&EPA
United States
Environmental Protection
Agency
Office of
Policy Analysis
Washington DC 20460
February 1985
EPA-230-05-85-006
Costs and Benefits of
Reducing Lead in Gasoline
Final Regulatory Impact
Analysis
-------
COSTS AND BENEFITS OF REDUCING LEAD IN GASOLINE:
FINAL REGULATORY IMPACT ANALYSIS
February 1985
Authors:
Joel Schwartz
Hugh Pitcher
Ronnie Levin
Bart Ostro
Albert L. Nichols
Economic Analysis Division
Office of Policy Analysis.
Office of Policy, Planning and Evaluation
U.S. Environmental Protection Agency
Washington, DC 20460
-------
ACKNOWLEDGEMENTS
Contributors;
Jane Leggett
Albert McGartland
George Sugiyama
The authors thank their colleagues Lane Krahl, Robert
Fegley, and June Taylor Wolcott for their help and expertise.
For technical assistance, often under very tight deadlines,
the authors thank Terry Higgins, Bill Johnson, and Steve Sobotka
of Sobotka and Company; Chris Weaver and Craig Miller of Energy
and Resource Consultants; Asa Janney and Jim Duffey of ICF;
and Ed Fu.
Many individuals have provided excellent secretarial
support, in particular Saundra Womack, Georgia Jackson, Hallie
Baldwin, Sylvia Anderson, Michelle Smith, Leslie Hanney, and
Yvette Carter.
The authors also are grateful to the many colleagues who
provided helpful comments on earlier drafts of this document
and its predecessors.
-------
TABLE OF CONTENTS
PAGE
EXECUTIVE SUMMARY E-l
CHAPTER Ii INTRODUCTION
I.A. Background Information 1-2
I.A.I. Previous Rulemakings 1-3
I.A.2. Continuing Problems 1-4
I.A.3. The August 1984 Proposal 1-8
I.A.4. Information Received After August 1984 Proposal 1-9
I.E. Need for Government Intervention 1-13
I.C. Alternatives to New Regulations 1-14
I.C.I. No Change in Policy 1-14
I.C.2. Public Education 1-15
I.C.3. Stepped-Up Enforcement 1-16
I.D. Market-Oriented Alternatives 1-18
I.D.I. Marketable Permits 1-19
I.D.2. Pollution Charges 1-21
I.E. Alternative Standards 1-23
-------
Table of Contents, page 2
PAGE
CHAPTER II: COSTS OF REDUCING LEAD IN GASOLINE
II.A. Price versus Cost Differences II-2
II.B. The Refinery Model II-6
II.B.I. Introduction to the DOE Model II-6
II.B.2. Overview of Refining Processes II-7
II.C. Base-Case Assumptions and Cost Estimates 11-18
II.C.I. Base-Case Parameter Values 11-19
II.C.I.a. Gasoline Demand 11-19
Il.C.l.b. Assumptions About Refinery Operations 11-24
II.C.2. Base-Case Results 11-34
II.D. Sensitivity Analyses 11-39
II.D.I. Level of Aggregation 11-40
II.D.2. Other Parameters 11-42
II.D.2.a. Assumptions Varied 11-44
II.D.2.b. Results of Sensitivity Analyses 11-48
II.E. Impact of Banking on Costs 11-60
II.E.I. Base-Case Banking Results 11-61
II.E.2. Sensitivity Analyses with Banking 11-68
-------
Table of Contents, page 3
PAGE
CHAPTER III: HUMAN EXPOSURE TO LEAD FROM GASOLINE
III.A. The Relationship between Lead in Gasoline
and Lead in Blood II1-2
III.A.I. Recent Studies III-2
III.A.2. Available Data Sets III-5
III.A.2.a. Gasoline-Use Data III-6
III.A.2.b. The NHANES II III-6
III.A.2.c. The CDC Screening Program for 111-10
Lead Poisoning
III.A.2.d. Chicago, New York, and Louisville, 111-10
Kentucky Data
III.A.3. Statistical Analyses of Exposure 111-10
III.B. The Question of Causality 111-21
III.B.I. Experimental Evidence 111-22
III.B.2. Does Cause Precede Effect? 111-23
III.B.3. Replicability and Consistency 111-24
III.B.4. Does a Dose-Response Relationship Exist? 111-25
III.B.5. Biological Plausibility 111-26
III.B.6. Control of Confounding Factors 111-26
III.B.6.a. External Validation 111-27
III.B.6.b. Seasonality 111-30
III.B.6.C. Other Time Trends 111-31
III.B.6.d. Geographic Sampling Pattern 111-31
III.B.6.e. Subgroup Analysis 111-33
III.C. Impact of Rule on Numbers of Children Above 111-34
Various Blood-Lead Levels
III.C.1. Estimation Procedure 111-35
III.C.2. Incidence Versus Prevalence II1-39
-------
Table of Contents, page 4
PAGE
CHAPTER IV: BENEFITS OF REDUCING CHILDREN'S EXPOSURE TO LEAD
IV.A. Pathophysiological Effects IV-3
IV.A.I. Effects of Lead on Pyrimidine Metabolism IV-7
IV.A.2 Effects on Heme Synthesis and Related Hematological IV-8
Processes
IV.A.2.a. Mitochondrial Effects IV-8
rv.A.2.b. Heme Synthesis Effects IV-8
IV.A.2.C. Impact of Lead on Red Blood Cell Abnormalities IV-10
IV.A.2.c.l. Effects of Lead Exposure on Blood Cell IV-11
Volume and Hemoglobin Content
IV.A.2.c.2. The Relationship Between Blood Lead IV-18
and FEP
IV.A.2.c.3. The Relationship Betwen FEP Levels and TV-21
Anemia
IV.A.3. Lead's Interference with Vitamin D Metabolism IV-24
and Associated Physiological Processes
IV.B. Neurotoxic Effects of Lead Exposure IV-28
IV.B.l. Neurotoxicity at Elevated Blood-Lead Levels IV-28
IV.B.2. Neurotoxicity at Lower Blood-Lead Levels IV-30
IV.B.2.a. Cognitive Effects of Moderate Blood-Lead Levels IV-33
IV.B.3. The Magnitude Impact of Lead on IO IV-41
IV.C. Fetal Effects IV-42
IV.D. Monetized Estimates of Children's Health Benefits IV-47
IV.n.l. Reduced Medical Costs IV-47
IV.D.2. Reduced Costs of Compensatory Education IV-52
IV.D.3. Summary of Estimated Benefits IV-53
-------
Table of Contents, page 5
PAGE
CHAPTER V: HEALTH BENEFITS OF REDUCING LEAD:
ADULT ILLNESSES RELATED TO BLOOD PRESSURE
V.A. The Relationship Between Blood Lead and Blood Pressure V-2
V.A.I. Earlier Studies V-2
V.A.2. Analysis of NHANES II Data V-5
V.A.2.a. Blood Pressure Measurements V-7
V.A.2.b. Initial Analysis V-7
V.A.3. Tests of Robustness V-9
V.A.3.a. Nutritional and Biochemical Variables V-9
V.A.S.b. Interaction Terms V-15
V.A.3.C. Marginally Insignificant Variables V-18
V.A.S.d. Nonnutrition Variables V-19
V.A.3.e. Other Age Groups V-23
V.A.4. Summary of Blood Lead - Blood Pressure Results V-23
V.B. Benefits of Reduced Cardiovascular Disease V-26
V.B.I. Reductions in Hypertension and Related Morbidity
and Mortality V-26
V.B.1.a. Hypertens ion V-27
V.B.l.b. Myocardial Infarctions, Strokes, and Deaths V-29
V.B.2. Monetized Benefit Estimates V-35
V.B.2.a. Hypertension V-35
V.B.2.b. Myocardial Infarctions V-37
V.B.2.C. Strokes V-40
V.B.2.d. Mortality V-42
V.B.3. Summary of Blood Pressure Benefits V-44
-------
Table of Contents, page 6
PAGE
CHAPTER VI: BENEFITS OF REDUCING POLLUTANTS OTHER THAN LEAD
VI .A. Emissions Associated with Misfueling VI-4
VI.B. Health and Welfare Effects Associated with Ozone VI-9
VI.B.I. Effects of a 1 Percent Reduction in Ozone Vl-11
VI.B.I.a. Health Effects of Reducing Ozone VI-11
