn than for Benton and Marion
Counties. Such results are consistent with the hypothesis that Linn County
land has the lowest potential convertibility to non-seed agricultural or
non-agricultural uses among the three counties. Qualitative evidence in
support of this hypothesis derives from (1) the poorer quality soil in Linn
as opposed to Benton and Marion Counties and (2) the lesser present, and
hence forecast future, urbanization of Linn County. Consequently, model
variant II (see Table E.3 and E.4) performs rather well with agricultural
land market values per acre as the dependent variable.
* Since dummy variables are essentially intercept shifters, no hypothe-
ses with regard to their signs are supported or rejected by the results
reported in Tables E.5 and E.6.
E-4
-------
Turning next to our findings when agricultural land value differentials
are the dependent variable, Tables E.7 through E.ll report poorer regression
results than those discussed above. This general finding cannot be regarded
as surprising, however, because land value differentials (1) have a much
greater variance than market values and (2) presumably reflect other factors
than those associated with current agricultural use. Without an increase in
the number of independent variables, therefore, one would reasonably expect
less satisfactory performance of our basic model. Nonetheless, the following
features of our results are noteworthy.
First, as before, model variant II performs better than variants I and
III. Tables E.9 and E.10 show all estimated regression coefficients have
their theoretically anticipated signs and relatively uniform R and P-statis-
tics for model variant II. The multiplicative functional form assumed for
variant II once again appears more appropriate than the additive form most
often used in land value investigations. Of course, the high degree of
statistical significance for the distance-to-town variable supports our basic
hypothesis that land parcels located more closely to existing urban areas
are more likely to be converted to non-agricultural uses in the future.
Additional evidence in support of this hypothesis may be derived from a
comparison of the size of the distance-to-town regression coefficients in
Tables E.3, E.4, E.9, and E.10. Such comparisons indicate that the coeffi-
cients estimated with differentials as the dependent variable are approxi-
mately twice the size of the coefficients estimated with market value data.
Finally, the lack of statistical significance for the estimated soil-quality
regression coefficients in Tables E.9 and E.10 indicate that agricultural
land value differentials in the Willamette Valley are apparently not "ex-
plained" by either intra- or inter-county variation in potential productivity
in agricultural uses.
Since land value differentials are calculated in a fashion to minimize,
though undoubtedly not eliminate, the influence of current earnings from
agricultural use, our land value differential regression results usefully
supplement the market value results and lead us to the following tentative
conclusions. First, the structure of agricultural land values in the Wil-
lamette Valley reflect urban influences in much the same fashion as location
theory would predict and as other studies have discovered elsewhere. Second,
prospective urbanization and urban influences rather than prospective continued
agricultural use dominate the structure of agricultural land values through-
out the Willamette Valley.
E.3 Conclusion
For the purposes of measuring the costs associated with control of open
field burning in the Willamette Valley, the most important results obtained
are those reported in Table E.4. Despite inter-county variation in the rela-
tive importance of the factors that statistically account for variations in
agricultural land values, our basic model performs rather well when tested
against data from (1) the three major and (2) all eight seed-producing coun-
ties. These results suggest the existence of a relatively "well-behaved"
agricultural land value surface extending across county boundaries in the
Willamette Valley.
E-5
-------
Table E.I
Market Values
Regression Results: Model I For Benton, Linn, and Marion Counties Separately
Constant
Soil quality
Distance to town
R2
R2
F-statistic
Degrees of freedom
Benton
268.1636
13.45419
(8.526)
23.19713
(-4.654)
.7859
.7733
62.4034
34
Linn
181.1212
10.8166a
(7.041)
-13.9725^
(-2.451)
.5834
.5667
35.0103
50
Marion
697.5259
1 .8875
(.061)
-62.71783
(-6.783)
.7632
.7450
41.9099
26
Note: Figures in parentheses are t-statisties
J Significant at the 1 percent level.
Significant at the 5 percent level.
E-6
-------
Table L.2
Market Values
Regression Results: Model I for Benton, Linn and Marion Counties and
for all Counties.
Constant
Soil quality
Distance to town
R2
R2
F-statistic
Degrees of freedom
Three Counties
312.8117
8.7354a
(7.028)
-21 .9800a
(-5.616)
.5415
.5336
68.4915
116
All Counties
233.5416
10.9942a
(9.598)
-15.1824a
(-4.409)
.4927
.4863
76.7273
158
Note: Figures in parentheses are t-statisties.
a Significant at the 1 percent level.
E-7
-------
Table E.3
Market Values
Regression Results: Model II for Benton, Linn and Marion Counties Separately
Constant
Soil quality
Distance to town
R2
R2
F-statistic
Degrees of freedom
Benton
2.0676
0.6759a
(10.154)
-0.46539
(-4.599)
.8197
.8091
77.2835
34
Linn
1.6597
0.7232a
(10.63)
-0.1432a
(2.97)
.7479
.7378
14.83
50
Marion
2.3416
0.4872a
(3.194)
-0.56913
(-4.801)
.7144
.6924
32.5105
26
Note: Figures in parentheses are t-statistics.
Significant at the 1 percent level.
E-8
-------
Table E.4
Market Values
Regression Results: Model II for Benton, Linn and Marion Counties
Combined and for all Counties
Three Counties All Counties
Constant 1.9587 1.8497
0.5944a 0.63331
(10.886) (12.975)
Soil quality 0.5944a 0.63333
Distance to town
R2
R2
F-statistic
Degrees of freedom
-0.2652a
(-5.625)
.6581
.6527
111.6228
116
-0.1731a
(-4.695)
.5980
.5928
117.5392
158
Note: Figures in parentheses are t-statistics
a Significant at the 1 percent level.
E-9
-------
Table E.5
Market Values
Regression Results: Model III for Benton, Linn and Marion Counties
Separately.
Constant
Soil quality
Distance to town
Dummy 1
Dummy 2
R2
R2
F-statistic
Degrees of freedom
Benton
133.8253
13.49683
(10.398)
-7.0758
(-1.236)
-22.4137
(-1.193)
110.8673a
(3.252)
.8643
.8473
50.9357
32
Linn
139.1472
10.67983
(6.512)
-13.5581b
(-2.313)
42.0792
(0.867)
58.2341
(1.120)
.5942
.5598
17.5690
48
Marion
687.4362
1.9988
(0.584)
-62.2614a
(-6.223)
7.2854
(0.156)
-4.1374
(-0.038)
.7636
.7242
19.3794
24
Note: Figures in parentheses are t-statisties,
? Significant at the 1 percent level.
Significant at the 5 percent level.
E-10
-------
Table E.6
Market Values
Regression Results: Model III for Three Counties; Benton, Linn and
Marion Combined.
Three Counties
Constant 326.6881
Soil quality 8.4352a
(6.583)
Distance to town -22.1400
(-5.466)
Dummy 1 -13.0011
(-0.515)
Dummy 2 14.4446
(0.432)
R2 .5467
R2 .5278
F-statistic 34.3711
Degrees of freedom 114
Note: Figures in parentheses are t-statisties.
a Significant at the 1 percent level.
E-ll
-------
Table E.7
Land Value Differentials
Regression Results: Model I for Benton, Linn and Marion Counties
Separately.
Constant
Soil quality
Distance to town
R2
R2
F-statistic
Degrees of freedom
Benton
279.9287
2.8872
(1.832)
-23.91253
(-4.804)
.4972
.4676
16.8120
34
Linn
165.4939
0.4198
(0.265)
-13.1903b
(-2.240)
.1045
.0687
2.9163
50
Marion
768.8398
-10.17983
(-3.4187)
-69.7724a
(-7.8276)
.7132
.6911
32.3231
26
Note: Figures in parentheses are t-statistics.
k Significant at the 1 percent level.
Significant at the 5 percent level.
E-12
-------
Table E.8
Land Value Differentials
Regression Results: Model I for Three Counties; Benton* Linn and
Marion Combined and for all Counties.
Constant
Soil quality
Distance to town
R2
R2
F-statistic
Degrees of freedom
Three Counties
311.3299
-2.0002
(-1.574)
-21.25693
(-5.311)
.1982
.1844
14.3379
116
All Counties
234.6726
0.3904
(0.333)
-14.91889
(-4.229)
.1171
.1059
10.4816
158
Note: Figures in parentheses are t-statisties.
a Significant at the 1 percent level.
E-13
-------
Table E.9
Land Value Differentials
Regression Results: Model II for Benton, Linn and Marion Counties
Separately.
Constant
Soil quality
Distance to town
R2
R2
F-statistic
Degrees of freedom
Benton
2.6075
0.3180b
(2.152)
-1.02409
(-4.599)
.4828
.4528
15.8718
34
Linn
1.9992
0.1838
(0.997)
-0.4046a
(-3.095)
.2072
.1755
6.5354
50
Marion
2.8629
0.3649
(0.097)
-1.0638a
(-3.625)
.4097
.3643
9.0237
26
Note: Figures in parentheses are t-statistics
?'Significant at the 1 percent level.
Significant at the 5 percent level.
E-14
-------
Regression Results:
Table E.10
Land Value Differentials
Model II for Three Counties; Benton, Linn and
Marion Combined and for all Counties.
Constant
Soil quality
Distance to town
R2
R2
F- statistic
Degrees of freedom
Three Counties
2.3788
0.0487
(0.361)
-0. 5185s
(-4.454)
.1730
.1588
12.1366
116
All Counties
2.1778
0.1325
(1.127)
-0.3492a
(-3.930)
.1130
.1018
10.0744
158
Note: Figures in parentheses are t-statistics,
a Significant at the 1 percent level.
E-15
-------
Table E.11
Land Value Differentials
Regression Results: Model III for Benton, Linn and Marion Counties
Separately.
Constant
Soil quality
Distance to town
Dummy 1
Dummy 2
R2
R2
F-statistic
Degrees of freedom
Benton
120.5948
3.1019b
(2.489)
-5.4457
(-1.006)
-21.1583
(-1.175)
126.47373
(4.008)
.7147
.6790
20.0400
32
Linn
130.6073
0.2898
(0.170)
112.86315
(-2.115)
35.2514
(0.700)
49.6798
(0.921)
.1204
.0456
1.6419
48
Marion
700.1061
-8.7482b
(-2.586)
-63.3860a
(-6.415)
4.5226
(0.098)
-2.9568
(-0.028)
.6670
.6115
12.0167
24
Note: Figures in parentheses are t-statistics,
b Significant at the 1 percent level.
Significant at the 5 percent level.
E-16
-------
APPENDIX F
DEMAND AND SUPPLY RESPONSE FUNCTIONS
FOR GRASS SEEDS RAISED IN THE WILLAMETTE VALLEY
This appendix reports the specification and estimation of demand and
supply-response functions for six grass seeds raised in the Willamette
Valley. In 1969, these six grass seeds (ryegrass, tall, red, and chewings
fescues, Merion-Kentucky bluegrass, and bentgrass) were grown on 85 percent
of the acreage devoted to grass seed production in Oregon and provided more
than 75 percent of the total value of sales of all grass seeds produced in
Oregon.* For these six crops, our best estimates of own-price elasticities
of demand range from -1.3 to -2.4, while our best estimates of long-run
price elasticities of supply extend from only +1.5 to +2.0. The appendix
concludes by presenting the linear demand and supply equations which are
the basis of the calculations of changes in producers' and consumers' sur-
pluses reported in the main text of this report.
F.I Specification of Demand and Supply-Response Functions
Demand and supply relationships in the grass seed industry are com-
plex and difficult to specify for a variety of reasons. Domestic and
world seed prices are determined by the simultaneous of supplies and de-
mand at various marketing levels (form, wholesale, and retail) and outlets
(domestic utilization, exports, and carryover) for which only limited, and
in some respects unreliable, data exists. More particularly, the perennial
nature of many seeds makes specification of supply-response functions diffi-
cult, and the fact that most seeds are sold at retail in various mixtures
makes a truly adequate specification of independent demand functions quite
impossible. Data on grass seed exports is non-existent prior to 1966, while
reasonably reliable price and quantity data at the wholesale level is avail-
able for twelve or fewer years. Consequently, satisfactory estimates of de-
mand and supply-response functions are possible only at the farm level.
Fortunately, however, such estimates provide the minimum information required
to measure the social costs of regulating open field burning in the Willamette
Valley.
Willis E. Middlemiss and Robert 0. Coppedge, Oregon's Grass and
Legume Seed Industry in Economic Perspective, Special Report 284, Coopera-
tive Extension Service, Oregon State University, Corvallis, April 1970,
Table 3, p. 3.
F-l
-------
In this section we specify and discuss two demand relationships, one for
current utilization and the other for ending stocks (i.e., carryover), and an
aggregate Oregon supply-response function. It is hypothesized here that,
given total commercial availability (i.e., current production plus beginning
stocks), the price of seed at the farm level is jointly determined by the
interaction of the demands for current utilization and ending stocks. Ideally,
of course, foreign and domestic demands would be treated separately, but lack
of data precludes this procedure.
Although the estimated relationships are undoubtedly inadequate for truly
accurate prediction of future grass seed outputs, prices, disappearance, and
carryover, none of our major hypotheses concerning grass seed supply and demand
functions have been rejected, and the estimated relationships statistically ex-
plain the observed price-quantity data reasonably well. On this basis, however,
the estimated demand and supply-response relationships reported here do appear
quite adequate for us to accomplish our primary research objective: the measure-
ment of changes in producers' and consumers' surpluses that would have occurred
in the recent past if open field burning regulations had been in existence in
Oregon.
Demand for Utilisation. As the introduction to this section has indicated,
the unavailability of retail price data for particular grass seeds, the short
time-series of wholesale price and quantity data, the considerable importance of
stocks carried forward to future years, and very limited export data lead us to
specify and estimate demand for utilization functions at the farm level. More-
over, although we fully recognize the desirability of incorporating behavioral
variables and relationships to reflect various activities at different marketing
levels, limited data availability precluded specification of as complete a
model as we wanted. Consequently, we have been required to assume constancy in
marketing margins, production coefficients in industries employing grass seeds
as inputs, etc.
There are four major hypotheses postulated here concerning the demands
for utilization of Oregon grass seeds at the farm level. Each of these hypo-
theses is equivalent to those commonly tested in econometric studies of demand
functions. Here we postulate that demands for seed utilization vary inversely
with own prices of seed but directly with prices of substitute seeds. Both of
these hypotheses are standard, of course, and empirical experimentation is
employed to determine the closest substitute for each forage and turf grass
seed. Our third hypothesis concerned the appropriate domestic and foreign de-
mand shift variables for forage and turf grasses. We hypothesized direct rela-
tionships between (1) demands for forage seeds and number of cattle in the 12
southern U.S. states and in Canada, West Germany, and France, and (2) demands
for turf seeds and the number of housing starts in the United States. Finally,
we would not expect consumers to adjust to price and other changes instantaneously
—more likely they would adjust their utilization gradually to the desired or
equilibrium level. This final hypothesis derives from Nerlove's brilliant
F-2
-------
insight that we rarely observe equilibrium prices and quantities in competi-
tive markets.*
In sum,your hypotheses concerning demands for utilization can be written
as follows:
C = Fi (pit> V zt> zt> fit)
where Q = desired or long-run equilibrium quantity-demanded of i—
grass seed for utilization in year t,
Pit = average price of i— grass seed received by farmers in
year t,
P.. = average price of,j— grass seed received by farmers in
year t, where j— seed is a substitute for the i— seed,
Z = domestic shift variable: number of cattle in 12 southern
states, or number of housing starts in the U.S.
Z = foreign shift variable: number ,of cattle in Canada, West
Germany, and France,
f = disturbance term
and d d _ d ( d* _ d
Qit Qi,t-l " Xi (Qit Qi,t-l)
where 0 < A. < 1. The first equation expresses our first three hypotheses in
a desired or long-run equilibrium demand equation, while the second equation is
our postulated adjustment function. The adjustment function states that the
actual observed change in utilization is some fraction of the difference between
the actual utilization in the previous year and the current desired level of
utilization. Substituting the desired demand-for-utilization equation for Q.
in the adjustment function equation, we obtain the required demand regres- 1
sion equation:
<4 ' xi Fi
-------
Demand for Endi-ag Stocks. Ending commercial stocks are the other
important outlet for grass seeds. Ending stocks as percentages of total
commercial availability vary considerably from year to year, as well as
among grass seed types. In 1969, for example, ending stocks as a percentage
of commercial availability varied from 18 percent for bentgrass to 30 per-
cent for Merion-Kentucky bluegrass. These stocks of seed are presumably
held in expectation of a higher price in the following season. Consequently,
following conventional economic theorizing, we hypothesize that the demand
for ending stocks is directly related to total commercial availability and
inversely related to current price.
More precisely,
where Q?S = quantity of ending stocks of i— seed held by farmers,
dealers, and government,
P. = average price of i— grass seed received by farmers,
Q?a = total commercial availability (i.e., current production
plus beginning stocks) of itn grass seed,
g. = disturbance term.
Supply-Response Function. Our specification of aggregate Oregon grass
seed supply-response functions is quite conventional. This specification
embodies three hypotheses often subjected to test by multiple-regression
analysis of time-series. Although supply functions can be estimated directly
from firm production functions, as well as by budgeting and linear programming
techniques, these approaches have been subject to considerable criticism re-
cently.* Moreover, for the present purpose of estimating changes in producers'
surpluses, what is often regarded as a weakness of the multiple-regression
approach becomes an important strength. Since we wish to estimate the changes
in surpluses that would have occurred in the recent past if open field burn-
ing regulations had been adopted, it is in fact an advantage that our esti-
mated supply-response functions are based on producers' typical past reactions
to changes in relevant variables. (Of course, this advantage becomes a defin-
ite disadvantage for research designed to predict future supply changes.)
*
For a very revealing critical analysis of supply functions derived
from production functions, see Larry J. Wipf and D. Lee Dawden, Reliability
of Supply Functions Derived from Production Functions, American Journal of
Agricultural Economics. February 1969, pp. 170-79. Wipf and Dawden con-
clude that supply functions derived from production functions are not
empirically relevant when compared with supply functions estimated by
regression analysis.
F-4
-------
The major hypotheses involved in our specification are as follows. First,
we postulate that producers base their production decisions on expected future
prices, which, we assume, equal current prices received by farmers. This hypo-
thesis concerning expectations is standard and predicts a positive relationship
between quantity-produced and the previous year's price received by farmers.
Second, we account for development and diffusion of new technology, introduction
of improved fertilizers and pesticides, new cultivation techniques, etc., by
including a trend variable in each of our supply-response functions. Once again,
we predict a positive sign for the regression coefficient of this variable.
Finally, we hypothesize that producers do not adjust their production perfectly
each year; we would expect a gradual adjustment to the desired or equilibrium
level of production.
The most convenient way of handling the imperfect adjustment phenomenon,
as well as obtain the basic supply-response regression equation, is to specify
(1) the supply-response function as a desired or equilibrium level of producti6n
equation and (2) the adjustment function. Stated formally,
where Q? = desired, long-run equilibrium quantity of the i — grass
seed produced in Oregon in year t,
T = time trend,
PA ^ = ?it lagged one year,
h. = disturbance term,
«Jt - «lt-i ' *J < - <$.t-i>
where 0 < X? < 1. The adjustment equation indicates that the actual change in
production 1 of a particular seed is some fraction of the difference between
the actual production in the previous year and the current desired level of pro-
duction. Substituting the desired-level-of-production supply-response equation
for Q?* in the adjustment function, one obtains the required regression
equation
which enables us to estimate short- and long-run price-elasticities of supply-
response. Of course, allowing for such an adjustment function creates special
problems because there is likely to be a high sample correlation between qi(_ and
-------
F. 2 Estimated Demand and Supply-Response Functions
In this section we report time-series, multiple-regression estimates
of demand and supply-response functions for six major grass seed crops (rye-
grass, tall fescue, red fescue, chewings fescue, Merion-Kentucky bluegrass,
and bentgrass) raised in Oregon. No function was estimated for orchardgrass
because data on Oregon production unfortunately is available for only 12
years, too few for a meaningful statistical analysis to be performed. The
time period of analysis varied among the grass seed crops: 1947-69 for
ryegrass, 1956-69 for Merion-Kentucky bluegrass, and 1950-69 for the remain-
ing four crops. Although both additive and multiplicative functional forms
were employed to estimate both demand and supply-response functions by
ordinary least squares (OLS), the two stage least squares estimation tech-
nique (TSLS) Was only used in the case of demand functions of the additive type.
The results are reported in Tables F. 1 through F. 7.
Demand for Utilisation. Since ryegrass and tall fescue are principally
used as livestock forage grasses, the number of cattle in the twelve southern
U.S. states and three foreign countries (Canada, West Germany, and France)
became our domestic and foreign demand-shift variables, respectively. For
the other four grasses which are chiefly used for turf, new housing starts
were hypothesized to be responsible for major shifts in demand. Not all of
these specified exogenous variables appear in the estimated equations re- ,
ported in Tables F. 1 - F.6, however. In our early regression runs we dis-
covered that the number of cattle in the three foreign countries was strongly
correlated with its domestic equivalent; the foreign demand-shifter was
deleted after we found that recasting the data in logarithmic first differences
did not eliminate the problem of multicollinearity. Certain other exogenous
variables were also deleted from some equations. In each instance, the regres-
sion coefficient was not statistically significantly different from zero, or
we again confronted the problem of multicollinearity.
Nonetheless, with the exception of tall fescue, the model performed
reasonably well, and the results were consistent with our theoretical expec-
tations. For most seeds, TSLS performed rather better than OLS when judged
by conventional statistical tests. However, in the case of red fescue (see
Table F.3), TSLS increased the statistical significance of the own-price and
price-of-substitute regression coefficients but resulted in a clearly unreal-
istic short-run own-price elasticity of demand. Consequently, a priori
reasoning and knowledge of the price-elasticities for other seeds suggested
that the TSLS red fescue regression coefficients were unreliable.
Typically the coefficients of adjustment (i.e., the regression coeffi-
cient of the lagged dependent variable) had its theoretically anticipated sign
and were statistically significant at the 10 percent level or better. It
should be noted, however, that specification changes not only affected the
statistical significance of the adjustment coefficient but also its sign. In
the case of ryegrass (see Table F.I), experimentation by deletion was under-
taken because of the multicollinearity that existed between the lagged depen-
dent variable and the demand shifter; the results reported in Table F. 1 only
indicate the deletion of the shift variable. In contrast, the demand shifters
were not statistically significant for red and chewings fescues (see Tables
F.3 and F.4).
F-6
-------
Table F.I. Estimated Demand Relationships for Ryegrass, 1347-69
Form at
method
estimat
Demand
O.L.S.
O.L.S.
L.L.S
L.L.S.
TSLS
TSLS
Demand
O.L.S.
L.L.S.
TSLS
id Regression Coefficients
-,f
:ion Z P.
for Utilization
8.7482a -11.
(5.226) (5.
- 6.
(2.
1.97323 - 0.
(4.088) (4.
- 0.
(2.
10.80063 -17.
(10.575) (11.
-20.
(9.
for Ending Stocks
-3.
(1.
-0.
(1.
-5.
(1.
5535a
504)
688 3b
302)
6132a
464)
3830b
262)
3563a
290)
3675a
197)
0136°
833)
6484°
977)
8808d
427)
*J
0.
(0
-0.
(0.
0.
(0.
-0.
(0.
1.
(3.
1.
(2.
t
6154
.929)
2035
205)
0558
784)
1369
147)
2263a
108)
0253b
090)
for
d
-0.
(0.
0.
(6.
0.
(0.
0.
(7.
-0.
(3.
0.
(6.
Qca
0461
289)
7027a
508)
1386
890)
71293
849)
31553
096)
4277a
834)
0.2142a
(4.419)
1.1317a
(6.083)
0.1524
(1.125)
Degrees Short-run Long-run Durbin
Inter- 2 F- of price price Watson
ceptor Statistic Freedom Elasti- Elasti- Statis
city city tic
-49.
(1.
88.
(3.
-0.
(1.
0.
(3.
-43.
(2.
200.
(10.
22.
(1.
-0.
(0.
51.
(1.
4281 .9311 60.8040
538)
4874 .8265 30.1762
124)
6408 .9285 58.4714
471)
9395 .8622 39.7400
462)
4817 .9771 192.1912
345)
8887 .9593 149.3020
230)
1242 .7883 37.2448
246)
4438 .8488 56.2071
713)
5389 .7708 33.6249
063)
18 -0.6412 2.0938*
19 -0.3712 -1.2486 2.6513**
18 -0.6132 -0.7119 1.9483*
19 -0.3830 -1.3341 2.5450**
18 -0.9633 2.7137**
19 -1.1304 -1.9752 2.1193*
20 -0.5810 2.1340*
20 -0.6484 2.4458*
20 -1.1326 1.6872*
Note: Figures in parentheses are t-statistic.