VI.B.l.b. Ozone Agricultural Effects VI-22
VI.B.I.e. Ozone Effects on Nonagricultural Vegetation VI-26
VI.B.l.d. Ozone Materials Damage VI-28
VI.B.I.e. Summary of Benefits of a 1 Percent VI-29
Change in Ozone
VLB.2. Linking NOX and HC Reductions to Ozone Effects VI-31
VLB.2.a. The Process of Ozone Formation VI-31
VI.B.2.b. Quantitative Estimates of Impacts of VI-33
HC and NOX on Ozone
VI.B.2.C. Ozone-Related Effects Per Ton of HC and VI-39
NOX Controlled
VI.C. Health and Welfare Effects Not Related to Ozone VI-43
VI.C.1. Hydrocarbons VI-43
VI.C.I.a. Impact on Sulfates VI-43
Vl.C.l.b. Impact on Benzene and Other Aromatics VI-44
VI.C.2. Nitrogen Oxides VI-47
VI.C.2.a. Visibility Benefits from Reduced NOx VI-48
VI.C.2.b. Health Benefits of Reducing NOx VI-50
VI.C.2.C. NOX Effects on Vegetation VI-52
VI.C.2.d. NOX Effects on Materials VI-53
VI.C.2.e. Acid Deposition Benefits VI-54
-------
Table of Contents, page 7
PAGE
VI.C.3. Carbon Monoxide VI-55
Vl.C.S.a. Health Effects of CO VI-56
VI.C.3.b. Change in Numbers of People Above
2.9 Percent COHb VI-59
VI.C.4. Ethylene Dibromide Emissions VI-64
VI.D. Monetized Benefit Estimates VI-66
VI.D.I. Value of Quantified Health and Welfare Benefits VI-66
VI.D.2. Implicit Value Based on Cost of Control
Equipment VI-68
VI.D.3. Summary of Benefits of Controlling Conventional
Pollutants Other than Lead VI-71
-------
Table of Contents, page 8
PAGE
CHAPTER VII: VEHICLE MAINTENANCE, FUEL ECONOMY,
AND ENGINE DURABILITY BENEFITS
VILA. Maintenance Benefits VII-3
VII.A.I. Exhaust Systems VII-3
VII.A.2. Reduced Fouling and Corrosion of Spark Plugs VII-9
VII.A.3. Extended Oil Change Intervals VII-11
VII.A.4. Summary of Maintenance Benefits VII-14
VII.B. Improved Fuel Economy VII-18
VII.B.I. Energy Content VII-18
VII.B.2. Reduced Fouling of Oxygen Sensors VII-19
VII.B.3. Summary of Fuel Economy Benefits VII-20
VII.C. Engine Durability VII-20
VII.C.I. Valve-Seat Recession VII-22
VII.C.I.a. Laboratory and Track Studies of Valve-
Seat Recession VII-22
Vll.C.l.b. Fleet Studies of Valve-Seat Recession VII-27
VII.C.I.e. Other Types of Engines VII-35
Vll.C.l.d. Alternatives to Lead to Avoid Potential
Valve Recession VI1-36
VII.C.2. Negative Effects of Lead on Engine
Durability VII-38
VII.C.3. Summary of Engine Durability Effects VII-43
-------
Table of Contents, page 9
PAGE
CHAPTER VIIIi COST-BENEFIT ANALYSIS OF ALTERNATIVE
PHASEDOWN RULES
VIII.A. Summary of Cost and Benefit Estimates VIII-2
VIII.B. Comparisons of Alternative Lead Levels VIII-7
VIII.C. Impact of Banking on Costs and Benefits VIII-27
of Final Rule
VIII.D. Summary VIII-33
REFERENCES R-l
APPENDIX A: Refinery Processes A-l
APPENDIX B: Fleet Model B-l
APPENDIX C: Supplementary Regressions of Blood Lead on
Gasoline Lead C-l
-------
LIST Of TABLES
CHAPTER I:
TABLE 1-1
Alternative Phasedown Schedules
PAGE
1-24
CHAPTER II:
TABLE II-l
TABLE I1-2
TABLE I1-3
TABLE I1-4
TABLE I1-5
TABLE 11-6
TABLE I1-7
TABLE 11-8
TABLE 11-9
TABLE 11-10
TABLE 11-11
TABLE 11-12
TABLE 11-13
TABLE 11-14
Functional Characterization of Refinery
Processes 11-13
Sample Process Data Table from Refinery Model:
Yields and Operating Cost Coefficients for 11-15
Crude Distillation Unit
Year-by-Year Estimates of Gasoline Demand 11-23
Estimated U.S. Refinery Processing Unit
Capacities for 1988 11-26
Prices of Crude Oil and Petroleum Products
in 1983 and 1985 11-31
Cost of the 0.10 gplg Standard with New Oil
Prices: New Model Run versus Repricing,
Assuming No Misfueling 11-33
Base-Case Results for 1985 - 1988, with
Partial Misfueling 11-36
Year-by-Year Estimates of Costs of Meeting
Alternative Rules, with Partial Misfueling 11-38
Costs of Meeting the 0.10 gplg Standard:
Comparison of National and Regional Results 11-43
for 1986
Parameters Examined in Cost Sensitivity Analyses 11-45
Effects of Varying Individual Parameters/
Assumptions: PADDs I-IV/VI 11-49
Effects of Varying Multiple Parameters/
Assumptions: PADDs I-IV/VI 11-52-53
Sensitivity Analyses for 1986: PADD V 11-58
Sensitivity Analyses for 1985: PADDs I-IV/VI 11-59
-------
List of Tables, page 2
TABLE I1-15
TABLE 11-16
TABLE 11-17
TABLE 11-18
TABLE 11-19
Alternative Phasedown Patterns with Banking
Refining Costs Under Alternative Phasedown
Patterns, with Partial Misfueling
Impact of Banking on Marginal Costs of Octane
Sensitivity Analyses for 1986 with Banking:
Alternative 1, PADDs I-IV/Vl
Sensitivity Analyses for 1986 with Banking:
Alternative 2, PADDs I-IV/VI
PAGE
11-63
11-65
11-67
11-69
11-70
CHAPTER III:
TABLE III-l
TABLE II1-2
TABLE III-3a
TABLE III-3b
TABLE II1-4
TABLE II1-5
TABLE II1-6
TABLE II1-7
NHANES II: Regression Results for Whites 111-15
Logistic Regression on Probability of
Blood Lead > 30 ug/dl for Children I11-16
6 months to 7 years
Regression of CDC Screening Data:
Percent of Children with Lead Toxicity on 111-18
Gasoline Lead
Regression of CDC Screening Data: Change
in Lead Toxicity on Change in Gasoline
Lead II I-18
Black Children in Chicago: Regression of
Average Blood-Lead Levels on Gasoline II1-20
Lead Levels
Lead in the Diet 111-28
Chicago Data: Probability of Blood Lead
> 30 ug/dl With and Without Lead Paint Hazard 111-30
in the Hone
Estimated Reductions in the Numbers of
Children Over Various Blood-Lead Levels, 111-38
Assuming No Misfueling
-------
List of Tables, page 3
CHAPTER TV:
TABLE IV-1
TABLE IV-2
TABLE IV-3
TABLE IV-4
TABLE IV-5
TABLE IV-6
TABLE IV-7
PAGE
Blood Lead Levels of Children in the United
States, 1976-80 IV-4
Variables Considered in the Regressions
of FEP, MCV, MCH, and Anemia IV-12
Computation of Joint P-Value from Epidemiological
Studies of Cognitive Effects from Low Level IV-39
Lead Exposure in Children
Year-by-Year Estimates of Gain in Person -
10 Points Under Alternative Rules, Assuming IV-43
No Misfueling
Estimated Decrease in Number of Fetuses
Exposed to 25 ug/dl of Blood Lead IV-47
Percent of Children Requiring Chelation IV-50
Therapy
Year-by-Year Monetized Benefits of Reducing IV-55
Children's Exposure to Lead Under Alternative
Rules, Assuming No Misfueling
CHAPTER V:
TABLE V-l
TABLE V-2
TABLE V-3
TABLE V-4
TABLE V-5
TABLE V-6
TABLE V-7
Variables Included in the Stepwise V-ll
Regression Analyses
Regression of Diastolic and Systolic
Blood Pressures in White Males Aged V-13
40 to 59
Weighted Logistic Regression on Probability
of Diastolic Blood Pressure Greater Than V-14