P. = Average price per 100 Ibs of timothy seed received by farmers.
a, b, c, and d indicate significance at the 1, 5, 10, and 20 percent levels, respectively.
*Reject the hypothesis that auto-correlation is present at the 95 percent confidence level.
**The Durbin-Watson Statistic is inconclusive.
-------
Table F.2. Estimated Demand Relationships for Tall Fescue;' 1950-69
Form and
method of
estimation
Demand for
O.L.S.
O.L.S.
L.L.S.
L.L.S.
TSLS
TSLS
Demand for
O.L.S.
L.L.S.
TSLS
Regression Coefficients for
4
Utilization
2.9439b
(2.392)
2.3215°
(1.851)
2.5472d
(3.003)
2.2049b
(2.617)
0.5938
(0.612)
0.6961
(0.863)
Pit
-0.0562
(0.308)
-0.0904
(0.468)
-0.1008
(0.922)
-0.1488d
(1.382)
-3.2594d
(4.358)
-3.1723a
(5.309)
»it
-------
Table F.3. Estimated Demand Relationships for Red Fescue, 1950-69
Form and
aethod of
estimation
Demand for
O.L.S.
O.L.S.
L.L.S.
TSLS
TSLS
Desand for
O.L.S.
L.L.S.
TSLS
Regression
Zt ?i
Utilization
2.6589 -0
(0.447) (1
-0
(1
-0
(1
-0.6039 -4
(0.417) (16
2.9499 -0
(0.478) (0
Ending Stocks
-0
(1
-0
(1
-0
(1
t
. 3844d
.508)
.3882d
.565)
.6841d
.447)
.5764a
.332)
.0365
.412)
.0839°
.697)
.5239d
.396)
.0680
.281)
Coefficients
'j« "'
-0.3562d 0
(1.385) (3
0.3603d 0
(1.441) (3
0.7274d 0
(1.544) (6
4.3649a-0
(16.245) (0
0
(3
for
d •, QCa
.6757a
.772)
.6609a
.859)
.7187a
.828)
.0210
.321)
.6982a
.660)
0.1540d
(1.356)
0.7525b
(2.094)
0.1769d
(1.483)
Inter- R2
ceptor
3.
(0.
7.
(1.
0.
(0.
31.
10.
22.
(0.
4.
(1.
0.
(0.
3.
(1.
3987 .7471
330)
4485 .7435
564)
3153 .8662
985)
3637 .9853
1866)
7394 .7089
211)
8231 .4391
462)
4586 .4970
515)
9104 .3997
114)
F-
Statistic
10.3372
14.4897
32.1746
235.1980
12.1779
6.2624
7.9094
5.3239
Degrees
of
Freedom
14
15
15
14
15
16
16
16
Short-run
price
Elasti-
city
-0.7605
-0.7680
-0.6841
-9.0539
-0.0722
-0.5298
-0.5239
-0.4294
Long-run Durbin
price Watson
Elasti- Statis-
city tic
-2.3450 1.9487*
-2.3682 2.0162*
-2.4319 2.6670**
1.1581**
-0.2392 2.0914*
1.8962*
2.1462*
1.7345*
Note: Figures in parentheses are t-statistic.
P. = Average price per 100 Ibs of chewings fescue received by farmers.
a, b, c, and d indicate significance at the 1, 5, 10, and 20 percent levels, respectively.
*Reject the hypothesis that auto-correlation is present at the 95 percent confidence level.
**The Durbin-Watson statistic is inclusive.
-------
Table F.4. Estimated Demand Relationships for Chewings Fescue, 1950-69
Form and Regression Coefficients for
method of ,
estimation Z P P 0
t it • jt Ht
Demand
O.L.S.
for Utilization
1.2321 -0
-2434b
(0.536) (2.237)
O.L.S.
L.L.S.
L.L.S.
TSLS
TSLS
Demand
O.L.S.
L.L.S.
TSLS
-0
(2
0.1708 -0
(0.274) (1
-0
(1
1.7758 -0
(0.850) (3
-0
(3
for Ending Stocks
0
(0
-0
(0
-0
(0
,2398b
.262)
.9567d
.433)
.9627d
.489)
.6603a
.061)
.6356a
.002)
.0156
.377)
.3499
.916)
.0010
.022)
0
(1
0
(1
0
(0
0
.1436d
.422)
.1430d
.451)
.4614
.719
.4760
(0.769)
0
(2
0
(2
.5195b
.655)
.5010b
.601)
0.
(2.
0.
(2.
0.
(2.
0.
(2.
0.
(0.
0.
(0.
ca
3268b
145)
3332b
248)
4404b
793)
4403b
833)
1073
629)
1267
756)
0.4985a
(3.012)
0.7928°
(2.015)
0.4418b
(2.532)
5
(1
7
(4
1
(3
1
(3
6
(2
9
(5
-2
(0
0
(0
-1
(0
Inter- 2 F-
ceptor Statistic
.5782 .7390 9.9084
.627)
.2219 .7336 13.7698
.846)
.1303 .7363 9.7570
.628)
.1443 .7349 13.8900
.844)
.8090 .7878 12.9913
.167)
.0863 .7768 17.4031
.583)
.1670 .6256 13.3698
.769)
.1935 .6360 13.9875
.215)
.0972 .6223 13.1818
.365)
Degrees Short-run Long-run Durbin
of price price 'Watson
Freedom Elasti- Elasti- Statis-
city city tic
14 -1.0572 -1.5704 2.9279**
15 -1.0415 -1.5620 2.9433**
14 -0.9567 -1.7096 2.7118**
15 -0.9627 -1.7200 2.7223**
14 -2.8679 -3.2126 2.4142**
15 -2.7606 -3.1611 2.4749*
16 0.1367 1.3352**
16 -0.3499 1.2359**
16 -0.0087 1.2782**
Note: Figures in parentheses are t-statistic.
P = Average price per 100 Ibs of red fescue received by farmers.
a, b, c, and d indicate significance at the 1, 5, 10, and 20 percent levels, respectively.
*Reject the hypothesis that auto-correlation is present at the 95 percent confidence level.
**The Durbin-Watson statistic is inconclusive.
-------
Table F.5. Estimated Demand Relationships for Merion-Kentucky Bluegrass, 1956-69
Form and
method of
estimation
Demand for
O.L.S.
O.L.S.
L.L.S.
TSLS
Demand for
O.L.S.
L.L.S.
TSLS
Regression Coefficients for
Zt
Utilization
2.4989°
(2.070)
3.5403b
(2.919)
0.8754
(1.231)
3.0925b
(3.038)
Ending Stocks
Pit
-0.0335a
(4.502)
-0.04243
(6.475)
-0.6799b
(2.288)
-0.0449a
(5.880)
-0.0251
(0.361)
-0.0519
X0.155)
-0.0138d
(4.692)
d ca Inter- 2
d c
0.0390 0.2652 0.9127 .8870
(1.656) (1.915) (0.628)
0.0507° 0.7409 .8410
(1.981) (0.454)
0.1678 0.3312° 1.2831 .8118
(0.722) (2.198) (3.351)
0.0622b 0.1333 0.8712 .9241
(2.898) (1.063) (0.731)
0.44563 -0.2839 .8755
(4.489) (0.264)
1.2680a -0.5782 .8456
(4.404) (0.717)
0.3110b 1.3887 .9001
(2.857) (1.121)
Degrees Short-run Long-run Durbin
F- of price price Watson
Statistic Freedom Elasti- Elasti- Statis-
city city tic
17.6527 9 -0.8523
17.6040 10 -1.0788
9.7039 9 -0.6797
27.4052 9 -1.1424
38.68 11 -1.1875
49.5247 11 -0.0519
11 -0.6529
-1.1600 2.0916*
-1.1360 1.4788**
-1.0163 2.8123**
-1.3181 2.9287**
2.3606*
2.6865**
2.5407*
Note: Figures in parentheses are t-statistic
P.t = Average price per 100 Ibs of chewings fescue received by farmers.
a, b, c, and d indicate significance at the 1, 5, 10, and 20 percent levels, respectively.
*Reject the hypothesis that auto-correlation is present at the 95 percent confidence level.
**The Durbin-Watson statistic is inconclusive.
-------
Table F.6. Estimated Demand Relationships for Bentgrass, 1950-69
Form and Regression Coefficients for
method of
estimation Z
Demand
O.L.S.
O.L.S.
L.L.S.
L.L.S.
TSLS
Demand
O.L.S.
L.L.S.
TSLS
for Utilization
1.1803
(0.549)
0.3131
(0.629)
0.3134
(0.651)
5.1896a
(4.068)
for Ending Stocks
Pit
-0.0689b
(2.433)
-0.0621b
(2.497)
-0.3692b
(2.099)
-0.3669b
(2.183)
-0.18993
(7.850)
0.0141
(0.896)
0.2759
(0.917)
0.0418a
(3.965)
>lt <-,
-------
Table F.7. Estimated Supply Response Relations for Selected Grass Seed Crops
Type of Form and
Regression Coefficients for
grass seed method of
estimation T P . Q.
Ryegrass O.L.S.
L.L.S.
Tall Fescue O.L.S.
L.L.S.
Red Fescue O.L.S.
L.L.S.
Chewings O.L.S.
Fescue
L.L.S.
Merion- O.L.S.
Kentucky
Bluegrass
L.L.S.
Bent grass O.L.S.
L.L.S.
4.3355a
(3.310)
1.9667a
(2.976)
0.3234a
(4.655)
3.0576a
(3.988)
0.0528
(0.580)
1.0001
(1.087)
0.0361
(0.312)
2.1467
(1.143)
9.4672b
(2.563)
0.70283
(3.424)
0.1693a
(4.557)
0.6092a
(4.002)
0.0242
(0.962)
0.3075
(1.237)
-0.0180
(0.310)
0.0233
(1.049)
0.0059
(0.544)
0.6317b
(2.180)
0.0188
(0.738)
0.2364d
(1.567)
0.4455b
(2.091)
0.6148a
(3.351)
0.62333
(5.635)
0.65823
(5.548)
0.8954s
(5.476)
0.83653
(5.071)
0.2683
(0.801)
0.3442
(0.084)
0.6438
(0.725)
0.6902b
(2.239)
0.8092a
(4.449)
0.84603
(7.110)
Inter- R2 F-
ceptor Statistic
245.3570 .7890 41.6731
-3.2355 .8494 11.2305
-19.4357 .8930 41.7161
-5.8607 .8462 27.4995
-0.0511 .7045 19.0690
(0.038)
-0.3180 .6695 17.6584
(0.745)
2.4719 .2361 1.6487
(0.344)
-1.2720 .3047 2.3367
(0.722)
-2.2675 .3710 1.9662
-5.0107 ,5339 3.8178
0.6093 .6583 15.4133
(0.312)
-0.2307 .8200 36.4598
(0.754)
Degrees Short-run Long-run Durbin
of price price Watson
Freedom Elasti- Elasti- Statis-
city city tic
19 0.5245 0.9459 2.0758*
19 0.7028 1.8245 1.9004*
15 0.4079 1.0828 2.1602*
15 0.6092 1.7823 2.6909**
16 0.1774 1.6960 2.6978**
16 0.3075 1.8807 2.8287**
16 -0.0744 2.4002*
16 0.0233 0.0355 2.4961*
10 0.3787 1.0632 2.0674*
10 0.6317 2.0390 2.3250*
16 0.1265 0.6630 2.3454*
16 0.2364 1.5351 2.5363**
Note: Figures in parentheses are t-statistic.
a, b, and d indicate significance at the 1, 5, and 20 percent levels, respectively.
*Reject the hypothesis that auto-correlation is present at the 95 percent confidence level.
**The Durbin-Watson statistic is inconclusive.
-------
Demand for Ending Stocks. Although our empirical results for ending-
stock demands are less robust than those obtained for utilization demands
(compare lower and upper halves of Tables F.1-F.6), none of our major hypotheses
were generally rejected. Theoretical considerations suggested that, other
things equal, demands for ending stocks would vary directly with total commer-
cial availability (current production plus beginning stocks) and inversely with
current price.
For all grasses, the regression coefficients for total commercial avail-
ability have their expected positive signs and are significantly different
from zero at the 20 or better percent level. In contrast, however, only four
grass seeds (ryegrass, red fescue, chewings fescue, and Merion-Kentucky blue-
grass) had own-price regression coefficients with the anticipated negative
signs. Tall fescue and bentgrass both had unexpected positively-signed own-
price regression coefficients, but these coefficients are statistically insig-
nificant except when estimated by TSLS. Unfortunately, a diligent search did
not reveal the cause(s) of these peculiar findings for tall fescue and bent-
grass.
For the grasses with negatively-signed own-price regression coefficients,
estimated own-price elasticities of demand for ending stocks varied from -0.52
to -1.13.* Our best estimates for individual grasses are as follows:
Ryegrass -0.65
Red Fescue -0.53
Chewings Fescue -0.35
Merion-Kentucky Bluegrass -0.65
Estimated Supply-Response Functions. Table F.7 reports our estimated
supply-response relationships for the six grass seed types under investigation
here. With the exception of chewing fescue, the model performed reasonably
well when evaluated by the usual statistical tests, and both OLS and LLS re-
gression estimates are rather similar. In the case of chewings fescue, five
of six estimated regression coefficients have their theoretically expected
signs, but none are statistically significantly different from zero. For the
other five grasses, however, all but two estimated regression coefficients
have their expected positive signs and are statistically significant at the
5 percent level: only the trend regression coefficient for Merion-Kentucky
bluegrass and the coefficient-of-adjustment regression coefficient for bent-
grass are not statistically significant.** Thus, none of our basic hypotheses
concerning grass seed producers supply-response can be rejected on the basis
of results reported in Table F.7.
*
This range of elasticity estimates does not include estimates derived
from statistically insignificant regression coefficients.
**
Problems of multicollinearity required the deletion of the trend
variable from the estimated functions for red fescue and bentgrass.
F-14
-------
Prices of several substitute seeds were included in our initial re-
gression experiments for each of the six grass seeds. Here, however, we
present regression results only for those substitute seed prices which
provided the best statistical results. As Table F.2 indicates, in the case
of tall fescue, all substitute seed prices proved to have the theoretically
unexpected sign or to be statistically insignificant. In contrast, the
regression coefficients for substitute seed prices did have the expected
sign and were statistically significant at a probability level of at least
20 percent for the other five seeds.
With the single exception of tall fescue, estimated own-price regres-
sion coefficients have their theoretically anticipated sign and are statis-
tically significant at the 20 or better percent level. Although our tall
fescue results are disappointing, they are relatively easy to explain.
Tall fescue is an important export crop, but unfortunately there is no data
on exports prior to 1966. Therefore, the inclusion of exports in data
measuring utilization may be the reason for the insignificant own-price
regression coefficients when estimated by OLS and LLS. Interestingly,
estimation of the tall fescue demand function by TSLS did result in a sta-
tistically significant own-price regression coefficient, but this strong
result is negated by the theoretically unexpected sign for the coefficient
of adjustment.
As one would expect, estimated short- and long-run own-price elasti-
cities of demand vary with the assumed functional form, estimation method,
and among grass seed types. When estimated at the mean, short-run elasti-
cities range from -0.37 to -2.87, and long-run elasticities vary from -0.71
to -3.21.* However, with the exception of the TSLS estimations of chewings
fescue demand reported in Table F.4, our estimated short- and long-run own-
price elasticities of demand for grass seeds are less than -1.30 and -2.50,
respectively.
Our best estimate of the long-run own-price elasticities of demand
for each seed crop are as follows:
Ryegrass -1.98
Red Fescue -2.37
Chewings Fescue -1.57
Merion-Kentucky
Bluegrass -1.32
Bentgrass -1.67
Unfortunately, as indicated previously, we were unable to estimate a satisfac-
tory demand function for tall fescue.
The range of elasticity estimates reported in the text does not encom-
pass own-price elasticity estimates derived from statistically insignificant
regression coefficients.
F-15
-------
Excluding estimates based on statistically insignificant regression
coefficients, our estimated short-run price-elasticities range from 0.24 for
bentgrass to 0.70 for ryegrass. Our best estimates of long-run price elasti-
cities are as follows:
Ryegrass 1.82
Tall Fescue 1.78
Red Fescue 1.88
Merion-Kentucky
Bluegrass 2.04
Bentgrass 1.54
These rather low long-run price-elasticity estimates are consistent with in-
dependent evidence that alternative crops or land uses to grass seed produc-
tion in the Willamette Valley are limited-*
F.3. Calculated Demand and Supply Functions
Tables F.8 and F.9 present the linear demand and supply functions
which underlie our computations of the changes in output, price, consumers'
surplus, and rents associated with alternative field burning control policies.
Each equation is calculated to satisfy two conditions. Each generates a
linear demand or supply curve that (1) passes through the 1965-69 average
observed prices and quantities with (2) a price-elasticity at that point equal
to the best multiple-regression based elasticity estimate for that seed type
reported earlier in this appendix. The first seven columns of Tables F.8 and
F.9 present the price, quantity and price-elasticity data used to calculate
all but one of the desired equations. Since we did not estimate non-Oregon
supply functions, we required that function to have a constant term equal to
Fisher's estimate of average variable costs for high-cost Willamette Valley
producers.**
*
For evidence concerning the limited alternatives available to farmers
now raising grass seeds, see Frank S. Conklin and R. Carlyle Bradshaw, Farmer
Alternatives to Open Field Burning: An Economic Appraisal, Special Report 336,
Agricultural Experiment Station, Oregon State University, Corvallis, October
1971, especially pp. 4-8 and 14.
**
The constant terms for the non-Oregon supply functions in Table F.9
were obtained from sample data reported in Douglas E. Fisher, An Economic
Analysis of Farms Producing Grass Seed in the Willamette Valley3 with Special
Attention to the Cultural Practice of Field Burning, unpublished Ph.D. disser-
tation, Oregon State University, June 1972.
F-16
-------
Table F.8. Linear Demand Functions for the 1965-69 Period, by Seed Type
Average Prices and Quantites, 1965-69
Quantities-Demanded Own-Price-Elasticities Demand Function Equations
Seed Utili- Ending Aggre- Utili- Ending
Type Prices Aggregate zation Stocks gate zation Stocks Aggregate Utilization Ending Stocks
(per (million (million (million
100 Ibs) Ibs) Ibs) Ibs)
Ryegrass 6.86 164.289 158.585 5.704 -1.94 -1.98 -0.65 P- 10.4074-0.0216Q P- 10.3247-0.0216Q P-17.4140-0,2706Q
Tall 12.08 74.743 60.040 14.703 -1.49 -1.85 0 P- 20.2085-0.1088Q P- 18.6098-0.1088Q Q6S=14 703
fescue
Red 26.50 26.164 20.094 6.070 -1.94 -2.37 -0.53 P- 40.1375-0.5212Q P- 37.6812-0.5565Q P-76.5000-8.2372Q
fescue x
Chewings 26.70 11.760 9.235 2.525 -1.53 -1.57 -0.35 P- 48.3591-1.8416Q P-.43.7090-1.84160 QeS-2 525
fescue
Merion- 68.90 7.262 4.512 2.750 -1.07 -1.32 -0.65 P-133.4991-8.8968Q P-121.1551-57410 P-175.1931-38 61000
Kentucky " x
Bluegrass
Bentgrass 37.70 9.962 7.444 2.518 -1.25 -1.67 0 P- 67.9208-3.0331Q P- 60.25113-3.0331Q Qes=2.518
Note: Functions are caluclated according to procedure described in text.
-------
Table F.9. Linear Supply Functions for the 1965-69 Period, by Seed Type
Average Prices and Quantities, 1965-69
Quantities-Supplied Price-Elasticities Supply Function Equations
Non- Non-
Seed Prices Aggregate Oregon Oregon Aggre- Oregon Oregon Aggregate Oregon Non-Oregon
Type (per (million (million (million gate
100 Ibs) Ibs) Ibs) Ibs)
Ryegrass 6.86 164.289 158.585 1.73 1.82 P- 2.9044+0.0241Q P- 3.0908+.0238Q
Tall 12.08 74.743 10.248 49.793 1.48 1.78 1.86 P- 3.9195+0.1091Q P- 5.2933+0.6622Q P= 5.5700+0.1307Q
fescue
Red 26.50 26.164 6.724 13.370 1.10 1.88 1.45 P- 2.4316+0.9200Q P-12.4048+2.0964Q P- 8.2202+1.3672Q
fescue
Chewings 26.70 11.760 6.800 2.435 1.35 1.75 1.44 P» 6.9446+1.6798Q P-12.0302+2.1575Q P- 8.2178+7.5873Q
fescue
M K blue 68.90 7.262 1.760 2.752 0.70 2.04 1.18 P—30.1719+13.6426Q P<=35.1324+19.1939Q P-10.5202+21.2314Q
grass
Bentgrass 37.70 9.962 7.048 0.396 1.15 1.54 1.63 P= 5.0367+3.2787Q P-13.2195+3.4734Q P-14.5205+58.4795Q
Note: Functions are calculated according to procedure described in text.
-------
APPENDIX G
-------
Table C.I. KmlmnU'il Cuiiuumern' Surplus, OI-PROII *ml Noii-OrtRoii Kt>iu« (or Kymrass In Supply Situation 1 Without oml^WHh
Allrnmllvn (>p,m Vlul tot tlm Mobile I'lclil S.-inl I Ir.vf
Of an FlPid Burning Control lyl < yA*J*SPJJUL*!lt. -C'."-(iA PCI' Acre
J""1^ Ban on Burning Once In Three Years Burnlng_ Alternate Year ..Burning
Situation ^$9 $TT ~~$5"$T $13 $5 $9 513
Price per 6.86 7.0568 7.2132 7.36971 6.9915 7.0959 7.2002 6.9589 7.0372 7.1154
100 Ibs.
Average 2.68000 3.09390 3.42494 3.74604 2.9595 3.1766 3.3974 2.8870 3.0525 3.2180
variable
cost per
100 Ibs.
Aggregate 1,643,890 1,551,225 1,478,788 1,406,337 1,581,416 1,533,125 1,484,823 1,596,510 1,560,291 1,524,066
Output
(100 Ibs)
Oregon 1,585,850 1,494,987 1,421,511 1,348,068 1,525,610 1,476,629 1,427,635 1,540,918 1,504,184 1,467,479
Output
(100 Ibs)
Non-Oregon .
Output
(100 Ibs)
Consumers' 2,913,994 2,598,767 2,361,765 2,136,008 2,700,948 2,538,510 2,381,077 2,752,758 2,629,270 2,508,605
Surplus
Change in 315,227 552,229 777,986 213,046 375,484 532,917 161,236 284,724 405,389
Consumers*
Surplus
Percentage 10.82 18.95 26.70 7.31 12.89 18.29 5.53 9.77 13.91
Change
Oregon 6,628,853 5,924,484 5,384,996 4,871,473 6,151,271 5,787,249 5,429,039 6,274,587 5,993,722 5,719,338
Rents
Change in 704,369 1,243,857 1,757,380 477,578 841,604 1,199,814 354,266 635,131 909,515
Oregon
Rents
Percentage 10.63 18.76 26.51 7.20 12.70 IB.10 5.34 9.58 13.72
Change
Bon-Oregon
Rents
Change in ^ ^___
Non-Oregon
Rents •
Percentage
Change
Sum of 9,542,847 8,523,251 7,746,761 7,007,481 8,852,219 8,325,759 7,810,116 9,027,345 8,622,992 8,227,943
Surplus
and Rents
Change in 1,019,558 1,796,086 2,535,366 690,628 1,217,088 1,732,731 515,502 919,855 1,514,904
Sum
Percentage 10.68 18.82 26.57 7.24 12.75 18.16 5.40 9.64 15.87
Change
aSee Part 111, sections A and R for a discussion of definitions and computation methods.
-------
Tnble C.2. Estimated Consumers' Surplus, Oregon and Non-Oregon Konts for Tall Fescue in Supply Situation I Without and With
Alternative Open Field Burning Control Policies at Varying Costs per Acre for the Mobile Flrld Sanitizer"
.