or Equal to 90 mm Hg in Men Aged 40 to 59
Nonnutrition Variables Tested in the Stepwise V-20
Regression
Regression of Diastolic and Systolic Blood
Pressures in White Males Aged 49 to 50 V-24
Weighted Logistic Regression Probability
of Blood Pressure Greater Than or Equal
to 90 mm Hg in Men Aged 40 to 59 V-25
Reductions in Cases of Hypertension in Males
Aged 40 to 59, Assuming No Misfueling V-28
-------
List of Tables, page 4
TABLE V-8 Reductions in Numbers of Cases of
Cardiovascular Disease and Deaths in Males
Aged 40 to 59, Assuming No Misfueling
TABLE V-9 Benefits of Reducing Strokes
TABLE V-10 Year-by-Year Estimates of Blood Pressure
Benefits, Assuming No Misfueling
PAGE
V-34
V-41
V-45
CHAPTER VI:
TABLE VI-1
TABLE VI-2
TABLE VI-3
TABLE VI-4
TABLE VI-5
TABLE VI-6
TABLE VI-7
TABLE VI-8
TABLE VI-9
TABLE VI-10
TABLE VI-11
TABLE VI-12
TABLE VI-13
Increase in Emissions Due to Misfueling VI-6
Misfueling Rates in 1983 VI-8
Year-by-Year Estmates of Reductions in
Emissions, Assuming No Misfueling VI-10
Regression Results for Portney and Mullahy
Study on Respiratory Symptoms Related to Ozone VI-15
Regression Coefficients from Hasselblad
and Svendsgaard Study on Respiratory and
Non-respiratory Symptoms Related to Ozone VI-19
Estimated Health Effects of a 1 Percent
Reduction in Ozone VI-23
Annual Agricultural Benefits of a 1 Percent
Ozone Reduction VI-25
Summary of Estimated Effects of a 1 Percent
Reduction in Ozone VI-30
Estimated Ozone Reductions from 1 Percent
Reduction in Rural and Metropolitan HC VI-34
and NOX
Estimated Ozone Reductions due to Eliminating
Misfueling VI-41
Quantified Ozone-Related Effects due to
Eliminating Misfueling in 1986 VI-42
Reductions in Benzene Emmissions in 1986,
Assuming No Misfueling VI-46
Monetized Benefits Per Ton of Hydrocarbons
Controlled VI-69
-------
List of Tables, page 5
TABLE VI-14
Year-by-Year Estimates of Benefits of Reduced
Emissions of Conventional Pollutants, Assuming
No Misfueling
PAGE
VI-74
CHAPTER VII:
TABLE VII-1
TABLE VI1-2
TABLE VI1-3
TABLE VI1-4
TABLE VI1-5
TABLE VI1-6
TABLE VI1-7
Summary of On-Road Studies of Spark Plug VI1-5
and Exhaust System Wear
Estimated Maintenance Benefits Per Mile of
Reducing Lead in Gasoline VI1-15
Year-by Year Estimates of Maintenance
Benefits, Assuming No Misfueling VII-17
Year-by-Year Estimates of Fuel Economy
Benefits, Assuming No Misfueling VII-21
Summary of Findings of Track and Dynamometer
Studies of Lead Levels and Valve Recession VII-24
Summary of Findings of Consumer and Fleet
Studies of Lead Levels and Valve-Seat Recession VII-29
Vehicle and Engine Types in U.S. Army
Unleaded Gasoline Test VI1-33
CHAPTER VIII:
TABLE VIII-1 Costs and Monetized Benefits of 0.10 gplg
in 1986, Assuming No Misfueling: Comparison VIII-3
of Current and Draft RIA Estimates
TABLE VI11-2 Non-monetary Measures of Health and
Environmental Benefits of 0.10 gplg in 1986, VIII-6
Assuminig No Misfueling: Comparison of
Current and Draft RIA Estimates
TABLE VIII-3a Costs and Monetized Benefits of Alternative
Lead Levels in 1986, Assuming No Misfueling VIII-8
TABLE VIII-3b Costs and Monetized Benefits of Alternative
Lead Levels in 1986, Assuming Full Misfueling VIII-9
TABLE VIII-3c Costs and Monetized Benefits of Alternative
Lead Levels in 1986, with Partial Misfueling VIII-15
-------
List of Tables, page 6
TABLE VIII-4a
TABLE VIII-45
TABLE VIII-4C
TABLE VII1-5
TABLE VI11-6
TABLE VIII-7a
TABLE VIII-7b
TABLE VIII-Vc
TABLE VI11-8
TABLE VI11-9
TABLE VIII-10
TABLE VIII-11
Costs and Monetized Benefits of Alternative
Lead Levels in 1987, Assuming No Misfueling
Costs and Monetized Benefits of Alternative
Lead Levels in 1987, with Full Misfueling
Costs and Monetized Benefits of Alternative
Lead Levels in 1987, with Partial Misfueling
Costs and Monetized Benefits of Alternative
Lead Levels in 1985, Assuming Full Misfueling
Present Values of Costs and Monetized
Benefits: Comparison of Proposed,
Alternative, and Final Schedules for
1985-1987
Year-by-Year Costs and Monetized Benefits
of Final Rule, Assuming No Misfueling
Year-by-Year Costs and Monetized Benefits
of Final Rule, Assuming Full Misfueling
Year-by-Year Costs and Monetized Benefits
of Final Rule, with Partial Misfueling
Present Values of Costs and Benefits of
Final Rule, 1985-1992
Alternative Phasedown Patterns with Banking
Costs and Monetized Benefits of Alternative
Phasedown Patterns, with Partial Misfueling
Present Values of Costs and Benefits of
Alternative Phasedown Patterns, 1985-87,
with Partial Misfueling
PAGE
VIII-16
VIII-17
VI11-18
VIII-20
VIII-21
VII1-23
VIII-24
VIII-25
VIII-26
VIII-28
VIII-30
VIII-32
-------
LIST of FIGURES
CHAPTER I:
FIGURE 1-1
Comparison of 1982 and Current
Projections of Lead Use in Gasoline
PAGE
1-5
CHAPTER II:
FIGURE II-l
FIGURE II-2
FIGURE I1-3
FIGURE I1-4
CHAPTER III:
FIGURE III-l
FIGURE II1-2
FIGURE II1-3
FIGURE II1-4
FIGURE II1-5
FIGURE II1-6
Schematic Diagram of a Simple Oil Refinery
(Topping Plant)
Schematic Diagram of a Hydroskimming Refinery
Schematic Diagram of a Fuels Refinery
Schematic Diagram of a High Conversion Refinery
Relationship Between Gasoline Lead and Blood
Lead in New York City
Lead Used in Gasoline Production and Average
NHANES II Blood Lead Levels
NHANES II Data: Blood Lead versus Gasoline Lead
CDC Data: Gasoline Lead versus Percent of
Children with Lead Toxicity
Chicago Data: Gasoline Lead versus Blood Lead
New York City Data: Gasoline Lead versus
Blood Lead
II-9
11-10
11-11
11-12
II1-3
III-8
III-9
III-ll
II1-12
II1-13
CHAPTER IV:
FIGURE IV-1
FIGURE IV-2
Multi-Organ Impact of Lead's Effects IV-2
on the Heme Pool
The Relationship Between MCV and Lead After
Adjusting for All Other Significant Variables IV-14
-------
list of Figures, page 2
FIGURE IV-3
FIGURE IV-4
FIGURE IV-5
FIGURE IV-6
FIGURE IV-7
FIGURE IV-8
The Relationship Between MCH and Blood Lead
After Adjusting for All Other Significant
Variables
Prediction of Percent Children with MCV < 74 fl
as a Function of Blood Lead, After Controlling
for All Other Significant Variables
The Relationship Between FEP and Blood Lead
After Controlling for All Other Significant
Variables
Prediction of Percent of Children with Anemia
as a Function of FEP Levels at Normal
Transferrin Saturation Levels
Effects of Lead on 10
Flow Diagram for Children with Blood Lead
Levels above 25 ug/dl
PAGE
IV-15
IV-17
IV-20
IV-23
IV-37
IV-49
CHAPTER V:
FIGURE V-l
FIGURE V-2
FIGURE V-3
Adjusted Systolic Blood Pressure versus
Blood Lead
Adjusted Diastolic Blood Pressure versus
Blood Lead
Adjusted Rates of Death and Heart Attacks
versus Blood Pressure: Framingham Data
V-16
V-17
V-30
CHAPTER VIII:
FIGURE VIII-1 Impact of Lead Levels in Misfueling Under VIII-11
Three Assumptions
FIGURE VIII-2 Net Benefits (Including Blood-Pressure- VIII-12
Related Effects) as Functions of Lead Level
and Misfueling
FIGURE VIII-3 Net Benefits (Excluding Blood-Pressure- VIII-13
Related Effects) as Functions of Lead and
Misfueling
-------
EXECUTIVE SUMMARY
The Environmental Protection Agency (EPA) is promulgating a
rule to reduce the amount of lead in gasoline from its current
limit of 1.10 grams per leaded gallon (gplg) to 0.50 gplg on July
1, 1985 and to 0.10 gplg effective January lf 1986. EPA's primary
objective in promulgating this rule is to minimize the adverse
health and environmental effects of lead in gasoline. To
increase flexibility in meeting the phasedown schedule, EPA has
proposed to allow refineries that reduce lead ahead of schedule
in 1985 to "bank" those extra lead rights for use in 1986 or
1987. The Agency also is considering the possibility of a
complete ban on leaded gasoline, to take effect as early as 1988.
This Regulatory Impact Analysis addresses only the final
phasedown rule; a separate document summarizes the costs and
benefits of a possible ban.
Basis for Action
Section 211(c)(l) of the Clean Air Act gives EPA's
Administrator broad authority to "control or prohibit the
manufacture ... or sale of any fuel additive" if its emission
products (1) cause or contribute to "air pollution which may be
reasonably anticipated to endanger the public health or welfare,"
or (2) "will impair to a significant degree the performance of
any emission control device or system ... in general use."
Reductions in the lead content of gasoline are justified under
both of these criteria.
Lead in gasoline has been shown to increase blood lead
levels, which in turn have been linked to a variety of serious
health effects, particularly in small children. Recent studies
linking lead to blood pressure in adult males also are a source
of concern about the health effects of lead in gasoline; because
these studies have just been published, however, EPA will not
rely upon that evidence until there has been a greater
opportunity for scientific review and public comment.
Lead in gasoline also impairs the effectiveness of
pollution-control catalysts. A 1983 EPA survey of vehicles in
use showed that about 15.5 percent of the vehicles that should
use unleaded gasoline to protect the effectiveness of their
pollution-control catalysts are misfueled with leaded gasoline,
resulting in significant excess emissions of hydrocarbons (HC),
nitrogen oxides (NOX), and carbon monoxide (CO). In addition to
these health and environmental effects, reducing lead in gasoline
will reduce vehicle maintenance costs associated with the
corrosive effects of lead on engines and exhaust systems, and
will improve fuel economy.
-------
E-2
EPA has considered a variety of alternatives for reducing
the health and environmental problems caused by lead in gasoline,
and has concluded that the most effective approach is to reduce
the amount of lead in gasoline as quickly as possible. EPA is
confident that the phasedown schedule mandated by the rule can be
met by the refining industry with existing equipment. The Agency
is not promulgating a complete ban on lead in gasoline at this
time because of continuing concern about the possible effects of
such a ban on certain engines that may rely on lead for
protection against valve-seat recession. EPA will consider the
valve-seat issue, along with the recent evidence on lead and
blood pressure, in reaching a final decision on a ban.
Costs of Reducing Lead in Gasoline
Since the 1920s, petroleum refiners have added lead to
gasoline as a relatively inexpensive way of boosting octane. To
meet octane requirements with less lead, refiners have several
options, including additional processing in reforming or
isomerization units and the use of alternative additives, such
as MMT or alcohols. At the margin, however, each of these
alternatives is more expensive than lead for producing octane.
Higher refining costs constitute virtually all of the rule's
social costs.
We estimated the additional refining costs using a model of
the petroleum refining industry developed for the Department of
Energy (DOE). This model employs a linear programming framework
to represent U.S. refining operations, and can find the minimum-
cost method for producing a particular product slate subject to
various constraints (including the amount of lead allowed in
gasoline). The model has been used by both EPA and DOE in
previous rulemakings, and several verification checks have
indicated that it performs well.
To estimate the costs of alternative rules, we first ran the
model specifying the existing lead limit of 1.10 gplg. We then
ran the model with a tighter constraint on lead. In both types
of runs, the model also was constrained to meet projected demands
for gasoline and other refined products. The difference between
the costs at the two lead limits is the estimated cost of the
rule. This procedure limits the possibility of underestimating
costs if the model assumes more flexibility than is in fact
possible; any overoptimization affects both estimates, and thus
has little impact on the difference between the two. In
addition, we placed many constraints on the model to reflect
real-world limitations in the ability of the refining industry to
fully optimize production.
-------
E-3
Our base case results suggest that the final rule will cost
less than $100 million for the second half of 1985, when the 0.50
gplg limit will apply. For later years, when the 0.10 gplg limit
will apply, the estimated costs range from $608 million in 1986
to $441 million in 1992. (The estimated costs fall over time
because of projected declines in the demand for leaded gasoline,
even in the absence of this new rule.)
We ran extensive sensitivity analyses to probe the limits of
feasibility. Those analyses focused on 1986, because that is the
first year in which the 0.10 gplg will apply, and refineries will
not be able to undertake substantial new construction by then.