Slt'">tl<>"
$5
Open Field Burning Control Policies anJ Mobile Santtlzt-r Cost per Acre
Ban on Burning _ Once in Three Years Burning _ Alternate Year Burning
' ' ~
$9
$13
?5
$9
$5
?9
$13
Price per
100 Ibs.
Average
variable
cost per
100 Ibs.
Agregate
Output
(100 Ibs)
Oregon
Output
(100 Ibs)
12.08
12.3703
4.66780 5.25880
Non-Oregon 497,930
Output
(100 Ibs)
12.60639
5.73162
12.84247 12.2719 12.4239 12.5867
6.20444 5.0618
5.3770 5.6922
12.2228 12.3408 12.4588
4.9633 5.1997 5.4361
747,430 720,422 698,723 677,024 729,463 714,997 700,531 733,983 723,134 712,285
102,480 97,946 94,371 90,796 99,437 97,053 94,669 100,181 98,393 96,606
520,298 538,362 556,425 512,773 524,814 536,857 509,009 518,041 527,073
Consumers' 3,037,742 2,823,406 2,655,855 2,493,490 2,894,713 2,781,045 2,669,647 2,930,702 2,844,701 2,759,998
Surplus
Change in
Consumers '
Surplus
Percentage
Change
214,336 381,887 544,252 143,029 256,700 368,095 107,040 193,041 277,244
7.06
12.57
17.92
4.71
8.45
12.12
3.52
6.35
9.14
Oregon
Rents
Change in
Oregon
Rents
Percentage
Change
759,602 696,543 648,779 602,707
63,059 110,823 156,895
8.30 14.59
20.65
716,955 684,448 652,695
42,647 75,158 106,907
5.62
9.90 14.08
727,259 702,633 678,437
32,343 56,969 81,165
4.26 7.50 10.69
Non-Oregon 1,620,762 1,769,091 1,894,063 2.023,292 1,718,287 1,799,934 1,883,488 1,693,155 1,753,766 1,815,461
Rents
Change In
Non-Oregon
Rents
Percentage
Change
148,336 273,301 402,530 97,525 179,172 262,726 72,393 133,004 194,699
9.15
16.86
24.84
6.02
11.15
16.21
4.47
8.21
12.01
Sum of 5,418,106 5,289,047 5,198,697 5,119,489 5,327,955 5,265,427 5,205,830 5,351,116 5,301,100 5,253,896
Surplus
and Rents
Change in
Sum
Percentage
Change
129,059 219,409 298,617 88,151 152,679 212,276
2.39 4.05 5.51 1.63 2.82 3.92
66,990 117,006 164,210
1.24 2.16 3.03
See Part lit, sections A and B for a discussion of deinitions and computation methods.
-------
Tab),. U.3. Kstlmuu-d Consumes' Sutplits. Oregon un.l Non-Or.-Kon R™t« for Ri'J tVscuu In Supply Situntiou I Without and With
Alternative Op#n Held Horning Control I'ulicJt-K at Vaiylng Costs por Acre for the Mobiln Meld SunUlzi-r*
— S£ai_LL°iiJ!!iaUj!i£L£gIlU°l-ggii£iiL'L-P-"'' Mobile SanltiEcr Cost.J>«r_Acrg_
Initial
Situation
?5
Ban on Burning
$9
$13
Once in Thrcc Years burning
$5 59" $13
Alternat e Year Burning
$5 " $9 "$13
Price per 26.50
100 Ibs.
Average
variable
cost'per
100 Ibs.
Agrega te
Output
(100 Ibs)
Oregon
Output
(100 Ibs)
6.7446
Non-Oregon 133,700
Output
(100 Ibs)
26.8243 27.0826 27.3409 26.7167 26.8889 27.0611 26.6629 26.7920 26.9212
7.6375 8.3517 9.0660 7.3399 7.8160 8.2922 7.1911 7.5482 7.9053
261,640 255,433 250,477 245,521 257,498 254,194 250,890 258,531 256,053 253,575
67,240 64,653 62,347 60,172 65,428 63,978 62,528 65,882 64,794 63,707
136,073 137,960 139,849 135,283 136,543 137,802 134,890 135,834 136,779
Consumers' 1,784,058 1,700,313 1,634,972 1,570,912 1,727,917 1,683,859 1,640,369 1,741,805 1,708,575 1,675,664
Surplus
Change in
Consumers'
Surplus
Percentage
Change
83,745 149,086 213,146 56,141 100,199 143,689 42,253 75,483 105,394
4.69
8.36
11.95
3.15
5.62
8.05
2.37
4.23
6.08
Oregon 1,328,353 1,238,183 1,167,815 1,099,639 1,267,787 1,220,243 1,173,581 1,282,842 1,246,886 1,24,444
Rents
Change in
Oregon
Rents
Percentage
Change
90,170 160,538 228,714 60,566 108,110 154,772 45,511 81,467 116,909
6.79
12.09
17.22
4.56
8.14
11.65
3.43
6.13
8.80
Non-Oregon 1,222,018 1,265,769 1,301,130 1,337,008 1.251,130 1,274,540 1,298,157 1,243,866 1,261,342 1,278,951
Rents
-------
Table G.4. Estimated Consumers' Surplus, Oregon and Non-Oregon Rents for Chewings Fescue in Supply Situation I Without anJ
With Alternative Open Field Burning Control Policies at Varying Costs per Acre for tlie Mobile Field Sanlti7.era
Initial
Situation
Price per 26.70
100 Ibs.
Average 6.74460
variable
cost per
100 Ibs.
Aggregate 117,600
Output
(100 Ibs)
Oregon 68,000
Output
(100 Ibs)
Open Field Burning Control Policies and Mobile Sanitizer Cost per Acre
Ban on Burning Once in Three Years Burning Alternate Year Burning
$5 $9 S13 $5 $9 $13 $5 $9 $13
27.1674 27.54091 27.91443 27.0117 27.2607 27.5098 26.9339 27.1206 27.3074
7.63750 8.35174 9.06602 7.3399 7.8160 8.2922 7.1911 7.5482 7.9053
115,072 113,044 111,016 115,918 114,565 113,213 116,340 115,326 114,312
66,022 64,443 62,864 66,680 65,628 66,455 67,009 66,220 65,430
Ron-Oregon 24,350
Output
(100 Ibs)
24,975 25,468 25,960 24,770 25,099 25,427
24,668 24,914 25,160
Consumers' 1,273,555 1,219,286 1,176,686 1,134,843 1,237,275 1,208,568 1,180,209 1,246,306 1,224,674 1,203,231
Surplus
Change in
Consumers'
Surplus
Percentage
Change
54,269 96,869 138,712 36,280 64,987 93,346 27,249 .: 3,881 70,324
4.26
7.61
10.89
2.85
5.10
7.33
2.14
3.84
5.52
Oregon 1,356,967 1,289,403 1,236,608 1,184,924 1,311,718 1,276,116 1,268,872 1,322,947 1,296,088 1,269,479
Rents
Change in 67,564 120,359 172,043 45,249 80,851 88,095 34,020 60,879 77,488
Oregon
Rents
Percentage
Change
4.97
8.87
12.68
3.33
5.96
6.49
2.51
4.49
5.71
Non-Oregon 225,021
Rents
Change in
Non-Oregon
Rents
Percentage
Change
236,634 246,061 255,665 232,763 238,980 245,269
11,613 21,040 30,644 7,742 13,959 20,248
5.16
9.35
13.62
3.44
6.20
9.00
230,845 235,473 240,148
5,324 10,452 «*l5,127
2.59
4.64
6.72
Sum of 2,855,543 2,745,323 2,659,355 2,575,432 2,781,756 2,723,664 2,694,350 2,800,098 2,756,235 2,712,858
Surplus
and Rents
Change in 110,220 196,188 280,111 73,787 131,879 161,193 55,445 99,308 142,685
Sum
Percentage
Change
3.86
6.87
9.81
2.58
4.62
5.64
1.94
3.48
5.00
See Part III, eectlons A and B for a discussion of definitions and computation methods.
-------
Tablo G.5. KotiindtuJ Consumer!)' Surplus, Oregon *i\J Nuii-Orogoii RfiHt lot Merlon Kenturky Mucgruaa in Supply Situation I
Without and With Alirnuiiivi> Open Field Burning Control I'diclm «t Varying Costs per Acre tor the Mobile Fli-ld
Sanltlzer
Initial
S3 tuation
Price per 68.90
100 Ibs.
Average 7.52170
variable
cost per
100 Ibs.
Aggregate 72,620
Output
(100 Ibs)
Oregon 17,600
Output
(100 Ibs)
Non-Oregon 27,520
Output
(100 Ibs)
Consumers' 2,345,593
Surplus
Change in
Consumers1
Surplus
Percentage
Change
Oregon 1,080,258
Rents
Change in
Oregon
Rents
Percentage
Change
Son-Oregon 803,309
Rents
Change in
Don-Oregon
Rents
Percentage
Change
Sum of 4,229,160
Surplus
and Rents
Change in
Sum
Percentage
Change
Ban on Burning Once in Three Years Burning Alternate Year BurninR
$5 $9 $13 $5 $9 $13 $5 $9 513
69.2123 69.46657 69,72078 69.1064 69.2758 69.4454 69.0533 69.1806 69.3077
8.32690 8.97097 9.61509 8.05S5 8.4679 8.9173 7.9243 8.2463 8.5684
72,258 71973 71,687 72,377 72,187 71,996 72,437 72,294 72,151
17,336 17,133 16,930 17,483 17,285 17,150 17,463 17,362 17,260
27,644 27,764 27,883 27,594 27,674 27,754 27,569 24,914 27,689
2,322,618 2,304,307 2,286,038 2,330,277 2,318,042 2,305,805 2,334,130 2,324,923 2,315,739
22,975 41,286 59,555 15,316 27,551 39,788 11,463 20,670 29,854
1.00 1.76 2.54 0.65 1.18 1.70 0.49 0.88 1.27
1,055,509 1,036,471 1,017,588 1,067,300 1,051,065 1,038,057 1,067,498 1,057,940 1,048,359
24,749 43,787 62,670 12,958 29,193 42,201 12,762 22,318 31,889
2.29 4.05 5.80 1.20 2.70 3.91 1.18 2.07 2.95
811,245 818,294 825,345 808,514 813,002 817,705 806,850 810,364 813,883
7,936 14,985 22,036 5,005 9,693 14,396 3,541 7,055 10,574
1.0 1.87 2.74 0.62 1.21 1.79 0.44 0.88 1.32
4,189,372 4,159,072 4,128,971 4,206,091 4,182,109 4,161,567 4,208,478 4,193,227 4,177,981
39,788 70,088 100,189 23,069 47,051 67,593 20,682 35,933 51,179
0.94 1.66 2.37 0.55 1.11 1.60 0.49 0.85 1.21
See Part III, sections A and B for a discussion of definitions and computation methods.
-------
Tnblo C.6. Kstlui.iteil Consume™' Surplus, Oregon and Non-Orison Kcnts (or Buntgrass in Su|i[ily Situation 1 without and With
Alternative! Open Field Burning Control Policies at Varying Costs yci Acic for the Mobile Field Sanltlzer"
Open Field Burning Control Policies and Mobile Sanltjger Cost per Acre
Initial
Situation
Ban on Burn\n
~$9~
$13
Once jn Three Years Burning
$5 " "$9 513
Alternate Year Burning
$5
$9
$13
Price per
100 Ibs.
Average
variable
cost per
100 Ibs.
Aggregate
Output
(100 Ibs.)
Oregon
Output
(100 Ibs.)
Non-Oregon
Output
(100 Ibs.)
37.70 38.4153 38.9855 39.5559 38.1776 38.5579 38.9381 38.0588 38.3439 38.6290
9.8754 11.3591 12.5460 13.7330 10.8645 11.6558 12.4471 10.6173 11.2107 11.8042
99,620 97,279 95,398 93,518 98,062 96,808 95,555 98,454 97,514 96,574
70,480 68,267 66,492 64,717 69,007 67,823 66,640 69,377 68,488 67,601
3,960
4,086 4,182 4,281 4,045 4,110 4,175 4,025 4,074 4,123
Consumers1
Surplus
Change in
Consumers'
Surplus
Percentage
Change
1,505,298 1,435,135 1,380185 1,326,314 1,458,340 1,421,284 1,384,721 1,470,019 1,442,083
70,163 125,173 178,984 46,958 84,014 120,577 35,279 63,215
4.66
8.32 11.89
3.12
5.58
8.01
2.34 4.20
Oregon
Rents
Change in
Oregon
Rents
Percentage
Change
Non-Oregon
Rents
Change in
Non-Oregon
Rents
Percentage
Change
1,961,088 1,847,043 1,758,015 1,671,183 1,884,792 1,824,577 1,765,359 1,903,809 1,858,295
114,057 203,063 289.895 76,286 136,501 195,719 57,269 102,783
5.82 10.35 14.78
3.89 6.96
9.98
2.92 5.24
45,879 48,817 51,156 53.588 47,847 49,397 50,972 47,371 48,528
2,938 4,277 7,709 1,968 3,518 5,093 1,492 2,649
6.40 11.50 16.80
4.29
7.67 11.10
3.25 5.77
1,414,413
90,885
6.04
1,813,387
147,591
7.53
49,700
3,821
8.33
Sum of 3,512,255 3,330,995 3,189,356 3,051,085 3,390,974 3,295,258 3,201,052 3,421,199 3,348,906 3,277,500
Surplus
and Rents
Change in
Sum
Percentage
Change
181,240 322,899 461,170 121,276 216,997 311,203 91,056 163,349 234,755
5.16 9.19 13.13 3.45 6.18 8.86 2.59 4.65 6.68
8See Part III, sections A and B for a discussion of definitions and computation methods.
-------
Table G.7. Estimated Consul's Surplus, Oregon and Non-Oregon Rents for Variou3 Grass Seeds in Supply Situation U
Field Saniti^'a AU"na"ve 0"en "leid Burni«S Control Policies at Varying Costs per Acre for the Mobile
Ryegrass
Price per 100
Seed Type
Ibs.
Average variable costs per 100 Ibs.
Oregon output
(100 Ibs.)
Initial
Situation
6.8600
2.6800
1,585,850
Open Field Burning Control Policies and Mobile Sanitizer Cost per Acre
Ban on Burning Once in Three Years
$5 $9 $13 $5 $9 S13
3.0939 3.4249 3.7460 2.9595 3.1766 3.1974.
Alternate Year Burning
$5 $9 $13
2.8870 3.0525 3.2180
Change in Oregon rents
Percentage change in Oregon rents
Tall Fescue
Price per 100 Ibs. 12.0800
Average variable costs per 100 Ibs. 4.6678
Oregon output (100 Ibs.) 102,480
Change in Oregon rents
Percentage change in Oregon rents
Red Fescue
Price per 100 Ibs. 26.5000
Average variable costs per 100 Ibs. 6.7446
Oregon output (100 Ibs.) 67,240
Change in Oregon rents
Percentage change in Oregon rents
Chevir.gs Fescue
Price per 100 Ibs. 26.7000
Average variable costs per 100 Ibs. 6.7446
Oregon output (100 Ibs.) 68,000
Change In Oregon rents
Percentage change in Oregon rents
Merion-Kentucky Bluegrass
Price per 100 Ibs. 68.4000
Average variable costs per 100 Ibs. 7.5217
Oregon output (100 Ibs.) 27,520
Change in Oregon rents
Percentage change in Oregon rents
Bentgrass
Price per 100 Ibs. 37.70
Average variable costs per 100 Ibs. 9.8754
Oregon output (100 Ibs.) 70.480
Change in Oregon rents
Percentage change in Oregon rents
656,383 1,181,363 1,706,438 437,584 787,565 1,137,625
9.90 17.82 25.74 6.60 11.88 17.16
5.2588
60,566
7.97
7.6375
60,039
4.52
7.6375
60,712
4.47
8.3269
22,159
2.05
11.3591
5.7316
109,020
14.35
8.3517
108.064
8.14
8.3517
109,286
8.05
8.9710
39,884
3.89
12.5460
6.2044
5.0618
5.3770
9.0660
7.3399
7.8160
9.0660
7.3399
7.8160
9.6151
8.0585
57,610 14,773
5.33 1.37
8.4679
26,589
2.46
13.7330 10.8645 11.6558
5.6922
157,475 40,377 72,680 104,984
20.73 5.32 9.57 13.82
8.2922
156,092 40,012 72,043 104,062
11.75 3.01 5.42 7,83
8.2922
157,857 40,546 72,857 105,238
11.63 2.99 5.37 7.76
8.9173
38,407
3.56
12.4471
104,571 188,225 271,881 69,714 125,484 181,254
5.33 9.60 13.86 3.55 6.40 9.24
328,192 590,682 853.219
4.95 8.91 12.87
4.9633
7.1911
7.1911
7.9243
5.1997
7.5482
7.5482
8.2463
5.4361
30,283 54,510 78,737
3.99 7.18 10.37
7.9053
30,019 54,032 78,046
2.26 4.07 5.88
7.9053
30,359 54.643 78,928
2.24 4.03 5.82
8.5684
11,080 IS,942 28,805
1.03 1.85 2.67
10.6173 11.2107 11.8042
52,286 94,113 135,940
2.67 4.80 6.93
See Part 111, sections A and B for a discussion of definitions and computation nethods.
-------
Table 0.8 Estimated Consumer's Surplus, Oregon and Non-Oregon Rents for Various Grass Seeds, in Supply Situation III
Without and With Alternative Open Field Burning Control Policies at Varying Costs per Acre for the Mobile
Field Sanitizer"
Initial
Seed Type Situation
Rye grass
Price per 100 Ibs.
Average variable costs per 100 Ibs.
Oregon output (100 Ibs.)
Change in Oregon rents
Percentage change in Oregon rents
Tall Fescue
Price per 100 Ibs.
Average variable costs per 100 Ibs.
Oregon output (100 Ibs.)
Change in Oregon rents
Percentage change in Oregon rents
Red Fescue
Price per 100 Ibs.
Average variable costs par 100 Ibs.
Oregon output (100 Ibs,)
Change in Oregon rents
Percentage change in Oregon rents
Chevings Fescue
Price per 100 Ibs.
Average variable costs per 100 Ibs.
Oregon output (100 Ibs.)
Change in Oregon rents
Percentage change in Oregon rents
Merlon-Kentucky Bluegrass
Price per 100 Ibs.
Average variable costs per 100 Ibs.
Oregon output (100 Ibs.)
Change in Oregon rents
Percentage change in Oregon rents
Bentgrass
Price per 100 Ibs.
Average variable costs per 100 Ibs.
Oregon output (100 Ibe.)
Change in Oregon rents
Percentage change in Oregon rents
6.86
2.68
1,585,850
12.08
A. 6678
102,880
26.50
6.7446
67,240
26.70
6.7446
68,000
68,90
7.5217
27,520
37.70
9.8754
70,480
Open Field
$5
3.0939
1,409,832
1,319,285
19.90
5.2588
93,563
96,408
12.69
7.6375
62,975
140,487
10.58
7.6375
65,668
105,171
7.75
8.3269
17,173
40,036
3.71
11.3591
66,208
217,100
11.07
Burning Control Policies and Mobile Sanitizer Cost per Acre
Ban on Burning
$9
3.42494
1,270,417
2,264,844
34.17
5.73162
86,422
211,000
27.78
8.3517
59,568
247,297
18.62
8.3517
60,546
246,053
18.13
8.9710
16,838
71,173
6.59
12.5460
62,791
381,634
19.46
$13
3.74604
1,131,639
3,104,929
46.84
6.2044
79,282
293,794
36.68
9.0660
56,161
349,243
26.29
9.0660
57,235
347,686
25.62
9.6151
16,502
101,938
9.44
13.7330
59,374
538,059
27.44
Once in Three Tears
$5
2.9595
1,467,815
903,641
13.63
5.0618
96,537
82,104
10.81
7.3399
64,395
94,536
7.12
7.3399
65,236
93,989
6.93
8.0585
17,376
23,076
2.14
10.8645
67,632
146,141
7.45
$9
3.1766
1,375,084
1.563,869
23.59
5,3770
92,779
133,723
18.13
7.8160
62,123
167,649
12.62
7.8160
63,029
166,729
12.29
8.4679
17,090
47,473
4.39
11.6558
65,354
258,987
13.21
$13
3.3974
1,282,311
2,188,723
33.02
5.6922
87,017
203,775
26.83
8.2922
59,852
238,581
17.96
8.2922
60,822
237,369
17.49
8.9173
16,446
93,782
8.68
12.4471
63,075
368,252
18.78
Alternate Year Burning
$5
2.8870
1,496,765
682,206
10.29
4.9633
98,025
62,005
8.16
7.1911
65,105
71,244
5.36
7.1911
65,925
70,839
5.22
7.9243
7,383
20,317
1.88
10.6173
68,344
110,133
5.62
$9
3.0525
1,427,227
1,194,686
18.02
5.1997
94,455
109,742
14.45
7.5482
63,4.01
126,788
9.54
7.5482
64,270
126,079
9.29
8.2463
17,215
36,105
3.34
11.2107
66,635
195,964
9.99
$13
3.2180
1,357,731
1,683,997
25.40
5.4361
90,885
155,791
20.51
7.9053
61,698
181,098
13.63
7.9053
62,615
180,137
13.27
8.5684
17,048
51,7:5
4.79
11.8042
64,926
279,766
14.27
See Part III, sections A and B for a discussion of definitions and coaputation methods.
-------
APPENDIX H
Effects of Air Quality Variation on
Tourist Behavior*
* The work upon which this report is based was performed pursuant
to Contract CPA 70-117 with the Environmental Protection Agency.
-------
TABLE OF CONTENTS
I. Introduction H-l
II. Research Design H-4
III. Regression Results H-10
IV. Analysis of Questionnaire Data H-20
V. Summary and Conclusions H-32
VI. Postscript H-37
APPENDICES
1. Theoretical paper "The Effects of Changes in Air Quality
on Tourist-Related Industries in a Region"
2. 1971 "Tourist Revenue Study", Oregon Highway Division
3. Interview Instrument
4. Questionnaire Derived Data
-------
I. INTRODUCTION
Mark Twain once observed that everybody talks about the weather,
but nobody does anything about it. A similar generalization might apply
to air pollution. That is, everybody talks about air pollution. The
question is, do they do anything different as a result of it? Do people
alter their behavior as a result of air pollution? If behavior is altered,
is there any reason to believe that social welfare has been affected nega-
tively? If it has, what is the magnitude of social cost and/or economic
loss associated with this change in behavior? The research results offer-
ed below constitute an attempt to respond to these questions. The focus
of the research will be rather narrow in that it will not exhaust all the
possible ways that air pollution might affect behavior. However, if the
existence of some social costs associated with air pollution can be iden-
tified, and if a preliminary estimate of the magnitude of these costs can
be obtained, the research should suggest other avenues for continued
inquiry.
Prior to testing empirical hypotheses about the effects of air
pollution on individuals it is necessary to establish the theoretical
existence of such effects. That is, it is necessary to provide an inter-
nally consistent theoretical model that generates meaningful hypotheses
relating the behavior of individuals to air pollution. Appendix I of this
report provides a model that does produce meaningful hypotheses. In that
model, individuals are treated as utility maximizers in the traditional
manner in economic analysis. The problem is set up as a nonlinear pro-
gramming problem in which there is both an income and a time constraint.
The consumer combines market commodities (purchased with his budget) with
time (in intervals of varying length) and he produces consumption activi-
ties. These activities are the arguments in the utility function.
To model the resource allocation problem of an individual is rela-
tively easy. To relate air pollution to this utility maximization,
production-consumption problem is more difficult. As noted in Appendix I,
it is customary to treat air quality as constant and effectively indepen-
dent of any optimal consumption set. If one recognizes the fact that air
quality does vary and takes casual remarks about individual preferences
for better as opposed to worse air quality at face value, it is still not
obvious how to relate variation in air quality to the utility maximization
problem. One possible tack would be to include a variable called "pollution"
in the utility function and assume that pollution affects the utility func-
tion in certain negative ways. Then one might produce hypotheses about the
effects of air quality variation. Unfortunately, there is no market for
pollution. There are no explicit prices for pollution. Hence it is not
possible to observe individuals consuming more or less pollution as the
quality (price) of air varies. Any hypotheses generated by taking this
tack would be operationally unmeaningful.
H-l
-------
The approach utilized in the analysis of Appendix I is less direct
than trying to put pollution in the utility function, but it does produce
meaningful and potentially testable hypotheses. That analysis is expres-
sed in a comparative static framework in which the consumer seeks out an
optimal allocation of his budget and his time. At equilibrium, the marginal
utilities of all consumption activities are fixed; the values of the fixed
marginal utilities are a product of the consumer's preferences for the
respective activities and the amounts of the activities that are presently
being produced and consumed.