Those sensitivity analyses show that the rule remains feasible
(i.e., product demands can be met with existing refining capacity)
under most conditions. Only when extremely unlikely combinations
of multiple adverse conditions are assumed does feasibility
appear to be in doubt, and then only for the peak-demand summer
months. Additional sensitivity analyses show that even in those
worst-worst cases, the 0.50 gplg limit in 1985 does not approach
infeasibility.
We also examined the potential impact on costs of the
banking rule that EPA recently proposed and which it may
promulgate shortly. Under the banking rule, refineries that
reduced their lead use below applicable limits in 1985 (1.10 gplg
prior to July 1 and 0.50 gplg for the second half of the year)
could bank those extra reductions and use them in 1986 or 1987.
Banking would allow refineries to follow their own least-cost
schedules of lead reduction, so long as their total usage over
the three years did not exceed the amount allowed by this rule.
Because banked rights would be freely transferable among
refineries, they also would increase individual refineries'
flexibility by allowing those refineries with relatively high
costs to buy rights from refineries with lower costs. Our
estimates suggest that banking would reduce the present value of
the rule's cost by about $200 million over the 1985-to-1987
period. Moreover, it appears that banking would alleviate the
potential infeasibility found in the most extreme sensitivity
analyses.
Benefits of the Rule
We estimated benefits in four major categories:
(1) children's health and cognitive effects associated
with lead; (2) blood-pressure-related effects in adult males due
to lead exposure; (3) damages caused by excess emissions of HC,
NOX, and CO from misfueled vehicles; and (4) impacts on mainten-
ance and fuel economy. In each category, our estimates do not
cover all of the likely benefits because of gaps in the data or
difficulties in monetizing some types of benefits. Nonetheless,
the estimates are substantial, and far exceed the costs.
-------
E-4
Human exposure to lead from gasoline. To predict the lead-
related health effects of the rule, we began by estimating its
impact on lead in individuals' blood. People are exposed to lead
from gasoline through a variety of routes, including direct
inhalation of lead particles when they are emitted from vehicles,
ingestion of lead-contaminated dust or inhalation of such dust
when it is stirred up, and ingestion of food that has been
contaminated with lead. Although it is difficult to estimate the
contributions of these individual pathways, several large data
sets make it possible to estimate the total contribution of lead
in gasoline to concentrations of lead in human blood.
These data sets include records of lead-screening programs
from the Centers for Disease Control (CDC), records from
screening programs in individual cities, and, most importantly,
the Second National Health and Nutrition Examination Survey
(NHANES II), which provides blood lead measurements (and other
important information) on a large representative sample of the
U.S. population surveyed during the late 1970s. By linking these
data to data on gasoline lead use, it is possible to estimate
statistically how blood lead levels respond to changes in
gasoline lead.
Several studies have shown remarkably strong and consistent
relationships between gasoline lead and blood lead. Figure E-l
plots those two measures over time using data from NHANES II.
Note how strong the relationship is; blood lead tracks both the
seasonal variations in gasoline lead (rising during the summer
months, when more gasoline is used) and the long-term downward
trend in gasoline lead. Multiple regression analyses show that
this relationship continues to hold after controlling for other
factors (such as socioeconomic status, nutritional factors, and
exposure to other sources of lead). Such studies suggest that
during the 1970s, gasoline was responsible on average for about
half of the lead in blood (other sources of lead include lead
paint, stationary sources, and lead solder in cans).
Experimental studies, where the isotopes in gasoline lead have
been modified so that its presence in blood can be distinguished
from lead from other sources, have yielded similar conclusions.
Statistical analyses indicate that gasoline lead not only
raises the average level of lead in blood, but also contributes
substantially to the incidence of lead toxicity in children.
Based on an analysis of NHANES II, we predict that the 0.10 gplg
limit will roughly halve the number of children with blood lead
levels above those recognized as harmful. Since 1978, the CDC
has recommended that children with blood lead levels above 30
micrograms per deciliter (ug/dl) receive follow-up testing
and possible treatment. The CDC recently reduced that recom-
mended level to 25 ug/dl. We estimate that in 1986 alone,
the rule will prevent 172,000 children from exceeding 25 ug/dl
blood lead.
-------
FIGURE E-l
110-1
100-
c
to
8
a
70-
•0
w
5d
M
O
0 60
o
o
o
£
3
in
50-
LEAD USED IN GASOLINE PRODUCTION AND
AVERAGE NHANES II BLOOD LEAD LEVELS
(FEB. 1976 - FEB. 1980)
LEAD USED IN
GASOUNE
AVERAGE
BLOOD
LEAD LEVELS
•/
1976
1977
1978
YEAR
1979
1980
16 g
14
13
12
11
10
0
g
H-
n
H
o
M
§
Ul
ID
n
c*
ro
-------
E-6
Children's health and cognitive effects. Elevated blood
lead levels have been linked to a wide range of health effects,
with particular concern focusing on young children. These
effects range from relatively subtle changes in biochemical
measurements at low doses (e.g., 10 ug/dl) to severe retardation
and even death at very high levels (e.g., 100 ug/dl). Lead can
interfere with blood-forming processes, vitamin D metabolism,
kidney functioning, and neurological processes. The negative
impact of lead on cognitive performance (as measured by IQ tests,
performance in school, and other means) is generally accepted at
moderate-to-high blood-lead levels (30 to 40 ug/dl and above).
Several studies also suggest cognitive effects at lower levels.
Changes in electroencephlogram readings have been found at levels
as low as 10 to 15 ug/dl.
For children's health effects, we estimated benefits in two
categories: medical care for children exceeding the new CDC
cutoff and compensatory education for a subset of those children
who may suffer cognitive effects from exposure to lead. These
estimates are conservative in that they do not include many
benefit categories (e.g., lasting health and cognitive damage not
reversed by medical treatment and compensatory education), nor do
they attribute any benefits to reducing lead levels in children
whose blood lead levels would be below 25 ug/dl even in the
absence of the rule.
To estimate reductions in medical care expenses, we relied
on recently published recommendations for testing and treating
children with blood lead levels above 25 ug/dl. Such treatment,
we estimate, costs about $900 per child over 25 ug/dl. (This
average reflects lower costs for most of these children, but much
higher costs for the subset requiring chelation therapy.)
The estimates for compensatory education assumed three years
of part-time compensatory education for roughly 20 percent of the
children above 25 ug/dl; that averages about $2600 per child
above that blood-lead level. Thus we estimated a total of $3500
in monetized benefits for each child brought below 25 ug/dl. Our
estimates of aggregate benefits in this category ranged from
about $600 million in 1986 to roughly $350 million in 1992.
Blood-pressure-related benefits. Lead has long been
associated with elevated blood pressure, but until recently most
of the studies have focused only on hypertension and relatively
high lead levels (typically found only in those occupationally
exposed to lead). Two recent studies, however, have found a
continuous relationship between blood lead and blood pressure
using data from a large representative sample of the U.S.
population (NHANES II, the same data set used to estimate the
relationship between gasoline lead and blood lead). These
studies show a strong relationship that has proved robust in the
face of exhaustive statistical tests involving many possible
confounding factors and alternative specifications of the model.
-------
E-7
These findings, if verified, have importanb implications for
the benefits of the phasedown rule. EPA has not relied upon them
in setting the final phasedown rule, because they are too recent
to allow widespread review and comment. They will be an
important element, however, in the decision on a final ban, so we
present estimates of blood-pressure-related benefits in this RIA
for informational purposes.
To calculate the benefits, we used logistic regression
equations estimated from NHANES II to predict how the rule would
affect the number of hypertensives. These estimates cover only
males aged 40 to 59, because the effect of lead on blood pressure
appears to be strongest for men and because estimates for that
age range are not confounded by a strong covariance between age
and blood lead. We estimate that the rule will reduce the number
of hypertensives in that group by about 1.8 million in 1986. We
valued reductions in hypertension based on estimates of the costs
of medical care, medication, and lost wages; they yielded a value
of $220 per year per case of hypertension avoided.
We also estimated how reductions in blood pressure would
affect the incidences of various cardiovascular diseases, based
on our projections of changes in blood pressure as a result of
the rule and estimates of the relationships between blood
pressure and heart attacks, strokes, and deaths from all causes.
The latter estimates were derived from several large-scale
epidemiological studies, primarily the Framingham study. Because
those studies included very few nonwhites, we restricted our
estimates to white males aged 40 to 59.
We valued reductions in heart attacks and strokes based on
the cost of medical care and lost wages for nonfatal cases (the
fatalities from heart attacks and strokes were included in the
estimate of deaths from all causes). That procedure yielded
benefits of $60,000 per heart attack and $44,000 per stroke
avoided. It is important to note that these estimates do not
account for any reductions in the quality of life for the victims
of heart attacks and strokes (e.g., the partial paralysis that
afflicts many stroke victims).
Valuing reductions in the risk of death is difficult and
controversial, with a wide range of estimates in the literature.
EPA's RIA guidelines, for example, suggest a range of $400,000 to
$7 million per statistical life saved. For our point estimate,
we used a value from the lower end of that range, $1 million per
case. The benefits of reduced mortality dominate our estimates
of total blood-pressure-related benefits, which range from almost
$6 billion in 1986 to about $4.5 billion in 1992.
Benefits of reducing pollutants other than lead. Reducing
the amount of lead in gasoline should decrease emissions of
several pollutants in addition to lead. Most of these reductions
-------
E-8
will result from less misuse of leaded gasoline in vehicles that
should use unleaded to protect the effectiveness of their pollu-
tion-control catalysts. EPA expects the rule to significantly
reduce misfueling because it will be more expensive to produce
leaded regular gasoline (at 89 octane) with 0.10 gplg than to make
unleaded regular (at 87 octane). As a result, the gap between the
retail prices of unleaded and leaded regular gasolines should
narrow, and may well reverse. (At present, leaded regular is
slightly cheaper to make than unleaded regular, but the retail
price differential is much larger than the manufacturing cost
differential, apparently because of marketing strategies by
retailers.)
Eliminating misfueling would substantially reduce emissions
of HC, NOX, and CO. All three of these pollutants have been
associated with damages to health and welfare, and contribute to
ambient air pollution problems covered by National Ambient Air
Quality Standards (NAAQS). To predict the emission reductions
that would be associated with eliminating misfueling, we used
survey data on the extent of misfueling, tests showing the effect
of misfueling on emissions per mile traveled, and estimates of
the numbers of miles traveled by vehicles of different ages and
types. The rule also should reduce emissions of benzene (a
hydrocarbon that has been associated with leukemia) and ethylene
dibromide (a suspected human carcinogen, which is added to leaded
gasoline to control excess lead deposits in engines).
We valued these reduced emissions in two ways. The first
involved direct estimation of some of the health and welfare
effects associated with these pollutants. This method is
conceptually correct, but suffers from various uncertainties and
the inability to generate estimates for some potentially
important categories. Virtually all of the benefits that we
could quantify were associated with projected declines in ozone
as a result of reductions in HC and NOX emissions.
Our second method valued the emission reductions based on
the implict cost per ton controlled of the emission control
equipment destroyed by misfueling. The final estimate, used in
computing total and net benefits, was the simple average of the
two different methods. Assuming that the 0.10 gplg rule will
eliminate 80 percent of misfueling, those estimates range from
$222 million in 1986 to $248 million in 1992. (The estimates for
this category increase over time because the total amount of
misfueling is projected to increase in the absence of the rule.)
-------
E-9
Maintenance and fuel economy benefits. Reducing lead in
gasoline will decrease vehicle maintenance expenses associated
with the corrosive effects of lead and its scavengers on engines
and exhaust systems. The rule also is likely to increase fuel
economy, both because additional reforming to replace the octane
now provided by lead will increase the energy content of gasoline
and because leaded gasoline fouls oxygen sensors in newer
misfueled vehicles.
Three categories of maintenance benefits were estimated:
exhaust systems, spark plugs, and oil changes. Estimates for the
first two categories were based on fleet studies of vehicles in
use, which showed that exhaust systems and spark plugs last
longer with unleaded than with leaded gasoline. Estimates of oil
change benefits were based on studies showing that oil maintains
its quality longer with unleaded than with leaded. Summing
together these three categories, we estimate that reducing lead
in gasoline from 1.10 gplg to 0.10 gplg will yield benefits of
$0.0017 per vehicle mile, or about $17 per year for a vehicle
driven 10,000 miles. Because of the large number of vehicles
affected, the aggregate benefits are large, ranging from about
$900 million in 1986 to roughly $750 million in 1992.
The fuel economy estimate, as noted above, has two
components. To estimate the gain in fuel economy due to higher
energy content, we used the change in fuel density predicted by
the DOE refining model and applied it to a fuel economy formula
developed by the Society of Automotive Engineers. To estimate
the portion due to reduced fouling of oxygen sensors, we
estimated the change in the number of misfueled sensor-equipped
vehicles and used experimental data on how much extra fuel is
consumed by vehicles with fouled sensors. Total estimated fuel
economy benefits exceed $100 million in most years.
Costs and Benefits of Alternatives
Table E-l summarizes several important non-monetary measures
of the benefits of the 0.10 gplg standard for a single year,
1986. These estimates assume that the rule will eliminate all
misfueling. We also examined a wide range of alternative
standards, however, and considered various assumptions about the
impacts of those rules on misfueling, ranging from the rule
eliminating all misfueling to the rule having no impact on
current misfueling rates. In addition, we computed net benefits
with and without the preliminary estimates of blood-pressure-
related benefits.
Regardless of the assumption about misfueling, and whether
or not the blood-pressure-related benefits were included, we
found that net benefits were maximized at the tightest of the
alternative standards considered, 0.50 gplg for the second half
-------
E-10
TABLE E-l. Nonmonetary Measures of Health and Environmental
Benefits of 0.10 gplg in 1986, Assuming No Mis-
fueling
Reductions in thousands of
children above selected
blood lead levels
30 ug/dl 52
25 ug/dl 172
20 ug/dl 563
15 ug/dl 1,726
Reductions in thousands of
tons of pollutants
Hydrocarbons 305
Nitrogen oxides 94
Carbon monoxide 2,116
Reductions in adult male
health effects
Thousands of
hypertensives 1,804
Myocardial
infarctions 5,350
Strokes 1,115
Deaths 5,160
-------
E-ll
of 1985 and 0.10 gplg for 1986 and subsequent years. Table E-2
presents year-by-year estimates of the costs and benefits of the
final rule under our "partial misfueling" case, which assumes
that misfueling will continue unabated under the 0.50 gplg
standard in 1985, and then will be reduced by 80 percent under
the 0.10 gplg standard in 1986 and subsequent years. The net
benefits without blood-pressure-related effects exceed $1 billion
per year in 1986 and later years. If the blood-pressure-related
benefits are included, net benefits exceed $5 billion per year.
Although many of the individual components of these estimates are
subject to uncertainty, the magnitude of the estimated monetized
benefits relative to the costs, together with the many
potentially important unmonetized benefits, indicate that rapid
reductions of lead in gasoline are amply justified.