It is usual in the analysis of consumer behavior to subject any
equilibrium to the shock of changing either income or prices. The compara-
tive static framework, then, allows one to compare different equilibrium
positions in order to derive qualitative information on the effects of
that shock. In the analysis of Appendix I, the shock is provided by a
change in air quality that is assumed to affect the marginal utilities in
the system in a predictable manner. If air quality goes down, the marginal
utilities in the system either fall or remain unchanged. If air quality
goes up, the opposite set of sign changes in the vector of marginal utili-
ties takes place. In this framework, one can vary air quality and the
analysis will produce certain implications about changes in behavior. The
observable theoretical effects of air quality variation will be reflected
in variance in the amount of various consumption activities consumed and,
hence, in variance in the amounts of certain market commodities purchased.
Given certain a^ priori information about the relationship between air qual-
ity and various consumption activities, one can test the hypotheses genera-
ted in a meaningful and positive manner.
The hypotheses suggested by the model in Appendix I, of both the
operating and theoretical type, are:
1. Variation in air quality may have no direct effects on the
consumption of some activities. The result of this may be
that some activities appear to be independent of air quality.
2. There is a subset of consumption activities whose level of
consumption varies directly with the quality of the air.
3. By implication, since consumption activities are produced
by combining market commodities and time, and since market
commodities are purchased from a binding budget constraint,
and since there is a binding time constraint, variation in
the consumption level of one set of activities (i.e., those
activities whose consumption level varies directly with air
quality) will produce variation in another set of consumption
activities whose consumption level varies inversely with air
quality.
4. Since air quality is imposed as an exogenous variable unre-
lated to either the income or time constraints, any impact
on the set of optimal consumption activities must, by im-
plication, produce a negative welfare effect.
H-2
-------
5. Lastly, since the theory predicts that a deterioration
in air quality will lead to a switching in the set of
consumption activities produced and consumed by consumers,
by implication, there should be observable market impacts.
These hypotheses suggest that air quality does make a difference in
what people do and in their welfare. The hypotheses are conceptually test-
able. Put otherwise, the hypotheses suggest that variation in air quality
should produce social costs which are potentially measurable. The theory
dictates that these hypotheses be tested indirectly. •"•-As 'it turns Out,
there are substantial difficulties in doing this. These difficulties can
be attributed to. more than just the directness of the testing procedures.
The analysis of Appendix I draws upon the work of. Gary Becker who,
in his 1965 article on the "Theory of The Allocation-of Time", viewed the
consumer as literally producing consumption activities by combining market
commodities and time in different combinations so as to'create an activity
he then consumes. Each activity's production is limited by a production
function. JWhile this is theoretically plausible and intuJLtiyely appealing,
it unfortunately provides a source of difficulty for-purposes of empirical
analysis>. 'There are as many production functions a$ there are consumer
activities and there are no empirical estimates of them.
Aside from the lack of .a priori information about' the consumption
activity production functions, a second difficulty arises from another
lack of a priori information. The comparative static '/analysis of Appendix
I is amenable to the imposition of a shock on the equilibrium configura-
tion such as:would be provided by assuming that a change in air quality
affects the relevant marginal utilities in the configuration. This proce-
dure is mathematically and theoretically valid as well as Intuitively
plausible^ But there is a substantial difference between hypothesizing
a general and unspecific change in the vector of marginal utilities on the
initial and theoretical equilibrium configuration, and quite another thing
to be faced with the requirement of specifying exactly which marginal
utilities are affected by this change in air quality. If one knew that a
deterioration in air quality reduced the marginal utilities of a particu-
lar subset of consumption activities, he could predict that the consump-
tion of those activities and the consumption of the;associated market
commodities would decline ceteris paribus. The theory of Consumer behav-
ior would suggest, however, that if a particular subset of activities faced
diminished demand there should be another subset of activities that act
as substitutes for the first subset and which should be confronted by a
positive change in demand. There is, then, the dual problem of anticipat-
ing which activities would be affected negatively by a 'deterioration in
air quality and which activities would be affected positively as they per-
form the role of substitutes. Prior research is not helpful at this
juncture.
H-3
-------
II. RESEARCH DESIGN
The consumption activities we are concerned with can be specified
in terms of the amounts of market commodities (or public facilities) and
time required to produce a given kind of consumption activity. This is
suggested in general form in equation (2) of Appendix I. There, we have
assumed that the proportions are fixed or, alternatively, that the tech-
nologies are linear. Unfortunately, there are as many production func-
tions as there are consumption activities and there are no prior estimates
of the multitude of functional characteristics. A preliminary empirical
testing procedure may be suggested by the linearity assumption employed
with respect to the production functions.
Consider a production function such as the following:
Z. = fj (,cljf x2j, Tljf T2.) .
This hypothetical activity, Z., requires two market commodities, x-^ and X£,
and an interval of time from J two hours of the day. Drawing upon equa-
tion (3) of Appendix I, each unit of Z. produced and consumed would require
the expenditure of Pib.^ + Pobo• of ^ dollars and t^. + t£j of time.
Variation in the amount of Z^ produced and consumed would require direct
variation in the amounts of x^ and X£ as well as time. Even given that one
did not know the precise technical amounts of x-^ and X£ or time required
per unit of Zs, if he knew that these market commodities were involved in
the production and consumption of this activity he would predict that any
variation in the consumption of this activity would produce corresponding
and positively related variation in the amounts of those activities
purchased.
Suppose that a particular consumption activity required the use of,
among other things, a portion of the facilities at a public swimming pool?
Or suppose it required the use of a portion of the facilities of a golf
course, or a public camp ground, or a portion of the state capitol facili-
ties? Regardless of the name of the consumption activity, and regardless
of the other market commodities or public facilities required in the
production and consumption of the activity, and regardless of the strict
proportions of time required per unit of the activity, if one could pre-
dict variation in the amount of the activity consumed, he could predict
variation in the use of those particular facilities. Utilizing this pro-
position and the operating hypothesis about the relationship between air
quality and a particular set of consumption activities, it is possible to
indirectly test that hypothesis by relating the variation in air quality
to the variation in use or attendance at particular facilities that might
be predicted to be associated with the activities in question. Previous
research in this project has utilized data on levels of low visibility for
the cities of Salem and Eugene, Oregon, for the years 1969, 1970, and 1971.
Also, data are available on a limited number of public or private recrea-
tion-related facilities in these cities. In particular, for the City of
H-4
-------
Salem, daily attendance records over the relevant period are available for
the number of tourist visits to the State Capitol as well as the attendance
at the public swimming pools. Similarly, for the City of Eugene, daily
records are available for the indoor and outdoor swimming pools, one pri-
vate golf course and for one state park located in the immediate proximity
of Eugene. By hypothesis, the attendance at these facilities should be
affected by, among other things, the quality of the air. Assuming that
low visibility is a meaningful proxy for air quality, we can partially and
indirectly test the hypothesized relationships here using data on low
visibility. This procedure is incorporated as one part of this research
design.
®n
-------
They make substantial expenditures on commodities and facilities involved
in the consumption of out-of-door activities. If, in fact, air quality
makes a difference in the behavior of people and hence on the kind of
activities they consume, since there is such a large number of potentially
identificable people drawn to the state on an annual basis who utilize
activities theoretically related to air quality, it makes some sense to
employ an empirical technique that draws upon these facts. Appendix II
of this report outlines selected data on tourist travel in Oregon taken
from the 1971 Out-of-State Tourist Revenue Study prepared by the Oregon
State Division of Highways.
The lack of a_ priori information about the consumer activities pro-
duction functions has, as indicated above, complicated the empirical
problem. However, the flow of a large number of out-of-state tourists
traveling in the state, particularly during the summer months who utilize
facilities and purchase commodities that are involved in the type of
activities theoretically positively related to air quality, provides a
potentially large source of data on the effects of air quality variation.
One theoretical and empirical question for which there is no a. priori
information relates to whether tourists are "tracked" in their consump-
tion behavior such that variation in air quality will affect the selection
of consumption activities in a small region, or do these consumers have
sufficient flexibility to lead them to look some distance for available
alternative activities? The latter alternative would suggest the effects
of air quality variation are quite strong. It was decided to attempt to :
determine the answer to this question as well as provide an indirect test
of the effects of air quality variation by employing an interview tech-
nique on a sample of out-of-state tourists during the month of August
when the air pollution problem in the Willamette Valley is potentially,and
empirically the greatest.
In Appendix III the questionnaire employed in this empirical research
is provided. It was administered to 401 out-of-state tourists traveling in
Oregon in the interval August 19 through August 31, 1971. The questionnaire
is designed to:
1. Select for out-of-state persons traveling in Oregon for pleasure.
2. Identify the type of area from which they came.
3. Identify the routes they planned to travel prior to embarking
on their vacation trip.
4. Identify the type of activities they planned to enjoy on their
vacation prior to departing home and to obtain an idea of the
relative amounts of time they planned to allocate to each
activity.
5. Determine whether those pre-vacation plans have changed enroute.
6. Determine why any such changes occurred.
7. Focus on the possible impact of air pollution or air quality
variation on those travel plan changes.
H-6
-------
8. Estimate the dollar value or cost of those changes.
9- Determine the perceptions of the relative amounts of air
pollution in Oregon, vis-a-vis areas from which they came.
10. Obtain various demographic data on the population sampled.
The sample frame chosen for obtaining these interviews is the set
of out-of-state parties traveling in Oregon for pleasure who stopped at
a particular set of probability determined motels or safety rest areas
overnight. This sample frame was further divided equally between the
Willamette Valley on the one hand and a proportionally determined and
randomly selected set of motels and safety rest areas located on the
Oregon Coast and in the central Oregon region. This within-frame
dichotomy was suggested by the fact that it is the Willamette Valley
that has the largest actual and potential air pollution problem at any
time of the year and.particularly during the summer, while the coastal
and interior regions are essentially without an air pollution problem
at this time. Dividing the sample frame between the so-called "pollution"
and "nonpollution" areas should permit a maximal opportunity for observ-
ing whether air quality variation does, in fact, make a difference.
For purposes of determining exactly which motels and safety rest
areas would be involved in the actual sample design, the following pro-
cedure was employed. The so-called high pollution region which is af-
fected by field burning smoke as well as other sources of air borne
pollutants can be roughly identified as the area from Cottage Grove at
the southern end of the Willamette Valley to Portland. This region is
spanned in its entirety by Interstate Highway 5. Portland is located
on the northern border of the state and is the center of business activity.
The motels and hotels in Portland are expected to be utilized by tourists,
traveling businessmen, and state residents passing through Portland for a
variety of reasons. Since there are no data on the relative proportions
of these different types of potential users of the Portland area motels and
hotels and given its location in the state, it was decided to treat Port-
land as a special case and exclude it from the sample frame. Even with
this exclusion, there remains a large number of motels located in cities
and towns within the Willamette Valley area. In deference to pragmatism
and budgetary constraint, for the motel portion of the sample frame it
was decided to utilize the set of motels listed in the combined lists pro-
vided by the Oregon Motel Hotel Association Touring Guide and the AAA
Tour Book for cities or towns or motels listed along Interstate 5 in the
relevant region. The combined list of motels provided by these two tour
books is large but clearly excludes many motels even in the cities and
towns covered. However, these tour books are tourist oriented and the
listed motels provide substantial variety in room prices. For these rea-
sons, along with budgetary and other pragmatic considerations, the set was
delimited to the combined list offered by the tours bock. All safety rest
areas in this region of the Willamette Valley were included in the
potential frame.
H-7
-------
The 200 interviews obtained from the so-called "non-polluted"
portion of the state were obtained from the central and the coastal regions.
Determination of the set of motels and safety rest areas included in the
frame was based partly on pragmatism extending from a budget constraint
and partly on the basis of population and traffic pattern considerations.
For the Central Oregon area, Highway 97 provided the guideline for locating
motels and safety rest areas. Again, the combined list of motels provided
by the Oregon State Motor Hotel Association and the AAA tour books extend-
ing from Klamath Falls at the south to The Dalles at the north provided
the potential sample frame. Safety rest areas along this length of highway
were included in the potential frame. Given knowledge of the number of
units at the potential motels and the number of parking spaces at the po-
tential safety rest areas, the precise numbers of interviews to be taken
at each motel or safety rest area was determined on a random basis.
On the Oregon Coast, the highway that provided the guide for select-
ing motels and safety rest areas was Highway 101. The relevant range of
this highway was chosen to extend from Coos Bay at the south to Astoria at
the north. The choice of Coos Bay as the southern boundary was based pri-
marily on budgetary constraint considerations. The same procedure for
selecting motels and determining the number of interviews to be taken at
particular motels or safety rest areas was identical to the central and
Willamette regions.
The map provided in Figure 1 outlines the portions of 1-5, Highway
97, and Highway 101 covered in the sampling procedure. The portions of
these highways included in the sample design are marked in heavy black ink.
A copy of the questionnaire employed in this interview procedure is
provided in Appendix III. The questionnaire consists of 33 questions which
are designed to permit indirect testing of hypotheses relating to:
1. The impact of air quality variation on the pattern of
consumption activities consumed.
2. The economic costs of these impacts.
3. The relative magnitude of air pollution in Oregon vis-a-vis
other areas.
4. The type of person or family unit influenced by air quality
variation.
5. The degree to which vacation travelers are "tracked" in their
travels or whether they change travel plans as a result of
perception or parameter changes enroute.
The data generated and the analysis of those data are provided in section IV.
H-8
-------
MAP OF SAMPLE AREA-
Figure 1. Map outlining the portions of 1-5, Highway 97, and Highway 101
covered in the sampling procedure. The portions of these highways
included in the sample design are marked in heavier black.
H-9
-------
III. REGRESSION RESULTS
In the last section, a plan for relating attendance or use at certain
public or private facilities in the cities of Eugene and Salem was outlined.
The objective of that procedure was to attempt to perceive whether relation-
ships anticipated on a^ priori grounds between certain dependent variables
on the one hand and an index of air quality as well as a limited number of
other determining variables on the other hand is supported by empirical data.
In general terms, we are postulating a relationship between a dependent
variable, Y, and three independent variables, X^. In general form, this
functional relationship can be expressed Y = F (X-^, X£, Xo). For lack of
other information, the function, F, is in each case assumed to be linear.
In all, there are seven relationships hypothesized. For Eugene,
there are five relationships. For Salem, there are two. The five dependent
variables used for the Eugene area are (1) the number of camper units
staying overnight at Armitage State Park, located on the outskirts of
Eugene, (2) the number of "nine holes" played at Green Acres Golf Course,
Inc., (3) the number of people attending the two public indoor swimming
pools in Eugene, (4) the number of people attending the two outdoor public
swimming pools in Eugene, and (5) the number of people attending all public
swimming pools in Eugene; this is a total of the indoor and outdoor atten-
dance. The availability of daily attendance or use data on public or
private facilities is very limited. For State parks in any relevant region,
there are daily counts of the number of camper units. Armitage is the only
State park in the immediate area. In the Eugene area there are approximately
ten public or private golf courses. Of this set, only one claimed to have
data on the number of people using the facilities daily that could be regard-
ed as accurate. In most cases, signup sheets were not regarded as an accur-
ate indication of the number of people playing golf on a particular day.
Regarding attendance at the public swimming pools in Eugene, since there are
two indoor and two outdoor pools and since attendance at these types of
facilities might theoretically be different under variation of air quality,
it was decided to initially treat the indoor and outdoor pool attendance
separately. Data on these dependent variables is on a daily basis extend-
ing over a period from July 15 through either September 30, 1970, or in the
case of the swimming pools, through Labor Day, as the pools close after that
date.
For Salem, the two dependent variables are (1) the number of visitors
recorded at the Oregon State Capitol at the information counter, and (2) the
total attendance at the three public outdoor swimming pools. These data, too,
are daily and- extend over the'same period as do those for Eugene.
The independent variables in each case are definitionally the same.
Data were obtained for visibility as measured at the Eugene and Salem air-
ports. The low measurement was taken as the first independent variable.
Low visibility is, again, used as an index of air quality. This is an
appropriate choice on a priori grounds, given the fact that the primary
H-10
-------
source of air pollution during the period under consideration is the
smoke from burning grass seed fields in the Willamette Valley. The
disadvantage of this index of air quality relates to the manner in which
it is measured at the airports. The measurement technique involves a
visual and hourly evaluation of maximum visibility over 360° on the hori-
zon and measured against known distances between the airport on one hand
and certain pre-established markers on the other. Given the location of
the airport particularly at Eugene, the measurement technique may produce
inaccurate estimates of visibility in the respective cities. However,
this is the only meaningful index of air quality presently available for
these cities. It is an index that has been used in other empirical work
in this study-* Hence, low visibility as measured at the airports was
used as the index of air quality.
Attendance or use at the facilities reflected in the set of depend-
ent variables might, theoretically, be related to air quality. However,
since these are facilities that might be classed as recreation and, in
many cases, out-of-door type, attendance or use might also be theoretic-
ally affected by the temperature. For this reason, high temperature at
the respective cities was included as an independent variable. This
datum was selected from a set of daily observations taken at the respec-
tive airports.
Also, it seemed reasonable to dichotomize the week between the work
days of Monday through Friday on one hand and the weekend days of Saturday
and Sunday on the other. The purpose of imposing a dichotomy is explained
by the fact that most of the dependent variables can be regarded as being
affected by the behavior of local populations. Therefore, it seems rea-
sonable to recognize the influence of the institutionalized work week.
The independent variable reflecting this dichotomized version of "Day" is
a dummy variable which takes on the value of 0 for days in the Monday
through Friday range and the value of 1 for Saturday and Sunday. Given
the fact that the functional forms are assumed linear in the regression
analysis, this dummy variable has the effect of just shifting the regres-
sion.
It should be clear that this regression model subsumes everything
that usually comes under "taste" under the functional operator. The in-
dependent variables or functional arguments, i.e., low visibility, high
temperature, day - operate in the model as marginal and exogenous inde-
pendent factors that influence the behavior of people. Clearly these
three arguments are not the sole determinants of attendance or use at the
public and private facilities in operation. They are, again, treated as
marginal determinants of that behavior.
What are the qualitative results that one would expect from these
regressions? Table III.l below illustrates the qualitative set of rela-
tionships that are hypothesized for the seven regressions which were per-
formed. The supporting rationale is as follows:
* This work related to the Pollution Production Function.
H-ll
-------
f
H*
N3
TABLE III.l
Participation in or Use of a Selected Set of Dependent Variables Against
Visibility, Temperature and Day; Predicted Regression Signs
Regression Number
Dependent Variable*
(1)
A
(2)
G
(3)
PI
(4)
P0
(5)
PT
(6)
CV
(7)
P
Independent Variable
Low visibility ++-++-
High temperature + + + + + +
Day + +--- +
* Dependent variables are: A = daily attendance at Armitage State Park, G = number of nine holes
of golf played at Green Acres Golf Course, Pj = attendance at Eugene indoor swimming pools,
PQ = attendance at Eugene swimming pools, P-p = total indoor and outdoor swimming pool attendance
at Eugene, CV - number of daily registered visitors at State Capitol, P = total daily attendance
at Salem public outdoor swimming pools.
-------
1. The Armitage State Park variable, A, measures the number
of camper units staying overnight at that facility measured on a daily
basis. On ja priori grounds which accept the proposition that camping
is related to the environmental quality, we would expect this variable
to be positively related to low visibility, positively related to high
temperature, and positively related to the day variable. Given the loca-
tion of Armitage, being on the McKenzie River and very close to Eugene,
the attaraction of this facility may be different than that of another
state park located in more remote areas. That is, the Armitage facility
may be more of an overnight accommodation for visitors to the area than
a purely camping and outdoor recreation oriented facility. Therefore,
the coefficient for low visibility might be expected to be smaller than
the coefficients for the other two variables.
2. The golf variable, G, is viewed as functionally similar to
the Armitage State Park or the outdoor swimming pool variables. That
is, it reflects an outdoor type activity that, on a priori grounds, is
influenced by the quality of the environment. Therefore, we would pre-
dict a positive relationship between the golf variable and low visibility,
a positive relationship with high temperature, and a positive relationship
with the day variable.
3. The indoor swimming pool variable, P^ is viewed as offering an
interesting cross-check on the role of air quality on recreation type
activities. Eugene has two indoor pools along with two outdoor pools.
Theoretically we should observe a difference in the impact of air quality
on the attendance of these two classes of swimming pools.
4. The outdoor swimming pool attendance variable, PQ, should be
more strongly related to air quality. Specifically, we would predict a
strong positive relationship between attendance at outdoor swimming pools
and our low visibility variable. As in previous predictions, we would
predict a strong positive relationship between the outdoor swimming pool
variable and high temperature while the day variable should be inversely
related for the same arguments.
5. The total swimming pool attendance, PT, which is the sum of
attendance at both indoor and outdoor swimming pools was regressed against
the same variables. It provides a useful cross-check on the broad hypothe-
sis about the effect of air quality variation. In this case, total outdoor
swimming pool attendance, P.J, would be predicted to have a positive
coefficient for the high temperature variable, and a negative coefficient
for the day variable.
6. The sixth regression relates the number of visitors to the
State Capitol in Salem, CV, to the same set of exogenous variables measured
in Salem. On a_ priori grounds, visiting the State Capitol might be viewed
as an indoor activity and, theoretically, might be treated as a substitute
for alternative outdoor activities. This proposition is not advanced with
confidence. It is possible that the activity represented by visiting the
State Capitol might be independent of air quality for variation within some
reasonable range. Nonetheless, we would predict that the coefficient for
H-13
-------
low visibility in this regression should be negative, the temperature
variable coefficient should be positive. The day variable offers some-
thing of a problem. It is not clear how this variable should be related
to the number of Capitol visitors. Some visitors to the Capitol come in
tours that are prearranged. The number of tours is not a function of
day of the week. One might predict that the weekends would attract more
families out for one-day family outings. On this basis alone, one would
predict a positive relationship between the number of Capitol visitors
and the day variable.
7. The City of Salem has three public swimming pools. They are
all outdoor. Following the same arguments applied for Eugene, we pre-
dict a strong positive relationship between swimming pool attendance and
low visibility, a positive relationship with high temperature, and an
inverse relationship with day of the week.
Table III.2 offers the empirical results for the seven regressions
outlined above. Regression coefficients significant at the .01 or .05
levels are indicated. All coefficients significant at levels higher than
.05 are assumed to be zero. These empirical results can be summararized
as follows:
1. Operating on a .05 level of significance, the data revealed
that low visibility and attendance at Armitage State Park are apparently
statistically independent. The regression coefficient is presumed to be
zero. The high temperature variable is positive as predicted and statis-
tically significant at the .01 level. The data suggest that attendance
at the Armitage Park and day of the week are statistically independent.
That is, the regression coefficient for day is assumed to be zero. Over-
all, the regression explained only 23% of the variation in attendance at
Armitage. However, the F-statistic is significant at the .01 level
suggesting that there is regression. However, the performance in this
regression would suggest that there is a set of other variables that
determine the overnight attendance at this particular state park.
2. The regression between the number of nine holes played at
Green Acres Golf Course in Eugene and our set of independent variables
also revealed that, on a statistical basis, low visibility - our index
of air quality - is independent of this use of golf facilities. The
regression coefficient here is assumed to be zero. However, the regres-
sion coefficients for high temperature and day are both positive as
predicted and significant at the .01 level. Overall the regression explains
50% of total variation in G. The F statistic is significant at the .01
level.
3. In the case of indoor swimming pool attendance in Eugene, low
visibility also reveals statistical independence. However, again, the
high temperature and day variables reveal the predicted signs and are
significant at the .01 level. Fifty-seven percent of total variation in
Pj is accounted for by the regression. The F statistic is again signifi-
cant at the .01 level.
H-14
-------
TABLE III. 2
Participation in or use of a Selected Set of Dependent Variables Against Visibility, Temperature and Day.
Regression Results: Salem and Eugene, Oregon, 1970
Regression Number (i) (2) (3) (4) (5) (6) (7)
Dependent Variable:
A G PZ PQ PT Cap. Vis. P
Regression Coefficients, Etc.
Constant
* Low Visibility
M
High Temperature
Day
R2
F-statistic
Sample size
17.8 -.73 -1197.30 -2973.64 -4228.56 -258.8 -2164.