-------
E-12
TABLE E-2. Year-by-Year Costs and Monetized Benefits of Final Rule, Assuming
Partial Misfueling (millions of 1983 dollars)
1985 1986 1987 1988 1989 1990 1991 1992
MONETIZED BENEFITS
Children1s
health effects 223 600 547 502 453 414 369 358
Adult blood
pressure 1,724 5,897 5,675 5,447 5,187 4,966 4,682 4,691
Conventional
pollutants 0 222 222 224 226 230 239 248
Maintenance 102 914 859 818 788 767 754 749
Fuel economy 35 187 170 113 134 139 172 164
TOTAL MONETIZED
BENEFITS 2,084 7,821 7,474 7,105 6,788 6,517 6,216 6,211
TOTAL REFINING
COSTS 96 608 558 532 504 471 444 441
NET BENEFITS 1,988 7,213 6,916 6,573 6,284 6,045 5,772 5,770
NET BENEFITS
EXCLUDING BLOOD
PRESSURE 264 1,316 1,241 1,125 1,096 1,079 1,090 1,079
-------
CHAPTER I
INTRODUCTION
The Environmental Protection Agency (EPA) is promulgating a
rule to sharply reduce the lead content of gasoline from its
current limit of 1.10 grams per leaded gallon (gplg) to 0.10 gplg.
The phasedown will take effect in two steps: effective July 1,
1985, the limit will be 0.50 gplg; effective January 1, 1986, the
limit will be 0.10 gplg. Although EPA believes that these
standards are feasible for the industry as a whole, to increase
individual refineries' flexibility in meeting that schedule,
EPA has proposed a rule change to allow refineries to reduce
lead use below allowable limits in 1985 and "bank" those credits
for use in 1986 and 1987.
EPA's goal in promulgating this phasedown schedule is to
minimize the adverse health and environmental impacts of lead in
gasoline. To further aid in meeting that goal, the Agency is
considering the elimination of all lead in gasoline.
The health and environmental consequences of lead in gasoline
include both the direct health effects of exposure to lead and the
effects of higher emissions of conventional pollutants (hydrocar-
bons, nitrogen oxides, and carbon monoxide) from vehicles whose
pollution-control catalysts have been poisoned by the misuse of
leaded gasoline. In addition to these health and environmental
benefits, which form the basis for this rule, reducing lead in
gasoline will provide benefits to vehicle owners in the form of
increased fuel economy and reduced maintenance expenditures for
lead-induced corrosion of engines and exhaust systems. On the
-------
1-2
other hand, lead is a low-cost octane booster, so the proposed
rule will raise the cost of producing gasoline.
EPA has determined that both the final phasedown rule and
the possible ban are "major" regulatory actions under the criteria
of Executive Order 12291, because higher gasoline production
costs would exceed $100 million per year under each action. For
major rules, the Executive Order requires a Regulatory Impact
Analysis (RIA); this document constitutes the final RIA for the
phasedown being promulgated. Although EPA has not relied on
banking in analyzing the costs or establishing the feasibility
of the final phasedown, this RIA also examines the impact of the
banking proposal on the costs and benefits of the phasedown
rule. A separate preliminary RIA has been prepared for the ban,
now under consideration for as early as 1988; it employs the
same methods used in this document, and summarizes the costs and
benefits of a ban. That preliminary RIA incorporates by reference
large parts of this final RIA.
The remainder of this chapter provides an overview of the
problem of lead in gasoline, a brief review of earlier rulemakings
and the proposal made in August 1984, an analysis of the rationale
for government intervention, and a discussion of alternative
regulatory approaches.
I.A. Background Information
The rules being promulgated and proposed are part of a
series of actions taken by EPA over the past eleven years to
address the health and environmental hazards posed by lead in
gasoline. This section provides some background information
-------
1-3
for these latest actions.
I.A.I. Previous Rulemakings
EPA has regulated lead in gasoline since 1973 to meet two
goals: assure the availability of unleaded gasoline for those
vehicles with pollution-control devices (catalysts) that are
rendered ineffective by leaded fuel, and reduce the adverse
health effects associated with exposure to lead.
EPA's original rule for lead in gasoline limited the lead
content per gallon averaged over all gasoline (both leaded and
unleaded) sold by each refinery (38 FR 33741; December 6, 1973).
EPA also temporarily set separate and less stringent interim
limits for small refiners.
In 1982, EPA promulgated new rules (47 FR 49331; October 29,
1982) that: (1) switched the basis of the standard from all
gasoline to leaded only, (2) set a limit of 1.10 gplg, (3) phased
in uniform treatment of all refiners regardless of size, and (4)
allowed "constructive averaging" or "trading" of lead use across
refineries (e.g., one refinery could produce a gallon with 1.20
grams of lead if it traded with another refinery that produced a
gallon with only 1.00 gram).
The purpose in switching from a standard based on the
overall pool average to one based only on leaded gasoline was to
reduce total lead usage as sales of leaded gasoline declined
with the retirement of the older vehicles allowed to use it.
(With the overall gasoline-pool standard, refiners would have
been allowed to increase the amount of lead per gallon of leaded
gasoline as the total demand for leaded declined.) When EPA
-------
1-4
promulgated the 1982 rule, it projected a steady and substantial
decline in total lead in gasoline over the next decade.
I.A.2. Continuing Problems
Since 1982, new studies and reanalyses have shown lead to be
a greater health risk than previously thought. In addition,
total lead in gasoline has not declined as projected in 1982.
Figure 1-1 compares the 1982 lead projections with actual lead
use in 1983 and with current projections for later years.
The major cause of higher lead levels than previously pro-
jected is misfueling, the use of leaded gasoline in vehicles
designed to use unleaded. EPA's 1983 survey of vehicles in use
indicated that about 15.5 percent of vehicles required to use
unleaded gasoline were misfueled with leaded (U.S. EPA, 1984e).
By comparison, in the 1982 survey, the overall rate was 13.5
percent, with lower average rates in areas with Inspection and
Maintenance (I/M) mobile source enforcement programs (6.2
percent) than in areas without such programs (15.1 percent)
(U.S. EPA, 1983a). The 1983 survey showed increased misfueling
in I/M areas, reducing the gap. These estimates probably under-
state the true extent of misfueling because owners were not
required to submit their vehicles for testing. Moreover, the
surveys show higher misfueling rates in older vehicles, so
misfueling rates are likely to grow in future years as the
average age of catalyst-equipped vehicles increases.
Misfueling not only increases lead emissions, but by
poisoning pollution-control catalysts, it increases emissions
of hydrocarbons (HC), nitrogen oxides (NOX), and carbon monoxide
-------
1-5
L.
o
0>
>x
E
2
o>
o
Q>
V)
O
0)
70
60
50 -
40 -
30 -
20 -
10 -
rrent Projection
1982 Proj«ctio
i i i i i i I I
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
Year
FIGURE 1-1. Comparison of 1982 and. Current Projections of Lead Use in Gasoline
-------
1-6
(CO). In vehicles equipped with three-way catalysts (virtually
all post-1981 cars), lead-induced poisoning of catalysts increases
emissions by 500 percent for HC, 100 percent for NOX, and 300 per-
cent for CO (U.S. EPA, 1983b). Each of these pollutants is
regulated by vehicle emission standards. Each is also covered
by a National Ambient Air Quality Standard (NAAQS) or contributes
to a pollution problem covered by a NAAQS (HC and NOX form ozone,
and NOX forms NO2).