-.3 .67 5.21 11.913 18.11b -1.04 10.07a
.89a 1.07a 20.83a 43.71a 64.88a 7.62a 34.52a
2.05 6.31a -182. Ola -199. 65a -421. 19a 64.59b -206. 50a
.23 .5 .57 .74 .66 .20 .74
7.50a 25.40a 22.39a 46.21a 31.65a 6,14a 48.40a
78 78 54 54 54 78 55
Notes:
a Significant at the .01 level
b Significant at the .05 level
-------
4. For attendance at outdoor swimming pools in Eugene, the
low visibility variable reveals a much stronger influence. In this
case, the regression coefficient is positive and significant at the
.01 level. A priori considerations would have suggested this should
have been the case. As in the previous regression, the high tempera-
ture and day variables revealed the predicted signs and are signifi-
cant at the .01 level. Overall, 73% of total variation in PQ is ac-
counted for by the regression. The F statistic for this regression
is also significant at the .01 level.
5. The regression for total pool attendance in Eugene produces
comparable results to the regression on outdoor pool attendance. That
is, the low visibility variable is positively related to total pool
attendance and significant at the .05 level. The coefficients for high
temperature and day both bear the predicted signs and are significant
at the .01 level. Sixty-five percent of the total variation in Pp has
been accounted for by regression. The F-statistic is significant at
the .01 level.
6. The regression relating number of visitors at the State
Capitol to the set of independent arguments is the weakest of all seven
regressions in terms of the amount of variation it accounts for. Low
visibility is revealed as statistically independent to the number of
visits to the capitol. The coefficient for this variable is assumed to
be zero. Capitol visitations do appear to be related to temperature.
The regression coefficient for high temperature bears the predicted
positive sign and is significant at the .01 level. The coefficient for
day is significant at the .05 level, is positive, and quite large. The
interpretation of this is somewhat obscure. Only 20% of the total var-
iation in capitol visitation is accounted for by regression. However,
the F-statistic is significant at the .01 level.
7. In the case of pool attendance for Salem, the regression
performance compares with the results for Eugene. That is, the coef-
ficient for low visibility is positive and significant at the .01 level.
This sign relationship is as predicted. For high temperature and day,
the coefficients bear the predicted positive and negative signs respec-
tively and they are significant at the .01 level. These variables and
this regression account for 74% of the total variation in P. The
F-statistic is significant at the -01 level.
What are the interpretations that one would apply to these seven
regressions, particularly with respect to the impact of air quality
variation? Clearly, the linear model has substantial power in account-
ing for variation in the respective dependent variables. This is evi-
denced by the consistent failure to refute hypotheses about lack of re-
gression. All F-statistics were significant at the .01 level. But there
is substantial variance in the amount of overall variation in the depen-
dent variables that the regressions account for. The range in the amount
of variance accounted for by regression extends for a high of 74% to a
low of 20%. The high temperature and day variables also reveal themselves
as powerful in explaining variation in the respective dependent variables.
H-16
-------
What do these regressions suggest about the role of air quality
variation in the determination of what people do? From the examination
of the limited set of activities associated with, the facilities examined
in regression, the answer is unclear but there are suggestions of
relationship.
Consider, first, the regressions for which low visibility proved
statistically insignificant. These were the regressions for overnight
attendance at Armitage State Park (A), the number of nine holes played
at Green Acres Golf Course (G), the attendance at indoor swimming pools
in Eugene (P,-), and the number of visitors to the state capital in Salem.
In each of these cases, low visibility revealed itself as being sta-
tistically independent (i.e., having a zero relationship) of the respec-
tive independent variables. These results are probably plausible in
terms of the underlying theory for this analysis.
The most difficult of these cases is probably the one of Armitage
State Park. We attempted to treat overnight attendance at Armitage State
Park as an outdoor "camping" activity identical to data that might apply
to state parks located in more remote and specific camping areas. How-
ever, as indicated in the description of the independent variables,
Armitage State Park is located on the McKenzie River and might be viewed
as effectively in Eugene. It is also easily accessible from Interstate
Highway 1-5. These characteristics make the Armitage State Park facility
attractive to tourists or to University of Oregon summer students who are
looking for overnight camping facilities. The City of Eugene does
attract tourist-visitors due to the number of lumber and plywood mills
in the area that offer free tours and because of the location of the
University of Oregon. Due to these circumstances, the Armitage State
Park facility might theoretically be used like any other overnight camp-
ing facility and for reasons that are independent of the quality of the
air, under conditions of normal air quality variation. Unfortunately,
there are no other state, federal or Bureau of Land Management camping
and park facilities in the area for which we have air quality data.
The regression for the golf variable, G, reveals a zero relation-
ship between low visibility and this dependent variable. This is somewhat
disconcerting. Playing the game of golf clearly involves an outdoor and
recreation type activity. The game is played in the open atmosphere.
Perception of the air quality component of the atmosphere should reason-
ably be influenced by visibility. Other factors involved in that percep-
tion would include atmospheric temperature, presence of rain, etc. In
the regression for this golf variable, high temperature did reveal a very
strong positive relationship. Why does the playing of golf appear to be
unaffected by air quality as measured by low visibility? An explanatory
hypothesis is not obvious and this study is not required to provide one.
However, the anser to the question may be tied up in the character of the
game itself and the type of people that play it.
The zero relationship between low visibility and attendance at
Eugene indoor swimming pools is also difficult to explain. If deterioration
H-17
-------
in air quality does influence the set of activities that individuals
enjoy, and if individuals have a choice of the type of facilities they
might use in the face of air quality variation (such as indoor as
opposed to outdoor swimming facilities), then one might expect (on
theoretical grounds) that deterioration in air quality would lead them
to substitute indoor for outdoor swimming facilities in the face of a
reduction in air quality, other things equal. That is, we would predict
an inverse relationship between the use of indoor swimming facilities
and low visibility. However, the empirical results reported here sug-
gest the relationship is neutral.
This does not refute the hypothesis being examined here. The
most that the hypothesis about relationship suggested was a weak inequal-
ity (less than or equal to). If one were to find support for the strict
inequality (less than), there would be stronger support for the hypothesis
about substituting one activity for another under variation in air quality.
When one is forced to accept the weak inequality, what remains is the un-
certainty over the manner in which activity substitution takes place if,
in fact, it does.
The case of the empirically suggested zero relationship between the
number of visitors to the State Capitol and low visibility is less
troublesome. The a_ priori considerations that lead to using the number
of visitors to the State Capitol as a dependent variable were not rigor-
ously derived. The argument was that the activity involved in visiting
the State Capitol is a recreation-type activity (although, admittedly,
it might also be viewed as educational) as well as indoor; therefore, one
might predict an inverse relationship between air quality and State Capitol
attendance. However, without other information, this relationship would
be predicted as a weak inequality (i.e., less than or equal to). As it
turned out, with the data employed in this analysis, the sign on the coef-
ficient relating air quality to attendance at the State Capitol was nega-
tive. However, the coefficient was not significant at any level and hence
the conclusion of no relationship. It may well be that the activity in-
volved in visiting the State Capitol is more educational than recreational
and, in that case, the role of air quality in affecting the demand for that
type of facility may, in fact, be neutral. The regression for this depen-
dent variable was conducted here primarily because of the availability of
data and on the casual assessment that the activity involved in visiting
the State Capitol may be more educational than recreational and, in that
case, the role of air quality in affecting the demand for that type of
facility may, in fact, be neutral.
The relationship between low visibility and attendance at the outdoor
swimming pools in both Eugene and Salem were, again, positive and highly
significant. These results might be regarded as "clean" in the sense that
they bear the predicted sign and were highly significant. These results
provide substantial presumptive and empirical support for the overall
hypothesis that air quality makes a difference. The most important aspect
of these results relates to the fact that the results are strong where we
would predict they should be strong. A priori considerations suggested
that swimming is definitely a recreation activity. Where the activity is
undertaken in the out-of-doors and if air quality makes a difference, the
results should be strong.
H-18
-------
On an overall basis, there is enough supportive empirical evi-
dence to suggest that air quality does make a difference and that
consumers do engage in some kind of activity substitution under the
variation of air quality. What is not clear is the character of the
substitution and the welfare and economic costs associated with it.
The answer to the latter question depends on both the magnitude of the
marginal rates of substitution between activities affected positively
by air quality and those that act as substitutes as well as on the tech-
nical or market commodity characteristics of the activities. The answer
to the welfare question depends on the character of the activity set in
the real world. That is, for analytical purposes we have assumed that
consumption activities are divisible on a continuous scale and that the
utility function is also continuous. If these conditions hold in prac-
tice, variation in any independent variable might lead to continuous
reductions (increases) in one activity with corresponding changes in
the associated market commodities and use of public facilities, as well
as continuous increases (decreases) in the consumption of substitute
activities with corresponding changes in associated market commodities
and use of public facilities. In actual practice, the consumption
activities we have reference to might be better measured on an integral
scale. In this case, consumption activities would be produced and
consumed in integral amounts. If this were the case, the consumer's
problem might be better described as an integral programming problem
in which case, variation could take place in the effective independent
variables over some range without affecting the level of consumption of
any optimal consumption activity set. Put otherwise, under these cir-
cumstances air quality might vary over a range of unknown width without
affecting the precise consumption set or welfare.
H-19
-------
IV. ANALYSIS OF QUESTIONNAIRE DERIVED DATA
Characteristics of Travelers
As indicated in Section 2 above, this research design involved inter-
viewing a set of out-of-state parties traveling in Oregon for pleasure
during the month of August, 1971. The questionnaire instrument provided in
Appendix III was administered to 401 out-of-state parties. The questions
and hypotheses which we were attempting to obtain insight to are outlined in
Section 2 above. Before returning to those questions and hypotheses and
the discussion of relevant data, it may be useful to outline some of the
characteristics of the parties interviewed.
Fifty-four percent of all parties interviewed came from California;
20 percent came from the state of Washington. Of the remaining 26 percent
of interviewed parties, 9 percent came from other western states while the
residual was distributed over the rest of the nation and Canada.
In addition to home state, interviewed parties were asked for city of
residence. Utilizing the 1970 census preliminary reports to establish city
populations, the cities listed by respondents were coded into seven popula-
tion categories with "less than 2500" at the lower end and "more than 500,000"
at the upper end. Twenty-two percent of all respondents indicated residence
in a city of more than 500,000 and twenty-three percent of the respondents
listed residences in cities in the "2500 to 24,999" class. The remaining
respondents were fairly evenly distributed in the other five population bands.
Several of the questions asked related to the activity plans of the
vacationer prior to leaving home. The questions asked for plans relating
to places they intended to visit, the number of days they intended to stay
in Oregon, and the set of highways they intended to use in their travels.
In terms of the places vacationers planned to visit in Oregon, eleven ident-
ifiable sets of places were established for coding purposes. Forty-one
percent of all respondents indicated plans to visit the Willamette Valley
cities. Other areas which respondents revealed as having high attractive-
ness included the coastal recreational areas for which 32 percent of respond-
ents indicated plans to visit, the coastal cities which 25 percent of respond-
ents indicated plans to visit, and the city of Portland, which 23 percent of
respondents indicated plans to visit. Interestingly, only 8 percent of re-
spondents indicated plans to visit Willamette Valley recreational areas.
These percentages summed to more than 100 percent due to the opportunity
for listing more than one area in the set of planned areas. One suggestion
obtained from these data is that, if the population interviewed in this
study reflects the characteristics of the population of all parties travel-
ing in Oregon for pleasure (even in the summer), the great bulk of these
travelers do not depart significantly from the main traffic arteries.
H-20
-------
The same observation applies when one considers the set of highways
that these travelers planned to use to their destination and in leaving
the state. Sixty-six percent of all parties planned to use Highway 1-5
to their destination. Forty-three percent indicated plans to use Highway
101 to their destination. The percentages of responses for other highways
are considerably smaller. These data are provided in Appendix IV.
Thinking in terms of pre-vacation plans, respondents were asked the
degree to which they intended to engage in nine specific outdoor activities
readily available to Oregon tourists. This set of activities included
camping, boating or water skiing, golfing, ocean beach activities, hiking
or walking, swimming, pleasure driving and sightseeing, salmon fishing, and
trout fishing. To each of these activities respondents were asked whether
they planned to engage in the activity a lot, quite a bit, a little, none,
or "don't know".
An analysis of the data generated by these questions revealed some
interesting results. For the activities camping, boating or water skiing,
golfing, salmon fishing, and trout fishing, more than 75 percent of the
respondents in each case said that they did not plan to engage in these
activities. This is superficially interesting in the sense that these ac-
tivities reflect several of the activities that are presumed to be prime
attractions for summer tourists. The response configuration to this set
of questions may in part be reflected by the sample frame. However, there
are no a^ priori reasons to presume that the sample frame was biased against
these activities.
In response to the question regarding the planned use of time in the
activity called "pleasure driving and sightseeing", 51 percent of the
respondents indicated "a lot" while 23 percent indicated they planned to
engage in this activity "quite a bit". The activities ocean beach activi-
ties, hiking and walking, and swimming received the responses "a little"
24, 50, and 25 percent respectively.
These questions on the planned use of time in specific outdoor activi-
ties along with the responses relating to planned uses of highways and the
planned areas of visitation seem to suggest that the out-of-state tourist
who is reflected by the sample frame used in this study is very markedly
"tracked" in terms of the path that he travels and the activities that he
engages in. This observation is supported by additional data to be cited
below. The full data configuration describing responses to the planned
activity questions are provided in Appendix IV.
Other interesting characteristics of the set of travelers sampled
relate to years of education completed, age of respondent, and sex. Regard-
ing education, approximately 37 percent of the respondents had completed
college, 25 percent had partial completion of college, 35 percent indicated
a high school education, while approximately 3 percent listed grade school
or no school as highest educational achievement.
H-21
-------
Regarding the age characteristics of respondents, the age bands that
respondents were offered to associate with were five in total. These in-
cluded the 20-29, 30-39, 40-49, 50-59, and 60 and above brackets. The age
distribution of respondents was quite symmetrical. Fourteen percent of the
respondents fell in the 20-29 bracket, 21 percent fell in the 30-39, 25
percent fell in the 40-49 bracket, 22 percent fell in the 50-59 bracket,
and 17 percent fell in the 60 and above.
The sex distribution of the respondents is also interesting. Of the
401 total respondents, 61 percent were male and 39 were female. Interviewers
were under instruction to attempt to balance the sex distribution of respon-
dents if possible. Since only 5 percent of all respondents indicated their
party size as one (1), there is reason to presume that there may have been
a choice available to interviewers in terms of the sex of respondent. The
actual sex distribution of respondents seems to suggest that either the
interviewers interjected some bias or the parties themselves revealed some
process by which sex made a difference in terms of which party member should
respond to the request of an interviewer. The precise reason for the uneven
sex distribution of respondents is not important for purposes of this study.
Characteristics of Parties Changing Plans
Two of the questions of this study set out to provide insight to ques-
tions related to whether vacation travelers actually did change their plans
en route due to variation in the parameters of their planning frame and
whether, for those who did change plans, air pollution could account for
any of those changes. Each of the parties interviewed was asked the ques-
tion "have your vacation plans for Oregon changed since you arrived in the
state?" Of the 401 respondents, 59 said that they have changed plans in
the state while 340 said there have been no changes and two responded they
did not know or did not answer. The 59 parties who indicated a change in
plans while traveling in Oregon were offered a list of factors that may have
influenced the plan changes they experienced. The reasons included rain,
heat, crowded facilities, air pollution, lack of money, automobile break-
down or accident, sickness in the family, and "other". Those that indicated
"other" were asked to specify what this meant. There were 30 affirmative
responses in the "other" category. When the 30 questionnaires with this
response were examined, it turned out that, where explanations were offered,
most comments spoke to the crowded facilities and lack of available rooms.
The second highest affirmative response related to rain; there were 15
responses in this category. Crowded facilities showed 11 affirmative re-
sponses. Of the 59 parties who indicated a change of plans, only (^ indicated
that air pollution had anything to do with their behavior.
H-22
-------
The parties that did change plans were also asked if, for reasons other
than air pollution, their change of plans had affected their planned expend-
itures. That is, they were asked whether planned spending had increased,
decreased, or remained the same; a fourth response of "don't know or no
answer" was also provided. Of the 59 respondents who changed plans, 22
indicated their spending had increased, 13 indicated their spending had
decreased while the residual indicated no difference or they did not know.
The distribution of the estimates of increased expenditures required by
vacation plans is fairly even with a median of about $100 increase. The
distribution of estimated reductions in spending resulting from vacation
plan changes is more skewed with a median and mode in the $51-100 range.
These data seemed to add further support to the observation that out-
of-state parties traveling in Oregon for pleasure have fairly fixed plans
about the path of their travels and the amount they will spend. Where in-
formation or events require parties to alter their plans, the factors re-
quiring changes may lead the parties to increase their spending over their
expectations slightly but, given that some parties have their planned ex-
penditures reduced by factors affecting their vacation plans, the net impact
of vacation plan changes on spending in the state may be approximately zero.
The data from this sample should not be advanced as strong evidence of other
conclusions.
THE IMPACT OF AIR POLLUTION
The primary purpose for administering the interview instrument was to
obtain insight into the question of whether air pollution actually led
vacationers or consumers of outdoor and recreation-type activities to alter
their plans, spending, and behavior. The questionnaire design also permit-
ted the generation of perceptions of the air quality situation in Oregon
and to obtain evaluations of comparative air pollution between areas visited
in Oregon and home areas.
For purposes of this analysis, the answer to the impact question can
be approached from two directions. First, of the 59 parties that indicated
their vacation plans had changed, only 6 indicated that air pollution had
any bearing on their behavior. These 6 responses constituted only 1.5 per-
cent of the total responses and cannot be viewed as statistically signifi-
cant in relation to any of the hypotheses of interest to this study.
While the number of those who changed plans and revealed air pollution
as a factor in this change was small, the observations of respondents in
this category are of interest. The six respondents that indicated air
pollution did make a difference in their vacation plans were asked probing
questions about what kind of air pollution affected their behavior and in
what ways. The responses along with some descriptive information about the
respondents are provided below:
H-23
-------
a. A male truck driver from California, in the age bracket 40-49
said: The field burning and smoke from Salem and Eugene
affected our plans. We were thinking of staying over in
Eugene, but we didn't because of the smoke. We just wanted
to get away from it and go somewhere where there isn't any
smoke. For one thing, you can't see the scenery, and the
smoke bothers my sinus. There is a lack of good, fresh air.
b. A television repair shop owner from California, in the age
bracket 40-49 said: Portland smog is bad because of indus-
trial and auto pollution. We skipped a day in Portland and
went to Vancouver instead.
c. The wife of a California salesman, in the age bracket 50-59
said: We were astonished at the sawdust burners used by the
mills, like near Roseburg. We were going to stay in Eugene,
but the sawdust burners in the area made us decide to come on
up north. It was a haze we didn't expect.
d. A male lawyer from California, in the age bracket 20-29 said:
Basically it is the countryside pollution or haze one sees
from the highway that bothered me. I feel it is worse than in
California. We shortened our stay due to the pollution. We
had some idea of moving to Oregon, but we're not too sold on
Oregon.
e. The wife of an Agricultural Inspector from California, in the
age bracket 50-59 said: (Didn't specify kind of air pollution)
We had planned to stay overnight in Portland; but because of
air pollution, we decided to go on east to The Dalles.
f. A male Steamfitter from California, in the age bracket 50-59
said: The smog in Southern Oregon is bad, all the way up the
Willamette Valley as far as Salem. Didn't travel any farther
north, but I'm sure Portland would have been worse. I don't
like smog. If there is going to be smog, a person might as
well go elsewhere.
Several observations about these comments are required. First, all
respondents were from the state of California. All the comments seemed
to indicate a set of preconceptions about the relative air pollution in
Oregon vis-a-vis their home state of California. For this sub-population,
experience seemed to indicate that the relative difference in air pollution
did not meet their expectations.
Another observation about the set relates to the qualitative character
of the pollution observed. They speak of field burning, sawdust burner
smoke, and "countryside pollution". One party observed the bad "Portland
smog". The interesting thing about these observations is, not that they
differ from observations generated by Oregon citizens, but rather they
H-24
-------
are so much like comments coming from within the state by residents and
that there were so few respondents who indicated air pollution affected
their behavior.
The question about whether air pollution in the state was statistic-
ally related to the change of vacation plans by travelers was tested in a
multivariate analysis of variance technique. This analysis will be discus-
sed below. For present purposes, we will continue with the more descrip-
tive material relating to perceptions of the relative air pollution in
Oregon vis-a-vis home area.
The questionnaire contains five questions having to do with percep-
tions of relative air pollution in Oregon and about the qualitative charac-
ter of the Oregon problem. Analysis of the data from these questions sug-
gest something about why so few people revealed air pollution as having
affected their plans and just how serious an air pollution problem they
think Oregon has.
Of the 401 total respondents, only 76 (19 percent) responded that the
Oregon problem was "quite" or "very" serious. One hundred eighty-one
respondents indicated the problem was "not too" serious and 136 indicated
there was "no problem". Eight respondents indicated they did not know.
In a similar question asking about how serious the air pollution prob-
lem is in their home area, 223 respondents (57 percent) indicated that the
problem was either quite or very serious, while 12 said it was "not too"
serious and 64 indicated there was "no problem". Responses to these two
questions suggest that, in terms of perceptions, it is relative rather than
absolute air pollution that makes a difference.
This is suggested further by another question that asked about percep-
tions of relative air pollution in Oregon vis-a-vis their home area. Two
hundred forty-five respondents (61 percent of the total) indicated that
Oregon had less air pollution than their home area. Seventy-one respond-
ents indicated Oregon had more pollution while 63 indicated it was about
the same. An interesting sub-set of the responses to this question were
coded "qualified less in Oregon". Respondents who observed either more or
less air pollution in Oregon vis-a-vis their home area were asked to explain
observations. The verbal comments are provided in Appendix IV, however,
it is interesting to note that, for those who provided a qualified "less
in Oregon" spoke quite specifically about the extensive field burning smoke
problem in the Willamette Valley.
Review of the verbal comments provided in Appendix IV and relating to
perceptions of relative air pollution reveals an almost unanimous agreement
on the distastefulness of field burning smoke. Comments included the obser-
vation that "the fire on the farms is quite bad" or that they had observed
"a heavy and obnoxious smoke". One respondent indicated that, in terms of
a statewide comparison, Oregon had a much smaller air pollution problem than
the home state of California but "where there is field burning, the problem
is much greater".
H-25
-------
MULTIVARIATE A.N.O.V.
In all, the interview instrument had 47 questions for which the re-
sponses could be listed as "closed ended". Most of the data from this set
of questions were in code form. These codes related to such matters as size
of city, state of origin, planned expenditure categories, qualitative per-
ception of air pollution, level of education, age, etc. In only three cases
were absolute numbers requested as estimates or statements of particular
values. These pertain to the estimated amount of planned expenditures to be
made in the state of Oregon, the number of days they planned to spend in the
state, and the size of the total party.
The data from this subset of questions in the interview instrument might
have been statistically processed in a number of ways. After observing the
very weak case to be made for air pollution in affecting the travel plans and
behavior of tourists traveling in Oregon for pleasure as evidenced by the
fact that only 6 parties indicated air pollution made a difference, it was
decided to slice and analyze the data along two lines that may provide insights
into the character of the population being examined and to the reasons for
their behavior. The points of data partition were:
1. At question 17 which asked whether vacation plans for Oregon had
changed since arriving in the state; the partition at this point
was between those who had changed and those who had not or did
not know.
2. At question 25, which asked the perception of respondents toward
the seriousness of Oregon air pollution; the data were partitioned
between those who responded "quite serious" and "very serious" on
the one hand and those who viewed it as "not too serious", "not
serious at all" or did not know on the other.
These data partitions made sense on several grounds. The question deal-
ing with the change in vacation plans was included to assist in finding out
whether the population in question is sensitive to unanticipated information
received in the course of travel and sensitive in the sense that parties
changed their behavior. The question on perception of Oregon air pollution
provided a meaningful point of partition because the focus of this research
is on the impact of air pollution on behavior and, since air pollution re-
vealed itself as having no apparent impact on the in-trip plans and behavior
of travelers, it seemed to make sense to focus on the question of whether
there are meaningful differences between the set of travelers that view the
pollution problem in Oregon as either quite or very serious as opposed to
those who saw the problem in Oregon as of more or less minor concern. The
latter partition would permit a continued focus on the impact of air pollu-
tion, although not in quite the same way as was anticipated at the outset.
H-26
-------
With the partitioning of the data set along the lines outlined here in
mind, it was decided to subject the data for a selected set of variables to
analysis through the multivariate analysis of variance (MANOV) technique.!
As used in this research, MANOV technique was used to examine hypotheses
about equality in means of variables generated by two groups. For purposes
of this research, the hypotheses in question at this juncture were:
1. The subset of respondents who changed vacation plans while in
Oregon and the subset of respondents that did not change vaca-
tion plans were drawn from the same population.
2. The subset of respondents who indicated that air pollution in
Oregon is either a "quite" or a "very" serious problem and a
subset that perceived no problem or not too serious a problem
were drawn from the same population.
The computer program used in performing this analysis was constrained to
a set of 20 independent variables. To accomodate this constraint, a set of
variables that included state of origin, a set of 10 planned outdoor activi-
ties, planned expenditures in Oregon, the question regarding spending changes,
perception of pollution at home, relative air pollution in Oregon, party size,
highest education achieved, occupation of major bread winner, age, and sex
was utilized. This same set of 20 variables was used to attempt to predict
those respondents who did change their vacation plans in Oregon versus those
who did not change plans en route on the one hand and to predict those that
viewed air pollution in Oregon as relatively bad versus those that viewed it
as not so bad on the other. A separate "interaction" analysis was performed
which attempted to discover a systematic relationship between those respond-
ents who saw pollution as "bad" in Oregon and changed plans while traveling
in the state on the one hand and those who did not view pollution as "bad"
and did not change plans while traveling in the state on the other. The
MANOV technique provides tests for hypotheses relying on assumptions about
linear statistical relationships. The major test statistic is the F statis-
tic which is used for examining hypotheses about linear fit. The computer
program used in this research also generates a computed discriminant function
the use of which provides a way of comparing vectors of mean values from two
populations and for assessing the relative importance of the contribution
of each in dependent variable involved in developing the computed value of a
particular test statistic.
^ For one discussion of the use of multivariate analysis of variance
in testing hypotheses about means, see Donald F. Morrison, Multivariate
Statistical Methods, McGraw-Hill Book Company, New York, 1967, Chapter 4.
H-27
-------
Change Vs. No Change
In the attempt to predict respondents who changed plans as opposed to
those who did not -from the set of 20 variables outlined above, the analysis
was not particularly clear. The computed F value with 20 and 380 d.f. was
1.686. The critical value of F at the 5% level was 1.60. The fact that the
computed F only marginally exceeded the critical value of F left serious
questions as to whether the results were really statistically significant.
In the first computer run, the discriminant function revealed a subset
of 10 independent variables that appeared to be most influential in predict-
ing the changers vs. the non-changers. The variables along with the dis-
criminant function coefficients were:
1. camping + .015
2. boating - .020
3. golfing - .021
4. ocean beach
activity + .011
5. salmon fishing + .011
6. trout fishing + .018
7. change of spending
variable - .034
8. highest education + .014
9. occupation + .012
10. sex + .027
Before commenting on the implications of the discriminant function
analysis, it should be pointed out that the signs on the discriminant func-
tion coefficients are not to be viewed like the signs on regression func-
tion coefficients. The discriminant function coefficients do not provide
similar qualitative information about relationship.
Precise interpretation of the ten variables shown above that discrim-
inate in the comparison of means between fhe set of respondents who changed
travel plans en route and those that did not is not clear. However, six
of the strictly outdoor "planned activity" variables emerged. Also, the
sex, occupation, and education variables emerge as important and, along with
the important planned activity variables suggest an outline of a behavior
model.
H-28
-------
Since the F analysis revealed a "fit" that was at least marginally sig-
nificant, it was decided to rerun the analysis of variance problem using the
subset of 10 variables listed above as the set of independent variables in
an effort to obtain an improvement in "fit" and to examine whether this might
change the size of the discriminant function coefficients.
The second run did increase the sizes of most discriminant coefficients.
More important, however, the computed value of F at 9 and 391 d.f. was 1.94.
The critical value of F at the 5% level is 1.90. Under the model employed
in this analysis, the data do not allow us to reject the hypothesis of equal-
ity of means. There is no suggestion that the respondents who change vaca-
tion plans are, in fact, from a different population than those who do not.
Pollution Bad Vs. Not So Bad
The same original set of 20 independent variables were used in an effort
to predict those respondents who viewed pollution as quite or very serious
in Oregon as opposed to those who viewed it as a less serious problem. In
this case, the MANOV test for equality of means provided some statistically
interesting results. The computed F value at 20 and 380 d.f. was 2.26 while
the critical F at the 1 percent level was 1.92. The fit is particularly
good. The conclusion is that there is a difference in mean values. The
suggested implication is that the set of respondents who view pollution in
Oregon as a serious problem comes from a different population than the set
that view pollution as not too serious.
The computed discriminant function reveals five variables that seem more
influential in accounting for this behavior difference. These are:
1. golfing + .021
2. hiking & walking - .010
3. pleasure driving
& sightseeing + .029
4. trout fishing + .012
5. education - .032
In this attempt to discriminate between respondents that view pollution
in Oregon as bad as opposed to those who view it as not so bad, the five most
influential variables turn out to include four "planned activity" variables
and one relating to personal characteristic. These five variables do not
readily reveal a model for differentiating between the two possible popula-
tions suggested here.
H-29
-------
It is interesting to note that the variables for "pollution at home
and "relative air pollution in Oregon" were not revealed as particularly
influencial in the discriminant function analysis. This is consistant with
comments derived from the open-ended questions relating to air pollution
that suggest it is relative air pollution that is important.
As in the analysis of the change vs. non-change data, the pollution
bad vs. not so bad problem was rerun using the subset of 5 variables listed
above as the independent variables. On the second run, the computed F value
with 5 and 395 d.f. was 7.40. The critical F value at the 1 percent level
is 3.06. These results are highly significant and require rejection of the
hypothesis of equality of means. They also suggest something about differ-
ences in the populations of respondents.
Interaction
As suggested above, an "interaction" analysis was performed in which
we attempted to discover a systematic relationship between those travelers
who changed travel plans and saw pollution as bad in Oregon on the one hand
and those who did not change plans and did not view pollution in Oregon as
bad. This analysis revealed a very poor fit. The computed F value was 1.22
with 20 and 380 d.f. while the critical value of F at the 5 percent level
was 1.60. The hypothesis of equal means must be accepted. There is no
evidence of systematic difference provided by these data.
Data Analysis Extension
For purposes of this research, it made sense to partition the data along
the lines provided by the "change in vacation plans" question and the ques-
tion relating to perceived air pollution in Oregon. Upon completion of this
data analysis, another data partition was suggested. That is, the question
was raised of whether the location of interview site made a difference in
data configuration. To respond to this question, the data were partitioned
between those respondents interviewed in the Willamette Valley and those in-
terviewed outside the Valley. The MANOV analysis performed on a data parti-
tion based on the geography of interview site provided some interesting re-
sults .
The data set was partitioned by geography at interview site and subjected
to the multivariate analysis of variance technique. A broad comparison was
made between Willamette Valley respondents and "outside" respondents. This
analysis produced very significant results. This analysis required rejection
of the hypothesis of equal means in the independent variables and, in effect,
suggested that the set of respondents interviewed in the Willamette Valley
came from a different population than the set interviewed outside the Valley.
H-30
-------
The discriminant function analysis revealed a set of seven variables as
contributing importantly in the comparison. This set included the planned
activities of camping, golfing, ocean beach activities, hiking and walking,
swimming, and pleasure driving. In addition, the variable reflecting per-
ceived pollution in Oregon emerged as a strong variable. More close examina-
tion of the data on these seven geographic specific variables provided one
interesting fact. Twenty-five (25) percent of respondents interviewed in the
Willamette Valley observed that Oregon had either a "very" or "quite" serious
air pollution problem. Thirteen percent of respondents interviewed outside
the Valley made the same observation. The interpretive conclusion from this
is not profound. The implication is that, if one asks people about their
perceptions of air pollution and the question is asked while the respondents
are in a so-called pollution zone, people will comment about the pollution.
More importantly, however, the mere tendency of observing pollution is ap-
parently not enough to require people to change their behavior.
Subsequent analyses comparing respondents in the Willamette Valley who
viewed pollution as bad as opposed to those interviewed outside the Valley
who viewed pollution as bad as well as the corresponding comparison of re-
spondents who viewed pollution as less severe both interviewed in the Valley
and out, produced significant differences in mean values of independent
variables. The discriminant function analysis also suggested that the ob-
served differences in mean values could be accounted for by differences in
the set of activities vacationers planned to engage in, their comparison of
air pollution in Oregon as opposed to their home area, education, age and
sex respondents. One inferential interpretation of this analysis relates to
the explanation of the variable configuration suggested by discriminant
function analysis.
In all the comparisons made thus far and for those that produced con-
clusions of significant difference in mean values of independent variables,
a number of the so-called planned activity variables continued to emerge
as apparently important in explaining the differences observed. This data
analysis that makes comparisons premised on a data partition specified by
geography of interview site may suggest that the respondents may not be
really different in terms of tastes and preferences but, rather, that the
location of interview site influences the weighting that respondents pro-
vide to questions about planned activities. That is, we have observed
through responses provided to other questions that the party visiting Oregon
usually makes a "loop" through the state. That is, they will utilize dif-
ferent highways into the state than they will out of the state. Their path
may provide them an opportunity to experience or brush close to activities
as extreme as ocean beach activities on the one hand and camping in the high
Cascades on the other. If the interview process "catches" a party On the
Oregon coast where there is ample opportunity to observe and enjoy ocean
beach activities or salmon fishing, it seems likely that their responses to
questions in these areas would provide more weight to these activities.
Other illustrations could be provided. The point of this is to suggest
that the site' of interview may substantially affect the data configurations
that we are observing for the set of planned activity variables.
H-31
-------
V. SUMMARY AND CONCLUSIONS
The primary motivational factor influencing the funding of this broad
project on the economic effects of air pollution was a recognition that
measuring the economic magnitude of the presumed effects of air pollution
was very difficult. Previous research on the measurement of air pollution
impacts had not reduced interest in this question.
This portion of the Oregon State University project has relied heavily
on the theoretical implications of a model that attempted to tie air quality
(or other perceived environmental factors) to the set of activities that
individuals and households consume. The underlying thesis of the research
has been subsumed in the joint propositions that (1) if air pollution really
does make a difference to individuals, one should be able to observe varia-
tion in the kind of activities people consume as a result of variation in
air quality and, (2) if the activity consumption set does change as a result
of changing air quality, one should theoretically be able to observe the
market consequences of that changed behavior. As a theoretical corollary
one should be able to estimate the dollar cost to the individual and to
society induced by changing air quality. The empirical methodology for
approaching these propositions proceeded along two main lines.
The regression analysis discussed in Section III was intended to
examine the first of the above propositions. Those regressions constituted
attempts to discern qualitative relationships between visibility (our proxy
for air quality) and attendance at a selected set of recreation related
facilities. Again, the purpose here was to establish an empirical tie be-
tween a set of arbitrarily specified outdoor related public facilities and
air quality. The regressions were not intended to provide estimates of
the concentration of pollution per se.
The second proposition outlined above was approached through an empiri-
cal method that involved the study of a set of out-of-state visitors to
Oregon for pleasure during August 1971. This procedure extends from the fact
that both the extent and the quality of air pollution in Oregon varies. None-
theless, both residents and non-residents alike acclaim the great bulk of the
Oregon environment as reflecting something of an environmentally pure environ-
ment. Many non-residents visit the state each year to participate in activi-
ties that utilize the environment.
Given that the Willamette Valley reveals rather marked seasonal air
pollution during the summer months as a result of field burning and that this
burning occurs at a time overlapping with the peak tourist flow, it made sense
to attempt to test the broad hypothesis about the effect of air pollution on
consumer behavior by studying the behavior of out-of-state tourists traveling
in Oregon for pleasure during the summer months and trying to relate this to
air quality.
The results of the analysis of survey data do not support either the
main theoretical hypothesis predicting that deterioration of air quality
would lead consumers of outdoor activities to alter their behavior or the pre-
conceptions that seem to prevail about the effects of air quality. Of the 401
H-32
-------
parties interviewed in this study while traveling in Oregon for pleasure,
56 apparently encountered parameter changes of sufficient magnitude to
warrant changing their vacation plans en route. Of the 56, only 6 parties
indicated that air pollution had anything to do with their behavior.
These 6 responses cannot be viewed as significant in any statistical sense
of the word.
The power of preconception in structuring one's thinking is strong. In
terms of the objectives of this research, the statistical results cited here
on the effects of air pollution on the behavior of tourists is negative. One's
initial reaction is to look for mitigating circumstances. Are there reasons
that would account for our inability to observe the anticipated statistical
relationship between air quality and tourist behavior? The answer to this
could be "yes" and the explanation may lie in both the condition of the
Oregon atmosphere during the interview period and the character of the sample
population.
Regarding the atmospheric conditions prevalent in Oregon during the
interview period, it is possible that the period was relatively pollution
free. This research has suggested at several points that primary source
of air pollution during the summer months in the Willamette Valley is the
smoke produced by burning grass seed fields. Automobile exhaust and the
smoke from saw, pulp, and planing mill operations also contribute signifi-
cantly on a regional basis. The relevant unanswered question for this re-
search relates to the condition of the atmosphere over the period August 19
through August 31, 1971. That is, is there reason to believe that the inter-
viewing period was characterized by relatively clean atmospheric conditions
and hence provided tourists, at least in the Willamette Valley, with rela-
tively low values of any pollution index they might compute.
In an attempt to provide partial answer to this question, we have ex-
amined U.S. Weather Bureau data relating to the relevant atmospheric condi-
tions and taken at the airports at Salem and Eugene over the relevant span
in August, 1971. In Table I below, we have provided data on low visibility
by day and city embellished with editorial observations made by weather
bureau observers on occasion. For purposes of this array, low visibility
data are taken from the daylight hours of 0700 through 1900 in order to
provide reasonable simulation of the environmental backdrop observed by
travelers in the Willamette Valley area in the period in question.
Examination of the visibility data and weather observer comments from
the Salem and Eugene airports reveals that the interview period was marked
by numerous days in which smoke and/or rain obscured the visibility in the
Willamette Valley. Comments about the existence of fires and smoke are more
prevalent from the Eugene station. We should note that the qualitative
content of the editorial remarks may be a function of variance in weather
station personnel or local procedure. Another point of observation, the
'low visibility' data for Eugene appear higher than the data for Salem on
the average. This might be accounted for by a difference in the availa-
bility and distance of usable distance markers in the relevant sectors
surrounding the airports. On the average, visibility in Eugene is probably
lower than it is in Salem. In addition to the smoke from field fires, Eugene
H-33
-------
TABLE IV-1
Low Visibility and Observer Comments
Taken at Salem and Eugene, Oregon, Airports
August 19-31, 1971*
SALEM
Date
8-19
8-20
8-21
8-22
8-23
8-24
8-25
8-26
8-27
8-28
8-29
8-30
8-31
Low
Visibility
10
6
7
10
IS
10
'5
8
10
4
20
3
10
Comment
K 50-®
RW;
R;Fk Ese-S
RW;
Vsby SE 3
K; K20-®
K layer; smkv
RW; Smkv
K
K; RW
RW
EUGENE
Low
Visibility
10
25
15
10
12
15
3
10
7
3
25
10
10
Comment
fires N&S K Iyer
fires NW-NE vsby NE %K
fires 1156-1856 RW; K
RW;
SE S Smky
Fires W&N 1156-1356; Vsby
Lwr NE & SE K
K
Fires N 0956-1055
RW
FK
R
RW
* Explanation of Meteorological terms and abbreviations
Abbreviations - in remarks and visibility columns:
9 - overcast, i.e., 10/10 sky cover
0 - broken, i.e., 5/10 to 9/10 sky cover
R - moderate rain.
RW - moderate rain showers.
K - smoke.
F - fog.
Shortened spellings:
VSBY - visibility.
SMKY - smokey.
LWR - lower.
LYR - layer.
Visibility - given in statute miles and fractions of miles. The visibility
given in a weather observation is defined as "Prevailing visibility" and is
the highest visibility which can be maintained throughout at least half of
the horizon.
Source: U.S. Weather Bureau
H-34
-------
is plagued with considerable airborne effluents from pulp, planing, and
plywood mills. Again, the data from Table 1 suggests that visitors to the
central Willamette Valley during the interview period were confronted with
obscured visibility. This was primarily accounted for by the smoke from
field burning fires and from background pollution provided by industry and
autos.
Given that there was a prevalence of obscured visibility during the
interview period, our attention is drawn to the question of whether the
sample population reveals characteristics that might help account for the
relative failure of respondents to be affected by the pollution they
encountered. As indicated above and in Appendix IV, 54 percent of all
respondents indicated California as their home state, 20 percent were from
Washington, with the balance distributed over the rest of the nation rela-
tively evenly. Perhaps more important, 68 percent of all respondents in-
dicated living in cities with populations in excess of 25,000. Fifty-six
percent indicated living in cities in excess of 50,000 population. Assum-
ing that there is a positive relation between home city size and airborne
pollution experience, one might expect the bulk of the sample to have been
exposed to heavier amounts of air pollution in their home areas.
One could make an even stronger assumption about the experience respond-
ents have had with air pollution. For example, one could assume that the
California visitors on the average and the visitors from metropolitan areas
in other states reside in areas where pollution is markedly worse than in
Oregon. The questionnaire data provides some support to this hypothesis.
For example, in response to Question 26 which asked about the relative
seriousness of the air pollution problem in their home area, 55 percent of
all respondents indicated either "quite" or "very" serious. Twenty-eight
percent indicated "not too" serious a problem. Question 27 of the inter-
view instrument asked about the relevant air pollution between Oregon and
the home area. Sixty-one percent of all respondents indicated there was
less pollution in Oregon, 15 percent indicated it was a problem of equal
magnitude, while only 18 percent indicated Oregon had more pollution than
their home area. These data, along with interpretations derived from the
qualitative content of verbal responses to the pollution questions, indicate
that, regardless of the amounts of smoke prevailing in the Oregon atmosphere
and the absolute minimums on horizontal visibility, visitors included in
the sample found less pollution in Oregon than they had observed in their
home area.
One can conceive a behavioral model in which there are parameters re-
flecting tastes and expectations. If one views the model as representing
an optimization problem and one that is being optimized through a linear
or non-linear programming technique, it is possible for these parameters in
the system to change over some small range without producing changes in
behavior. It is possible that the threshhold values of the parameters—i.e.,
those values of the parameters which, if reached, produce a change in be-
havior—are established on the basis of experience in the home area. For
those system parameters that are affected by atmospheric quality and for
visitors that find the atmosphere in Oregon "cleaner" than the home areas,
one might expect the parameters of the relevant behaviorial systems to move,
but not by enough to reach the threshholds requiring behavioral change.
H-35
-------
In terms of the hypotheses about the effects of air pollution that this
research set out to examine, the conclusions must be viewed as negative. In
terms of scientific investigation, negative conclusions are not viewed as
disappointing per se.
In spite of the results of this analysis, this writer maintains the
prejudice that there are social costs associated with air pollution. If
this prejudice is in fact true, the question remains of how to measure those
costs. The negative results produced here may provide a suggested direc-
tion. This research accepted on an
-------
VI. POSTSCRIPT
To repeat the conclusion expressed above, as far as this research
has gone, air pollution has not revealed itself as particularly influ-
ential in affecting the set of consumption activities that consumers
enjoy. This conclusion extends from the analysis of both the regression
analysis and the analysis of questionnaire-derived data.
From the outset of this research, one of the innovations that this
author contended to offer to this research on the impact of air pollution
was the manner in which pollution was treated in the theoretical model
relating air pollution to consumer behavior. Pollution was regarded to
affect tastes for various classes of consumption activities and, hence,
served as a shift parameter that could shift the relevant demand curves
to the right or left, depending upon direction of change of pollution
and the class of activity.
The empirical models employed in this research have been defended
on more than one occasion as "reasonable" on both theoretical and empiri-
cal grounds. This assertion still applies; however, a footnote is
required.
In both the regression and the MANOV analyses outlined above we
have, in effect, been attempting to relate variation in air quality
with certain predicted directional changes in the consumption level of
various classes of activities. In terms of traditional economic jargon,
we have been trying to observe the logical analogs to traditional sub-
stitution and income effects in the analysis of demand curves. It
should be pointed out, however, the analysis of substitution and income
effects takes place in a framework in which tastes and preferences are
assumed constant.
After reading a draft version of the above material, my friend,
colleague, and co-author on some materials, R. Charles Vars, observed
that it was inappropriate to set up an empirical research model de-
signed to examine traditional relationships such as is provided by the
substitution and income effects of demand curves when, by hypothesis,
we were going to allow tastes and preferences to vary as a function of
air pollution. What difference does this observation make?
The importance of this observation relates to the problem of
measuring the magnitude of economic impacts of air pollution on consumer
behavior even if those impacts exist. Air pollution has been treated
here as a shift parameter. That is, air pollution is treated much like
income in the traditional analysis of demand curves. A change in this
parameter should theoretically result in a lateral shift of the demand
curves for several classes of activities. For those classes of demand
curves affected by variation in the pollution parameter, we are saying
that consumers would "consume" more or less of the activity (depending
upon class of activity) at any given "price". This is a statement
H-37
-------
having qualitative content only. The regression and MANOV analyses
performed above were designed to discover empirical support for this
proposition. This empirical work was not intended to provide quanti-
tative estimates of the cost to consumers of switching between con-
sumption activities as a result of variation in air quality — even
if the analysis had provided strong evidence of switching.
If one wanted to estimate the dollar cost of the impact of air
pollution on consumer behavior, he would essentially be committed to
an argument in which the gross cost of air pollution to an individual
consumer would be viewed as the sum total of the air pollution-related
cost changes associated with the consumption of the relevant separate
consumtion activities. The separate cost changes associated with each
consumption activity affected would be computed, in terms of the usual
theory, as the dollar cost of the sum of the income and substitution
effects that result from varying the pollution parameter in the system.
This is what Professor Vars1 observation was about.
Even if there were data available to estimate the dollar cost
to consumers of the pollution model counterpart to income effects,
there is still the problem of estimating the dollar cost associated
with the pollution model analogs to the substitution effects. Given
the character of the model in which pollution is viewed as shifting
the relevant demand curves, there is, by implication, no opportunity
to estimate the sizes of those substitution effects. This last prob-
lem is complicated, according to the empirical observations outlined
above, by the apparent absence of a continuous relationship between
the pollution parameter and the set of consumption activities. Given
the character of pollution variation and the character of the data set
on the purchase and use of inputs to the consumption set, it is
apparently not practical to observe or estimate the magnitudes of the
relevant substitution effects. The corollary to this proposition is
that coming up with meaningful estimates of the cost of air pollution
to consumers may not be practicable. This is a substantive conclusion.
Its importance should not be minimized.
H-38
-------
APPENDIX 1
The Effects of Changes in Air Quality
on Tourist-Related Industries in a Region
-------
The Effects of Changes in Air Quality on
Tourist-Related Industries in a Region:
A Theoretical Analysis*
Gary W. Sorenson
Economics Department
Oregon State University
*The work upon which this paper is based was performed pursuant
to Contract No. CPA 22-69-86 with the National Air Pollution Control
Administration, CPE, Public Health Service, Department of Health,
Education and Welfare.
-------
There is a presumption, generally accepted, that deterioration in air quality
has among its effects an impact on the kind of consumption activities enjoyed
by consumers. If this is true and the activities affected negatively are
supplied by an industry that contributes significantly to a regional GNP, then
regional economic planners would be well advised to implement policies to protect
the industry from the external diseconomies being generated by other industries.
Like most policy arguments having high intuitive appeal, this logic has its
flaws.
Presumably it is the region's welfare that is of interest. Then the benefits
of relieving the burden of external diseconomies from one industry must be
viewed vis-a-vis the costs to the originating sectors. Hence, one's policy
analysis cannot afford to be "partial" in perspective.
Perhaps a more basic flaw in the above argument., however, is that it is
based on casual empiricism—largely on a presumption. What theoretical justi-
fication is there to think reduced air quality does affect consumption patterns?
And even if it does, how large is the effect? How would one test for whatever
hypotheses might apply here? This paper is addressed to these questions.
The objectives of this paper are three-fold. First, 1 will attempt to
establish some of the theoretical effects of variation in air quality on the
allocation of non-work time and income by individuals and families. Put another
way, this objective can be viewed as an attempt to discern whether some of the
social costs associated with consumer activities are theoretically meaningful.
Second, I will try to relate these theoretical effects to a regional economy
such as the economy of the Willamette Valley in Oregon. This objective will
necessitate that the theoretical model will be rather specific. That is, both
the model with which we will work and the implications that are drawn are not
expected to hold in all cases at all times but will necessarily draw upon the
unique characteristics of the region in question as well as other known insti-
tutional and social restrictions. Third, I will attempt to outline possible
testing procedures and the data necessary for testing any emergent hypotheses.
Any empirical testing of hypotheses, while not to be performed here, would
constitute an attempt at testing for the existence if not the size of social
costs associated with a class of consumption activities—i.e., these tests
would be examining hypotheses of the type 'total cost Xj^ market cost X^,'
where X^ is a manufactured commodity. As a point of departure, we will begin
with a relatively recent contribution to the neoclassical theory of consumer
behavior. This model will extend from Gary Becker's work [Economic Journal,
September 1965], I have modified the Becker model for analytical convenience.*
*For my treatment of the Becker model see my Income Changes and Labor
Force Participation, a monograph forthcoming from Oregon State University
Press.
-------
Assume an individual with utility function,
1) U - FU), j - l...n.
Regarding F, we assume this function to be continuous and quasi-concave.
Contrary to the neoclassical theory, the arguments in F are not market
commodities per se. They are activities. These activities are both pro-
duced and consumed by the individual. From the production side, the indivi-
dual is seen to combine market commodities, x^ j , and time,TnJ, in combinations
that are technically defined, and produce the consumption activities, Z . That
is, for each j there is a production function and we can represent this as,
2) Z = f (x.. ..,x ; T . . .T .) or, in vector form, Z = f.(x., T').
If the technologies represented by the f . are linear, the x. and T' can be
represented as, •* J J
3)
T5
In (3), b. and T'. are vectors of technical coefficients.
The individual is assumed to maximize F but this process is subjected to two
or more constraints. In the simple general case, the constraints are,
j
4) I t.Z. + ET = T, time constraint
J J s
j
I, c.Z + Zw T = G, Income constraint.
j j s s
In (4), the new notation is t. = T'*l, T = time worked in occupations,,
c.t = dollar cost per unit of market commoaities used in Zj and ws = the wage
rate applicable in occupation s — we are assuming both average and marginal
wage rates are equal. G is 'other' nonwage income and T is total time.
The optimization problem viewed here can now be represented as follows:
5) Maximize
u = F(Z.J)
s.t.
I tjZj + £Tg - T
EwsTs
Z >0f Tg>0.
-------
The consumer's problem represented here is a non-linear programming problem.
That is, it has a non-linear objective function, two linear constraints and
the equilibrium values of the independent variables, Zjand Ts, are permitted
to enter an optimal solution at the zero level. At this juncture, we must
make some statement about the time period being considered. In the neoclassical
theory of consumer behavior, the relevant time period is the 'income period.'
In its general form, the problem expressed in (5) can have any reasonable time
span. That is, by setting T equal to the sum of hours in a week or month,
the time period is specified. However, even here there would be an implicit
assumption about regularity in the work-consumption-production process. Certain
habitual or institutionally imposed constraints would be recognized—i.e.,
people may arbitrarily impose sleep periods on the problem between 11 p.m.
and 7 a.m. or employers may require that the time period from 8 a.m. to 5 p.m.
Monday through Friday be arbitrarily blocked out for work. In the present
analysis, however, we are particularly interested in non-work and vacation
type activities. Hence, because of the present orientation, the optimization
period being viewed here is one in which T =0 and total time is viewed as
being allocated to consumption and production of activities. This implies
that all income is viewed as 'other1 income—an observation that will not
affect the analysis.
Let us impose an additional simplifying assumption on the present framework
by viewing the consumer's problem as one of finding a conditional or planned
equilibrium configuration to his optimization problem. If one's vacation
involves consumption of activities that are new or novel in the experience
field then, to the extent that they are included in the planned equilibrium
configuration, the initial decision made toward these activities is dependent
upon a set of parameters whose values are, perhaps, based upon incomplete
knowledge. For example, actual technical coefficients relating market goods
and time to activity levels may be different than the assumed values. When
experience provides accurate values for the coefficients, a person may alter
his planned level of activity consumption. By viewing this problem as es-
tablishing a conditional equilibrium, we leave more flexibility in allowing
for the effects of experience on equilibrium activity levels. In this frame-
work and in terms of a generalized Lagrangean function, the problem evolves to
6) V(Z, A) = F(Z ) 4- A (T - Zt Z ) + X (G - Zc Z ), to be maximum
j j j y j »j
s.t.
Z., \ > 0
j ~
j « 1... n.
If an optimal solution for this system exists, Z and X , then we must have
the following: *
7) V(Z, X°) <.V(Z°, X°)<_V(Z°, X)
and
-------
8)
3F
, , j .
In (6) and (7) the analysis has required the addition of the multipliers
T and A . It can be shown that, in terms of this analysis, A is the
marginal utility of time and A is the marginal utility of income.
In (7), we have also added some set notation. We have added the set J
which is the subset of indicies j that enter the planned equilibrium con-
figuration at positive levels and J which is the subset of j that enter
at the zero level. Analysis requires A , A and Z to be non-negative.
The Kuhn-Tucker necessary conditions for anyequilibrium are provided in (8).
If optimal solution vectors, Z and A , exist to the problem expressed in
(7), then those solution vectors must be such that conditions expressed in
(8) hold. The conditions in (8) are the so-called Kuhn-Tucker conditions.
These conditions are very similar to those that hold in the traditional con-
sumer equilibrium. That is, for activities consumed at the positive level
the marginal utility of each activity must equal the weighted sum of the
marginal utilities of time and income—the weights are the appropriate
technical coefficients' for time and income respectively. For activities
consumed at the zero level, the marginal utility of these activities is,
reasonably, less than the appropriate weighted sum of the time and income
marginal utilities. The equilibrium configuration of consumption activities
as well as the optimal values of the relevant marginal utilities are clearly
a function of taste, income, the relevant time span, and the technical
coefficients relevant for each activity. Variation in any of these can
lead to variation in optimal consumption levels of any of the activities.
Before considering the question of sensitivity analysis let us inquire
further into the character of this equilibrium configuration.
The optimal solution vector, Z , describes the planned production-con-
sumption levels of activities for this assumed vacation period. With no
further information, there is very little we can say about Z° except the
obvious. That is, it will necessarily contain some sleep and eating activities
but the character of these activities is not revealed in this general pre-
sentation. Among the set of all vacationers, there will be many subsets.
We can conceive a subset with strong preferences toward sleeping and, per-
haps, gardening; this subset is of no interest here. A second subset may
have preferences for and income only adequate to enjoy activities in some
close proximity to their home. We can also conceive a subset having tastes
for and income -adequate to enjoy travel and outdoor activities. It is this
group that we want to examine.
-------
If an optimal solution to (7) exists then the Kuhn-Tucker conditions of (8)
hold. Let us denote the marginal utilities of activities expressed in (8)
as $j These are constants since they are evaluated at a particular point.
In the theory of non-linear programming, the optimal solution need not
be regarded as permanently fixed. The theory permits variation of parameters
and the optimal solution is regarded as holding within some small epsilon
area around the optimal solution. The shortest approach to sensitivity
analysis would be to regard the Kuhn-Tticker conditions of (8) as the con-
straints of a linear programming minimizing problem. In this case the
variables are the marginal utilities of time and income and the prices in
the objective function are expressed in differential terms. This is shown
in (9) below:
(9) tX + CX t
t Xm + c X > d>
n T n y — yn
dTX_+ dGX = dU (min)
'T, \ i °
All linear programming problems that have an optimal solution have associated
with them a dual problem which also has an optimal solution. The dual may
be more convenient here. This is shown in (10) below:
10) t.dZ....+ t dZ < dT
11 n n —
c,dZ....+ c dZ < dG
11 n n —
4> dZ.^..* <(>ndZn = dU (max)
In (10) the variables are differentials in consumption activities and the
prices in the objective function are the relevant marginal utilities of the
activities. This is a maximizing problem.
Parameters that can effect the equilibrium solution include the technical
coefficients, available total time, income, and taste for the various con-
sumption activities—taste is embodied in the marginal utilities, t«. tf
parameter changes occur and precipitate changes in optimal activity'levels,
the non-linear programming analysis requires a compensating set of changes
in the vector \ . As it turns out, it is easier to analyze the impact of
parameter changes on activity levels by looking at the impact on the multi-
pliers, A^, and X .
* *•
The dimensions of these multipliers can be easily obtained from (10). De-
fining B as a basis matrix from (10), it can easily be shown that,
11) X - B"1.
-------
In this case, the basis matrix would have coefficients from two columns
of the constraints in (10) and associated with activity levels consumed at
positive levels. For convenience, assume the basis includes! coefficients
for activity 1 and activity n. The basis matrix would then be,
B
±,t
1 n
cnc
1 n
Since all bases from (10) would "look alike," assume B is the optimal basis,
In this case we would have,
12)
*lcn " *ncl
Vn - Vn
Vl "
Vn - Vn
It is usual in the analysis of consumer behavior to treat air quality as
constant and independent of the optimal consumption set. Casual experience,
however, seems to deny both the assumptions. To discern the effects of
variation in air quality on the optimal consumption set, though, an additional
assumption must be made. It seems reasonable to assume that deterioration of
air quality, as measured by an increase in either eye irritation or horizontal
visibility would have negative or zero effects on optimal levels of consumption
activities. Put otherwise, we are postulating that an increase in either eye
irritation or a reduction in horizontal visibility, will have negative or zero
effects on the marginal utilities of activities. With this in mind let us con-
sider the effects of a general change in the "prices" of (10) and, specifically,
the effect of a change in the prices _ in (11). Should something occur to
change the marginal utilities associated with the optimal consumption set, the
effect would be as follows:
13) AA
-1
In this two constraint problem and with the assumed optimal basis matrix, B,
the effects of (13) are,
14)
Prior to looking for empirical implications in (14), a word on the possible
effects of parameter changes. A change occurring in the vector 4, , if 'small',
may leave the optimal solution vector, Z , unchanged in terms of its activity
composition — only the levels of activity consumption may change. On the other
hand, "large" changes in $ may require a whole new solution to our programming
problem. Similar observations would apply in the case of other parameter changes
The important thing is that we have no a priori way of knowing what the effect
of any given vector change will be on the optimality of a given solution. We
are able to make some statements toward the possible changes in activity levels
if the equilibrium configuration remains unchanged. With this in mind, we will
begin looking at (14) and consider the theoretical effects.
-------
In (14), the term k is the determinant of the basis matrix, B. It cannot
be zero. It can be positive or negative. If k is greater than zero, this
implies t^c^ .. t^c^ > 0, or, alternatively, c /t > c /t^ In words, this
says that dollar expenditures per unit of timeninnthe nth activity are greater
than the dollars per unit of time in the first activity - i.e., the nth activity
is relatively income intensive while the first good is relatively time inten-
sive. The relative income or time intensity of the goods is important here.
If k is less than zero, we are implying c1/t1 > c It — this implies the rela-
tive time and income intensiveness of these goodsnisnreversed. In any event,
a priori, the sign of k is unknown. Suppose k > 0. If something should happen
to <{> to reduce it, ceteris paribus, and the change is sufficiently small, we
can observe the effects on the relevant marginal utilities of time and income
by way of (14). These are:
15) AXT > 0
AXY < 0 .
By assuming k greater than zero we have postulated that the nth activity is
"relatively income intensive" and the effect of having the marginal utility
of this activity go down has been to raise the marginal utility of time and
reduce the marginal utility of income. If the initial change had been in -,
the opposite set of sign changes would have occurred in the marginal utilities.
Please note, the sign changes in the marginal utilities depend on both the
activity whose marginal utility has changed and the sign of k. But it is
not the activity so much as its relative time intensity or income intensity
that makes the difference. If k had been assumed to be negative, this would
have implied the opposite relative time and income intensities for our two
activities and in that case a reduction in $.. would have raised the marginal
utility of time and lowered the marginal utility of income. This is perfectly
consistant with the implications already observed.
The changes in the two marginal utilities were precipitated on a change in
the marginal utility of one consumption activity. These changes, expressed
in (15), have a straightforward interpretation but a word of warning is
justified. One must divorce his thinking from the prejudice provided by
the sensitivity analysis performed on the traditional theory of consumer
behavior. There, parametric changes generally occur in income or market
prices and if one observed a reduction in the marginal utility of income
this would have implied an increase in the consumption of goods using income.
In the present case, the technical income and time resetrictions have been
held intact while the utility function has been rotated in favor of one good
and away from another. That is, since the utility function has been changed
pne cannot rely on the assumed quasi-concavitity of that function to inter-
pret changes involving time and income. In the present case, the reduction
in $ has had the effect of lowering the opportunity cost of time intensive
activities (Z.)-with respect to income intensive activities (Z ). That is,
one's preference for time or time intensive activities has increased while
his preference for income or income intensive activities has decreased. This
is tantamount to predicting, as we would expect, that the consumption of '/.-
should increase while the consumption of Z should decrease.
-------
The implications of this are illustrated diagramraatically in Figure 1, shown
below. The two constraints in Figure 1 have been drawn with the relative
slopes implied in having k > 0. The point A denotes the consumer's equilibrium
prior to any change in preferences in this nonlinear programming problem. The
reduction in <(>n has had the effect of twisting the utility surface and the
consumer's new equilibrium must be at a point like B. That is, all indifference
curves have increased in relative slope and this must move the optimal point
to the right. The diagram illustrates what we have just said—that consumption
of Zj should increase while consumption of Z should decrease. In this case,
Z- is the time intensive activity while Z is the income intensive activity.
n
slope te - *
The above analysis has outlined some meaningful and potentially testable
implications. These are:
1) Tests for the existence of social costs of the type being con-
sidered here require the classification or specification of con-
sumption activities and some look at the relative income and time
intensities.
:2) Given (1), the theoretical implications outlined above can be
tested against empirical data on air quality and sales in various
product markets—and the hypotheses tested will be relatively 'clean1
or unambiguous.
3) If vacation type activities are 'income intensive,' analysis
suggests Oregon's economy will suffer in the face of reduced air
'quality. If these activities are 'time intensive,' the converse
is true.
4) The magnitude of these effects depends on the available alternative
activities and the character of available activities in adjoining states,
-------
APPENDIX 2
Selected Tables From:
OREGON STATE HIGHWAY DIVISION
Planning Section
Economics Unit
1971
OUT-OF-STATE
TOURIST REVENUE STUDY
-------
TABLE II
SUMMARY OF EXPENDITURES DATA FOR
OUT-OF-STATE TRAVELERS
1971
Of the 8,616 drivers interviewed, the number of vacationers were 5,087
Average Passengers Per Car 2.7
Average Miles Traveled in Oregon 409
Average Days in Oregon 3.4
Average Daily Expenditures Per Car $ 32.41
Average Trip Expenditures Per Car $ 110.19
Total Cars Outbound 3, 344,900
Total Persons 9,092, 800
Average Daily Expenditures Per Person $ 12.00
Average Trip Expenditures Per Person $ 40.81
Of the 8, 616 drivers interviewed, the number of business travelers 586
Average Passengers Per Car 1.6
Average Miles Traveled in Oregon 299
Average Days in Oregon 2.9
Average Daily Expenditures Per Car $ 29.22
Average Trip Expenditures Per Car $ 84.75
Total Cars Outbound 402,982
Total Persons 644,530
Average Daily Expenditures Per Person $ 18.26
Average Trip Expenditures Per Person $ 52.97
-------
D at a R e latedt o
Bu sT r a v e lers
Total Persons
Average Trip Expenditures Per Person
802,700
48. 09
Data Rj^atejc^Jio A 11^ J/isitorss andAll
1.
2.
3.
Out- of -State Passenger Cars
Expenditures on Recreation and Pleasure
Out -of -State Passenger Cars
Expenditures on Business
Out -of -State Travelers
Air, Rail, Bus
Total
$ 365,506,000
$ 34, 153, 000
$ 38,602,000
$ 438,261,000
The statistical data in the first two items in Table II were obtained from a survey conducted
during 1971 at all major highway exits from the State. The third item was obtained by analys-
ing information from various sources, including the National Transportation Census for 1967,
and preliminary survey data from the Port of Portland's Intermodal Transportation Study.
The statistical data contained in Tables III through XI pertain only to vacationists — out-
of-state passenger cars.
-------
CLASSIFICATION OF DATA BY STATES & REGIONS
The number of car parties from each of the states of California, Idaho and
Washington was large enough to give statistically valid results by classifying the
data from these states separately. In addition, Hawaii was classified separately
and Canada and Alaska were combined since they do not fit into any conventional
regional classification of states.
The following states were grouped by region (See Figure 1), so that mean-
ingful results could be obtained from the sample data:
ROCKY MOUNTAIN STATES
Arizona
Colorado
Montana
New Mexico
Nevada
Utah
Wyoming
SOUTH CENTRAL STATES
Arkansas
Louisiana
Oklahoma
Texas
Mexico
NORTH CENTRAL STATES
Iowa
Minnesota
Missouri
Nebraska
North Dakota
South Dakota
Kansas
EAST CENTRAL STATES
Illinois
Indiana
Michigan
Ohio
Wisconsin
5. NORTHEASTERN STATES
Connecticut
Maine
Massachusetts
New Hampshire
New Jersey
New York
Pennsylvania
Rhode Island
Vermont
6. SOUTHEASTERN STATES
Alabama
Delaware
District of Columbia
Florida
Georgia
Kentucky
Maryland
Mississippi
North Carolina
South Carolina
Tennessee
Virginia
West Virginia
7. CANADA AND ALASKA
8. NOT REPORTED
-------
DISTRIBUTION OF OIT-OF-STATE DRIVERS INTERVIEWED AND EXPENDITURES BY
STATE OR REGION 9F REGISTRATION
LEGEND
CAR EXPEMD-
PARTIES ITTJRES
1:
4.8% CANADA
AND
ALASKA
WASHINGTON
NORTHEASTERN
WEST
CENTRAL STATES
EAST
CENTRAL STATES
CALI-
ORN1A
ROCKY MOUNTAIN
STATES
SOUTHEASTERN
STATES
SOUTH CENTRAL
STATES
Figure 1
-------
OVERVIEW
Based on information obtained from an interview survey of out-of-state passenger
car travelers, it is estimated that out-of-state visitors traveling for pleasure spent
$365,506,000 in Oregon in 1971 for food, lodging, recreation, car expenses and other
purposes incidental to travel. The significant 1971 data are summarized below with
comparative data for 1969 and 1970.
1969 1970 1971
Expenditures for Recreation
& Pleasure $ 260,964,000 $ 272,761,000 $ 365,506,000
Out-of-State Car Parties 2,847,000 2,815,200 3,344,900
Total Persons 7,710,000 7,613,500 9,092,800
Average Days in Oregon 3.4 3.5 3.4
Average Trip Expenditures
in Oregon $91.66 $96.89 $110.19
These data show an increase of 18.8% in the number of out-of-state car parties visiting
in 1971 compared to 1970, and an increase of 17.5% compared to 1969; with a 34.0% increase
in total expenditures over 1970 and 40.0% over 1969.
As has been the case in recent years, tourist expenditures in 1971 increased, and this
year the number of tourists visiting Oregon showed a substantial increase. This increase in
expenditures is the result of a combination of factors including a large increase in the number
of tourists coupled with an increase in per capita spending by these tourists.
-------
The very substantial increase in the number of tourists this year of 19.4% over
1970 is in main due to the abnormally low number of tourists visiting the State in 1970,
which registered a loss of 1.3% over 1969. Thus, the average annual increase over
the two year period (1969-71) was a more modest 8.9% per year. It appears that the
large increase in 1971 put the industry back on its larger term growth curve with respect
to number of tourists.
The increase in number of tourists, then accounts for 19.4% of the 34.0% increase
in expenditures between 1970 and 1971 with the remaining increase being the result of
the increase in average expenditures per person per day.
-------
ORIGIN OF VISITORS
The data in Tables III, IV and Figure 1, show that the adjacent states of
California and Washington provide the bulk of Oregon's tourist trade. From these
two states come the greatest number of tourists, as well as the greatest contribu-
*
tion of tourist dollars spent in Oregon in 1971. <
In 1971, over 1.38 million car parties from California vacationing in Oregon
spent 184.7 million dollars, accounting for 41.4% of the passenger car tourists coming
to the State, and 50.5% of me total expenditures made by such tourists. The State of
i
Washington provided a significant 22.8% of the total car parties vacationing in Oregon
in 1971, but contributed only 12.5% of the expenditures due to their much shorter length
of stay; an average of 2.0 days in Oregon as compared to the 3.4 day average for all
visitors.
The Rocky Mountain States accounted for the third largest group of visitors to
Oregon. Motoring tourists vacationing from this area stayed an average of 0.5 days
more in Oregon than did the average visitor, and spent more than four dollars per day
per car party more than the average. Thus, while this area provided only 8.9% of the
car parti.es visiting Oregon, it accounted for 11.7% of the expenditures.
Following Washington, California and the Rocky Mountain States in dollar volume
of tourist trade, were the North Central States. This area had 4.4% of the out-of-state
car parties, and 5.1% of the total tourist expenditures in Oregon.
-------
After the North Central States, the greatest tourist expenditures came from re-
sidents of Canada and Alaska, followed in the order of dollar volume by East Central
States, Idaho, South Central States, Southeastern States, and the Northeastern States.
The regions having the highest dollar return per car party while in Oregon were
the Rocky Mountain, California and South Central, while the States of Washington and
Hawaii experienced the lowest. The primary factor contributing to this difference in
return per car party was that tourists from these regions stayed longer or had larger
car parties on the average than did tourists from Washington and Hawaii. To illustrate,
one car party from the Rocky Mountain States in terms of expenditures was worth about
2.4 car parties from the State of Washington.
-------
TABLE III
DISTRIBUTION OF CAR PARTIES AND EXPENDITURES BY ORIGIN
State or Region
California
Hawaii
Idaho
Washington
Rocky Mountain States
South Central States
North Central States
East Central States
Northeastern States
Southeastern States
Number of
Car Parties
(000)
1,383.1
2.0
123.4
763.0
297.0
91.0
148.2
126.8
82.3
80.6
%of
Car Parties
41.3
0.1
3.7
22.8
8.9
2.7
4.4
3.8
2.5
2.4
Number of
Persons
(000)
3, 898. 1
10.0
300.1
1,970.4
792.9
238.2
480.1
370.1
217.3
182.8
%of
Persons
42.9
0.1
3.3
21.7
8.7
2.6
5.3
4.1
2.4
2.0
$
Expendi-
tures
(000)
184, 653
37
13,523
45, 652
42,728
11,952
18,568
14,511
6,396
9,576
% of
Expend!
tures
50.5
- -
3.7
12.5
11.7
3.3
5.1
4.0
1.7
2.6
Canada and Alaska
243.2
7.3
622.8 6.8 17,581
4.8
Not Reported
4.3
0.1
10.0 0.1
329
0.1
TOTAL
3,344.9
100.0
9,092.8 100.0 365,506 100.0
-------
TABLE IV
AVERAGE DAYS IN OREGON, AVERAGE PEOPLE PER CAR, AVERAGE
DAILY EXPENDITURES PER CAR, AVERAGE MILES TRAVELED
IN OREGON, AND AVERAGE EXPENDITURES PER CAR PARTY
Average
State or Region
California
Hawaii
Idaho
Washington
Rocky Mountain States
South Central States
North Central States
East Central States
Northeastern States
Southeastern States
Average
Days in
Oregon
4.0
1.1
3.8
2.0
3.9
4.9
4.1
3.3
2.7
3.7
Average
People
Per Car
2.8
5.0
2.4
2.6
2.7
2.6
3.2
2.9
2.6
2.3
Avg. Daily
Expendi-
Per Car ($)
33.53
16.20
28.48
29.55
36.68
26.83
30.89
34.96
28.94
31.92
Avg. Miles
Traveled in
Oregon
441
341
431
313
449
43.1
472
406
450
455
Expenditures
Per Car
Party ($)
134.12
17.82
108.22
59. 10
143.05
131.47
126. 65
115.37
78.14
118.10
Canada and Alaska
2.4
2.6
29.92
394
71.81
Not Reported
3.1
2.3
24.31
377
75.36
TOTAL
3.4
2.7
32.41
409
110.19
-------
APPENDIX 3
Interview Instrument
-------
August 9, 1971
OREGON STATE UNIVERSITY
Budget Bureau Number OMB l!>H-'./1014
irial
Hello, I'm working on a survey for Oregon State University and I would like to ask you a few
interesting questions, if you don't mind.
1-
City
State
First, which city is your permanent residence
located in or near?
(NOTE TO INTERVIEWER: If respondent resides in Oregon, terminate interview; otherwise
continue)
2- 1 Recreation (Ask Q3)
H 2 Other only (TERMINATE)
(Specify )
3 Both recreation & other (Ask Q3)
Are you traveling in Oregon for recreation or for
other purposes?
1 Auto
2 Bus (commercial)
3 Airplane
4 Other (Specify
What type or types of transportation are you using
on your vacation or recreation trip in Oregon?
C4- Thinking back to just before you left on your vacation or recreation trip, which place or
places were you planning to visit in Oregon?
]5- Again, thinking back to just before you left on your vacation or recreation trip, how
many days were you planning to spend in Oregon?
'6- Again, thinking back to just before you left on your vacation or recreation trip, which
highway or highways were you planning to use in Oregon to get to your destination(s)?
7- And, which highway or highways were you planning to use to leave Oregon?
8- L_ QB_ L_I_ N. DK
43210 Camping
4321
0 Boating, Water
skiing
Golfing
Ocean beach
activities
Hiking, walking
Swimming
Pleasure driving
sightseeing
Salmon fishing
Trout fishing
2 1 0 Other (Specify _
4
4
4
4
4
4
4
4
3
3
3
3
3
3
3
3
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
Now, here are some specific out-of-door activities.
Thinking back to just before you left on this trip,
how much time, if any, were you planning to spend
participating in each of these activities in Oregon?
As I read each activity would you tell me how much
time you planned to spend on each one--a lot, quite
a bit, a little or none at all?
-------
9- $
Including everyone in your party, about how much,
in dollars, were you planning to spend in Oregon
for your vacation in the state this summer? Just
your best estimate?
10- Which place or places have you already visited in Oregon on this particular vacation or
recreation trip?
11- Considering your vacation plans now, which place or places are you planning to visit in
Oregon?
12- And, considering your vacation plans now, which highway or highways are you planning to
use to get to your destination(s) in Oregon?
13- Which highway or highways are you planning to use now when you leave Oregon?
14- How many days will you spend in Oregon?
15- L QB LI N DK
4~ 3 2 1 0 Camping
43210 Boating, water
skiing
43210 Golfing
43210 Ocean beach
activities
4 3 21 0 Hiking, walking
43210 Swimming
43210 Pleasure driving
sightseeing
43210 Salmon Fishing
43210 Trout Fishing
43210 Other (Specify
16- 1 More (Ask Q16a)
2 Less (Ask (]16a)
3 Same (Skip to Q17)
4 DK (Skip to Q17)
Here are some out-of-door activities. How much
time, if any, are you planning to spend now
participating in them in Oregon. As I read each
activity will you tell me the time you plan to
spend now on each one — a lot, quite a bit, a
little or none at all?
)
The way things look now, will you probably spend
more or less money on your vacation in Oregon than
you originally planned to?
I6,i- What are the main reasons you will probably spend (more) (less) than you originally
planned to? (PROBE)
17-
1 Changed (Ask Q18)
2 Not changed (Skip to 025)
3 DK, NA (Skip to Q25)
Have your vacation
you arrived in the
plans for Oregon changed since
state, or not?
-------
18- I have here a list of reasons some people have given for changing vacation plans in
Oregon. As I read each one, would you tell me if it had any effect or not on your
plans?
Yes No DK, NA
2
2
2
2
2
2
2
2
1 0
1 0
1 0
1 0
1 0
1 0
1 0
1 0
(INTERVIEWER:
Rain
Heat
Crowded facilities (What? )
Air pollution
Lack of money
Automobile breakdown or accident
Sickness in family
Other (What? )
Always ask #19 if code 2 on Air pollution is circled.)
If Air pollution code 2 is not circled, skip to #23)
19- What kind or kinds of air pollution in Oregon affected your vacation plans in the state
this summer? (PROBE for type and location of air pollution)
20- How or in what way or ways did air pollution in Oregon affect your vacation plans
this state? (PROBE!)
21- 1 Increased (Ask Q22)
2 Decreased (Ask Q22)
3 No difference (Skip to Q23)
4 DK, NA (Skip to Q23)
Would you say that air pollution in Oregon increased
or decreased the amount of money you planned to
spend for your vacation in the state this summer?
8
7
6
5
22- 4
3
2
1
$501
$301
$201
$101
$ 51
$ 26
$ 11
$ 1
or more
- $500
- $300
- $200
- $100
- $ 50
- $ 25
- $ 10
(HAND RESPONDENT CARD A)
Would you please look at this card and tell me
which one of these groups best fits the (increase)
(decrease) in dollars, your party is spending
in Oregon for your vacation because of air
pollution?
23- 1 Increase (Ask 024)
2 Decrease (Ask Q24)
3 No difference (Skip to Q25)
4 DK, NA (Skip to Q25)
(Not counting air pollution) Would you say that
the reasons you gave for changing vacation plans
in Oregon increased or decreased the amount of
money you planned to spend for your vacation in
the state this summer?
8
7
6
5
24- 4
3
2
1
$501
$301
$201
$101
$ 51
$ 26
$ 11
$ 1
or more
- $500
- $300
- $200
- $100
- $ 50
- $ 25
- $ 10
(HAND RESPONDENT CARD A)
(Not counting air pollution) Would you please
look at this card and tell me which one of these
groups best fits the (increase) (decrease) in the
amount of money you planned to spend for your
vacation in the state this summer?
-------
25- I
2
3
4
5
Very
Quite
Not too
Not at all
DK
From what you have seen in Oregon, how serious an
air pollution problem, if any, do you feel we have-
very serious, quite serious, not too serious or not
serious at all?
26- 1 Very (Ask 26a)
2 Quite (Ask Q26a)
3 Not too (Ask Q26a)
4 Not at all (Ask Q26a)
5 DK (Skip to Q27)
How serious an air pollution problem, if any, do
you feel you have in the area where you live--
very serious, quite serious, not too serious or
not serious at all?
26a How or in what way or ways is air pollution a
live? (PROBE)
_problem in the area where you
27- 1 More in Oregon (Ask Q27a)
2 Less in Oregon (Ask Q27a)
3 Same (Skip to Q28)
4 DK (Skip to Q28)
Thinking about clean air again, would you say then
is more or less air pollution in Oregon than where
you live?
27a How, or in what ways, do you think Oregon has (more) (less) of an air pollution problem
than the place where you live?
28- ASK OF EVERYONE
no. of persons
Including yourself, how many persons are traveling
in your party on this particular vacation or
recreation trip in Oregon?
29- 4 College - complete
3 College - partial
2 High School
1 Grade or no schooling
Would you mind telling me the last grade you
completed in school?
30-
_(Type of industry) What type of work does the chief breadwinner
in the family do?
(Specific job)
31- 1 20-29
2 30-39
3 40-49
4 50-59
5 60 & above
May I ask your approximate age?
32- BY OBSERVATION
1 Male
2 Female
-------
33- 1 Less than 30 minutes
.2 30 minutes - 1 hour Length of interview
3 1 hour - 1 1/2 hours
4 More than 1 1/2 hours
34- License number
35- 3 Positive
2 Neutral Rating on cooperation.
1 Negative
X I hereby certify this interview was actually taken at the following address and represents
a true and accurate account of the interview.
(Address(City)" (Interviewer's Signature;
Phone Number (For Name_
verification only)
FOR OFFICE USE ONLY: Interview verified by Date
-------
APPENDIX 4
Questionnaire Derived Data Arranged by
Closed Ended and Open Ended Type Questions
-------
A. Closed Ended Question Responses
Ql) Location and Size of Permanent Residence
State:
California
Washington
Other western states
Canada
Midwestern states
Southern states
Eastern states
New England states
Total
(Sample)
Total
54%
20
9
6
5
4
2
_J_
100%
(401)
Size of Home City:
Less than 2,500
2,500 - 24,999
25,000 - 49,999
50,000 - 99,999
100,000 - 299,999
300,000 - 499,999
500,000 or more
Total
(Sample)
9%
23
12
15
13
6
22
100%
(401)
Note: # = Less than 0.5%
-------
Q2) Purpose of Trip
Purpose:
Recreation
Other only
Both recreation and other
Total
Q3) Type of Transportation
Mode:
Auto
Bus (commercial)
Airplane
Other
Auto & Airplane
Auto & Bus
Total
Q5,14) Length of Vacation in Oregon
Number
Percent
Time:
3 days or less
4-6 days
1 week - 1 week, 6 days
2 weeks - 2 weeks, 6 days
3 weeks - 3 weeks, 6 days
4 weeks or longer
Undecided
Totals
Samples
341
0
60
401
Number
362
9
5
8
16
1
401
i
(Q5)
Number of Days
Originally
39%
21
22
8
2
4
4
100%
(401)
85.0
0
15.0
100.0
Percent
90.3
2.2
1.2
2.0
4.0
0.2
100.0
(Q14)
Planning to Spend
Currently
39%
24
23
6
2
4
2
100%
(401)
-------
Q8) Planned Use of Time in Specific Outdoor Activities
Activity
Campi ng
Boating, Water skiing
Golfing
Ocean beach activities
Hiking, walking
Swimming
Pleasure driving,
sightseeing
Salmon fishing
Trout fishing
Other
A Lot
# %
44 11.0
17 4.2
10 2.5
54 13.4
36 9.0
40 10.0
206 51.3
21 5.2
8 2.0
37 9.2
Quite
A Bit
# %
17 4.2
6 1.5
6 1.5
45 11.2
48 12.0
45 11.2
91 22.6
8 2.0
16 4.0
18 4.5
A Little None Don't Know Total
# %
26 6.5
29 7.2
26 6.5
97 34.1
200 49.9
99 24.7
62 15.4
21 5.2
31 7.7
29 7.2
# %
303 75.6
341 85.0
348 86.8
202 50.3
175 43.6
210 52.4
41 10.2
341 85.0
336 83.8
312 77.8
# %
11 2.7
8 2.0
11 2.7
3 0.7
2 0.5
7 1.7
1 0.2
10 2.5
10 2.5
5 1.2
401
401
401
401
401
401
401
401
401
401
Q9) Amount of Money Planning to Spend in Oregon
Dollars:
Under $100
$100 - $199
$200 - $299
$300 - $399
$400 - $499
$500 - $599
$600 - $699
$700 - $799
$800 - $999
$1,000 - $1,499
$1,500 or more
Undecided
Total
(Sample)
Average (Median)
Amount Planning to
Spend
Total
26%
20
10
4
3
2
#
2
2
25
1
5
100%
(401)
$206.40
Note: # = Less than 0.5%
-------
Q15) Present (in trip) Planned Use of Time In Specific Outdoor Activities
(read down)
A
Lot
Quite
A Bit
Activity
Camping
Boating, Water
skiing
Golfing
Ocean beach
activities
Hiking, walking
Swimming
Pleasure driving
sightseeing
Salmon Fishing
Trout Fishing
Other
#
36
12
11
44
34
34
201
17
4
33
Q16) Have Spending PI
Response
Yes - Will spend
Yes - Will spend
No change
Don't know
more
less
%
9.0
3.0
2.7
11.0
8.5
8.5
50.1
4.2
1.0
8.2
ans for
#
15
9
6
36
39
38
85
8
13
15
%
3.7
2.2
1.5
9.0
9.7
9.5
21.2
2.0
3.2
3.7
Oregon
A
Little
# %
28 7.0
23 5.7
22 5.5
96 24.0
136 33.8
94 23.4
55 13.7
21 5.2
30 7.5
22 5.5
Been Changed?
Number
97
41
237
26
None
Don
't
Total
Know Responses
#
311
344
347
221
188
227
59
344
343
325
77
85
86
55
46
56
14
85
85
81
(read
%
.6
.8
.5
.1
.8
.6
.7
.6
.5
.0
down)
#
11
13
15
4
5
8
6
11
11
6
%
2.7
3.2
3.7
1.0
1.2
2.0
1.5
3.0
2.7
1.5
401
401
401
401
401
401
401
401
401
401
Percent
24.
10.
59.
6.
2
2
1
5
Total 401
Q17) Have Vacation Plans for Oregon Changed Since Arriving in the State?
(read down)
Response
Changed
No change
Don't know or no answer
Number
Percent
14.7
84.8
.5
Total 401
-------
Q18) Reasons for Changing Vacation Plans, (read down)
Yes No
Reasons # % # %
Rain 15 3.2 44 9.3
Heat 0 0.0 59 12.5
Crowded facilities 11 2.3 48 10.2
Air pollution 6 1.3 53 11.2
Lack of money 2 .4 57 12.1
Auto breakdown or accident 6 1.3 52 11.0
Sickness in family 5 1.1 54 11.5
Other 30 6.4 29 6.2
Don't Know
No Answer
#
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Total
59
59
59
59
59
59
59
59
Q21) Impact of Air Pollution on Spending Plans, (read down)
Response
Spending increased
Spending decreased
No difference
Don't know or no answer
Number
0
5
1
0
Total
Percent
0
83.3
16.7
0
Q22) Magnitude of Impact of Air Pollution on Spending Plans in Oregon.
(read down)
Dollar Impact
$501
$301
$201
$101
$ 51
$ 25
$ 11
$ 1
or more
- $500
- $300
- $200
- $100
- $ 50
- $ 25
- $ 10
Number
0
0
0
2
1
2
0
0
Percent
0
0
0
40.0
20.0
40.0
0
0
Total
Q23) Exclusive of Air Pollution Have the Reasons for Changing Vacation Plans
in Oregon Increased or Decreased Spending Plans in the State.(read down)
Direction of Change
Increased
Decreased
'No difference
Don't know, no answer
Number
22
13
21
3
Percent
36.8
22.8
33.3
7.0
Total 59
-------
Q24) Exclusive of Air Pollution, the Dollar Magnitude of Vacation Plan
Change Impact.
Dollar Impact
Don't know
$501 or more
$301 - $500
$201 - $300
$101 - $200
$ 51 - $100
$ 26 - $ 50
$ 11 - $ 25
$ 1 - $ 10
(read down)
Number
2
2
1
5
2
12
7
3
1
Increase
2
2
1
5
1
5
3
3
0
Decrease
0
0
0
0
1
7
4
0
1
Percent
5.7
5.7
2.9
14.3
5.7
34.3
20.0
8.6
2.9
Total
35
22
13
Q25) How Serious an Air Pollution Problem does Oregon Have.
Seriousness Number
Very
Qui te
Not too
No problem
Don't know
37
39
181
136
8
••••••MHBM
Total 401
(read down)
Percent
9.2
9.7
45.1
33.9
2.0
Q26) Seriousness of Air Pollution Problem in Home Area,
Seriousness Number
Very
Quite
Not too
No problem
Don't know
149
74
112
64
2
(read down)
Percent
37.1
18.4
27.9
15.9
.7
Total 401
Q27) Regarding Air Pollution in Oregon vis-a-vis Home Area.
Relative Pollution Number
More in Oregon
Less in Oregon
Same as home area
Don't know
Qualified "less in Oregon"
71
245
63
6
16
Total 401
(read down)
Percent
17.7
61.1
15.7
1.5
4.0
-------
Q28) Number of Persons in Party (read down)
Number in Party Number
1
2
3
4
5
6
7
8
9
21
159
62
81
43
23
6
4
2
Total 401
Q29) Educational Characteristics of Respondents, (read down)
Education Number
Completed college
Partial college
High school
Grade school or no school
147
99
141
14
Total 401
Q31) Age Characteristics of respondents, (read down)
Age Number
20 - 29
30 - 39
40 - 49
50 - 59
60 & above
56
85
100
90
70
Total 401
Percent,
5.0
40.0
15.0
20.0
11.0
6.0
1.0
1.0
.5
Percent
3.5
35.2
24.0
36.7
Percent
14.0
21.2
24.9
22.4
17.5
Q32) Sex Characteristics of Respondents, (read down)
Sex Number
Male
Female
243
158
Percent
60.6
39.4
Total 401
-------
Q33) Length of Interview, (read down)
Time Number Percent
Less than 30 minutes 304 75.8
30 minutes - 1 hour 82 20.4
1 hour - lh hours 15 3.7
More than 1^ hours 0 0.0
Total 401
Q35) Rating on Cooperation, (read down)
Attitude Number Percent
Positive 361 90.0
Neutral 33 8.2
Negative 7_ 1.7
Total 401
-------
B. Open Ended and Multiple Response Questions
Q4,10,ll) Areas in Oregon Included in Vacation Plans. (Read down)
(Q10)
(Q4)
Original
Places
Planning
Area(s): to Visit
Valley cities 41%
Valley recreational
areas 8
Costal recreational
areas 32
Coastal cities 25
Portland 23
Crater Lake 13
Central cities 10
Central recreational
areas . , 4
Columbia River 5
Interior cities 3
Interior recreational
areas • #
No plans 9
Totals I/ .... 173%
(Samples) .... (401)
Places
Already
Visited
42%
19
27
22
11
12
7
8
2
1
10
170%
(401)
(Qll)
Current
Places
Planning
to Visit
22%
15
14
15
14
8
3
3
2
1
35
137%
(401)
II Results total more than 100%, due to vacation planning which included
~ more than one area.
Note: # - Less than 0.5%.
-------
Q6,7,12,13) Use of Oregon Highways. (Read down)
(06) (Q7) (Q12) (Q13)
Originally Planning Currently Planning
to Use: to Use:
to to
to Leave to Leave
Highways(s): Destination Oregon Destination Oregon
1-5 66% 51% 49% 50%
101 43 27 29 25
97 19 14 18 13
30 or SON 13 7 7 7
20 7 3 5 3
26 6 # 2 #
99 5 5 3 4
22 3 2 2 2
58 2 2 2 2
395 2 1 1 1
126 2 # 6 1
All others; airlines 19 10 14 13
No plans 2 4 9 3
Totals I/ 189% 126% 147% 124%
(Samples) (401) (401) (401) (401)
If Results total more than 100%, due to use of more than one highway.
Note: # = Less than 0.5%.
-------
Q16a) Reasons for Spending Less Money In Oregon than Originally Planned
(Read down)
Reason(s): Total Spending Less
Cut vacation short; injuries forced
us to leave; illness . 39%
Lodging is less expensive; stayed
with relatives; camped more;
gas is less • 24
Air pollution, smog 7
Rain, didn't get to fish 5
Season closed for clams 2
No vacancies 2
No sales tax 2
Well-marked roads 2
No reason; just saved to have
money left over ^
M i see 11aneous 5
Undecided '
Total I/ 102%
(Sample)
II Results total more than 100%, due to multifile resoonse.
-------
Q16a) Reasons for Spending More Money in Oregon than Originally Planned.
(Read down)
Reason(s): Total Spending More
Unplanned activities; more to do
than we thought 51%
Lodging is expensive; used camper
less than planned; high(er) prices 20
Picked up souvenirs and gifts 18
Car repairs; needed tires 8
Purchased fishing gear 3
Illness; doctor bills and medicine 1
Undecided 2_
Total I/ 103%
(Sample) (97)
Q26a) Reasons for Feeling Air Pollution is NOT a Serious Problem in Home Area.
(Read down)
Reason(s): Total Feeling MOT Serious
Less industry 41%
Open country; beach winds;
open spaces; high altitude 24
Less pollution from automobiles 1?
Fewer people 8
Not as many lumber mills;
lumbering has less pollutants
than other industry 5
I/ Results total more than 100%, due to multiple response.
-------
Q26a) Reasons for Feeling Air Pollution is NOT a Serious Problem in Home Area,
(Read down)(continued)"~
Reason(s): Total Feeling NOT Serious
Ban on field burning; less burning 2
Fewer pulp and paper operations 2
Absence of respiratory ailments 2
Fewer farms , 1
Airports and air traffic not
as bad 1
Miscellaneous; proper preventive
legislation; well-controlled .... 7
Undecided 20
Total I/ 130%
(Sample) (176)
Q26a) Reasons for Feeling Air Pollution is a Serious Problem in Home Area.
(Read down)
Reason(s): Total Feeling Serious
Industrial pollution; chemicals;
oil refinery; iron smelters;
aluminium plants; fertilizer
factories 55%
Traffic; abundance of cars and trucks 53
Respiratory ailments; eye irritation;
breathing difficulties; allergies 28
Population; dense metropolitan areas . 14
Airports and planes 9
Pulp and paper mills 6
Lumber mills 4
I/ Results total more than 100%, due to multiple response.
-------
Q26a) Reasons for Feeling Air Pollution is a Serious Problem in Home Area.
(Read down) (continued) ~~
Reason(s): Total Feeling Serious
Farming in general; crop dusting;
smudge pots 2
Field burning specifically 2
Open burning; trash burning 2
Mine dust 1
Miscellaneous; highway construction;
poor visibility; litter 6
Undecided 1_
Total I/ 183%
(Sample) (222)
Q27) Reasons for Feeling Oregon has More of an Air Pollution Problem than
Hometown!(Read down)
Reason(s): Total Feeling More
More industry; canneries 39%
Lumber mills; saw mills;
timber industry 31
Field burning 18
Paper mills 17
Cars and trucks 14
More population; greater
urban areas 11
Open burning; backyard trash
burning; smoke 7
Respiratory ailments; physically
bad; can see and smell pollution 4
\l Results total more than 100%, due to multiple response.
-------
Q27) Reasons for Feeling Oregon has More of an Air Pollution Problem than
Hometown'(Read down) (continued)
Reason(s): Total Feeling More
Less wind; hometown at higher
elevation 4
Miscellaneous; air just isn't
clean; can't see 8
Undecided 3
Total I/ 156%
(Sample) (71)
Q27) Reasons for feeling Oregon has Less of an Air Pollution Problem than
Hometown\(Read down)
Reason(s): Total Feeling Less
Less industrial pollutants % 40%
Smaller population; fewer
people in urban areas
40
Less traffic congestion;
not as many cars ................. 27
Rain; climate cleans air; circulation
of air; sea breeze; close to ocean ........ 17
Forests; trees consume carbon;
high mountains .................. 6
No respiratory ailments .............. 4
Fewer lumber and paper mills ........... 2
Less airports and airplane traffic ........ 2
Not as much field burning; burning
isn't harmful as are chemicals .......... l
Miscellaneous; absence of litter;
clean roads ....................
Undecided ..................... -I4—
Total I/ ........... 155%
(Sample) ....... . . . . (245)
I/ Results total more than 100%, due to multiple response.
-------
Q27) Respondent Comments on Clean Air In Oregon.
(Note:The following comments reflect persons who felt Oregon has less
air pollution than where they live, but also volunteered a critical
comment -- coded 5 on question 27)
V/hen they field burn, you can't do anything about it.
(M, 50-59, Investment Broker, Washington)
Oregon is less industrialized, but there isn't so
much field burning in Washington. (F, 20-29, Wife
of Lumber Supervisor, Washington)
The last few days I think it is a very serious problem
with farmers burning. (M, 30-39, Structural Engineer,
Tennessee)
The country air isn't clean in Oregon. (M, 50-59,
Dentist, Texas)
The field burning is really the only problem. (F,
30-39, Wife of Carpet Layer, California)
On the whole the state Is good, but the field
burning in places is bad. Albany is a poor area
also. Just have these spots of field burning,
which I'm sure is just for part of the year. Of
course, you do have the pulp and paper mills, but
they are trying to convert those to clean them
up. (F, 40-49, Wife of Accountant, California)
What is that burning, anyway? At first I was
feeling so sorry for all those people, but here
they are doing it deliberately. There can't be
any reason that could justify all this smoke.
(F, 40-49, Wife of Teacher, California)
V/hen we came in, they were burning; and the sky
was not quite as clear and pretty as Vancouver,
B.C. where we were just at. (F, 30-39, Wife of
Advertising Agent, California)
I've seen some bad field burning here. (M, 50-59,
Bus Driver, Canada)
There was that awful field burning. (F, 30-39, Wife
of Electrical Contractor, California)
Here it is worse than in the same areas in Washington
because we don't burn. (M, 20-29, Navy Reserve, Wash-
ington)
-------
More Respondent Comments on Clean Air in Oregon.
The only thing we saw was the burning. We have
banned outside burning, like at private residences.
(M, 30-39, Teacher, California)
I had heard before that people travelling through
the Willamette Valley had been disappointed be-
cause of the burning. I hope Oregon will clean
the air. (M, 30-39, Teacher, California)
The fire on the farms is quite bad. (M, 30-39,
Fireman, California)
Where there is field burning, the problem is much
greater. (M, 30-39, Manufacturer of Burglar Alarms,
California)
The only place that we have observed any pollution
is below Portland about 75 miles. There is a heavy
and obnoxious smoke. I wouldn't want to stop in
that area. (M, 60 or over, Writer, California)
------- |