Continued use of lead in gasoline, whether to meet the
demands of misfuelers or of legal users, poses a serious threat
to health. Several studies have shown a strong relationship
between lead in gasoline and levels of lead in children's blood,
with blood lead levels following gasoline lead closely, tracking
seasonal fluctuations as well as long-term trends. Analyses
using several different data sets show that this relationship
remains strong and statistically significant even when other
potentially confounding variables are controlled for using multiple
regression and other statistical techniques. These analyses
show that gasoline lead is related to blood lead levels in both
adults and children.
Analyses also show that gasoline lead not only increases
average blood lead levels, but also raises the number of children
with dangerously high blood-lead levels. Statistical analyses
indicate that lead in gasoline accounts for about half of the
number of children above 30 raicrograms per deciliter (ug/dl),
until recently the level set by the Center for Disease Control
(CDC) for follow-up testing and possible treatment. Lead paint,
-------
1-7
emissions from stationary sources, and other non-gasoline sources
of lead exposure account for the other children at high levels,
although some of these additional exposures, including lead in
food, partially reflect past emissions of lead from gasoline.
The adverse health consequences of high levels of lead in
children are well accepted. They include damage to the kidney,
the liver, the reproductive system, blood creation, basic cellu-
lar processes, and brain functions. The CDC recommends that
children with blood lead levels above 25 ug/dl receive follow-up
testing and possible medical treatment; upon further testing,
about 70 percent of children with blood lead levels above 25
ug/dl are expected to be classified as "lead toxic" under CDC
criteria.
Increasing evidence also points to health effects at blood
lead levels below 25-30 ug/dl. These effects include inhibition
of certain enzyme activities, changes in EEC patterns, impairment
of heme synthesis, increases in levels of a neurotoxic chemical,
possible interference with neurotransmission processes, impairment
of vitamin D activity, and impairment of globin synthesis.
Several studies also have found indications of lead affecting
cognitive functions (as measured by IQ tests and other means), as
well as having other neurobehavioral effects, at levels well
below 30 ug/dl. Although EPA has not reached definite conclusions
regarding the specific blood lead levels at which such effects
occur, the Clean Air Scientific Advisory Committee (CASAC) has
recommended that the Agency consider these studies in the
Criteria Document for the ambient air quality standard for lead.
-------
1-8
Concerns about the health effects of ambient exposure to
lead traditionally have focused on children. Although lead has
a variety of adverse effects on the health of adults, most of
the known effects appear not to be of substantial concern except
at blood-lead levels in excess of 30-40 ug/dl. Recent analyses,
however, indicate a strong and robust relationship between blood
lead levels and blood pressure, with no apparent threshold.
Those findings have important implications for estimating the
benefits of reducing lead in gasoline, because high blood
pressure, in turn, is linked to a variety of cardiovascular
diseases including hypertension, myocardial infarction and strokes,
I.A.3. The August 1984 Proposal
Concerns about the health risks posed by lead in gasoline
and about the growing misuse of leaded gasoline in catalyst-
equipped vehicles led EPA to consider additional restrictions on
the lead content of gasoline. In August 1984, the Agency pro-
posed to reduce the allowable lead limit to 0.10 gplg, effective
January 1, 1986. EPA also indicated that it would consider
alternative phasedown schedules with more than one step that
would start earlier than 1986. In addition, it offered two
alternatives for the final elimination of lead in gasoline: a
ban to take effect in the mid-1990s, or allowing market forces
to eliminate leaded gasoline as demand shrank.
The proposal and the alternatives reflected EPA's desire-to
eliminate lead in gasoline quickly, tempered by two concerns.
First, the Agency wanted to ensure that the phasedown schedule
-------
1-9
could be met without excessive costs or gasoline shortages due
to lack of time to build additional refining capacity. The
Agency's analyses, based on the Department of Energy's (DOE)
refinery yield model, indicated that refineries could meet the
0.10 gplg standard with existing equipment, but in the proposal
EPA solicited comments and additional data on that issue.
Second, EPA was concerned that eliminating lead in gasoline
altogether might pose a serious risk of premature valve-seat wear
for certain older engines. These engines include those in older
automobiles (primarily those manufactured before 1971) and some
gasoline-powered trucks and off-road vehicles (including farm
equipment) that do not have hardened valve seats. Dynamometer
and track tests have shown that such engines can suffer premature
erosion of valve seats ("valve recession") with unleaded gasoline
under severe conditions (sustained high speeds and heavy loads),
but studies of vehicles in normal use generally have failed to
find excessive wear. However, based on the available studies,
the proposal allowed 0.10 gplg to provide a margin of safety for
those engines that might need lead to protect against undue
valve seat recession.
I.A.4. Information Received After August 1984 Proposal
Several pieces of information received after the August
1984 proposal have strengthened EPA's determination to reduce
lead in gasoline as quickly as possible. The CDC has now lowered
its recommended cut-off level for follow-up testing (from 30 ug/dl
to 25 ug/dl) as well as its definition of lead toxicity. The
CDC cautions, however, that even this new, lower level should
-------
1-10
not be regarded as "safe"; rather, it reflects tradeoffs between
protecting children against the adverse effects of lead and the
risks associated with treatment, as well as the limited ability
of screening tests to reliably measure blood-lead levels below
25 ug/dl. The reduction in this cut-off level more than triples
the number of children at risk under CDC criteria.
Additionally, comments and data from refiners, plus additional
sensitivity analyses using the DOE refining model, have reinforced
the Agency's conviction that the refining industry as a whole
can meet a 0.10 gplg standard within a year of promulgation
using existing equipment. Moreover, these same sources indicate
that it is possible to achieve significant reductions even sooner.
Thus, the Agency has decided to add an interim standard of 0.50
gplg to take effect on July 1, 1985.
Although EPA believes that both standards are feasible for
the industry as a whole, EPA's recent proposal to extend the
lead trading program to allow "banking" also will increase flexi-
bility and help to ensure that individual refineries can meet
the phasedown schedule in the most cost-effective manner. Under
that proposed rule change, refineries using less lead than allowed
under the applicable limits in 1985 could "bank" those extra
lead rights for use in meeting the tighter limits in 1986 and
1987. Banking should reduce significantly the overall costs
of meeting the phasedown limits and increase a refinery's
flexibility in meeting unexpected problems, without permitting
any additional use of lead over the three-year period, 1985
through 1987.
-------
1-11
Three additional types of information have contributed to
EPA's decision to consider a ban on lead in gasoline by 1988.
First, the Agency has received data indicating that valve-seat
recession with unleaded gasoline may be a much less serious
problem than earlier feared. The Agency has reviewed data from a
large test performed by the U.S. Army in the mid-1970s that in-
volved switching many types of vehicles — including heavy-duty
trucks, construction equipment, motorcycles, and stationary engines
(such as generators), as well as light-duty vehicles — from
leaded to unleaded gasoline. The Army's study showed no detectable
increase in valve-seat problems and resulted in all of the armed
services switching to unleaded gasoline (where available) by 1976.
In addition, the U.S. Postal Service switched its heavy-duty
trucks to unleaded in 1980, and a review of their computerized
maintenance records shows no evidence of abnormal rates of
valve seat recession. Moreover, reanalyses of other studies
suggest that lead may cause serious engine durability problems.
EPA is reviewing these and other studies relevant to this issue.
Second, although EPA continues to believe that the phasedown
rule will have a significant impact on misfueling, it is less
confident that the problem will be eliminated so long as any
lead remains in gasoline. Moreover, the 1983 EPA misfueling
survey, the results of which became available after the August
proposal, showed that misfueling continues to be a serious problem,
and that the rates appear to be increasing.
Finally, two recently published studies (Pirkle et al., 1985;
Harlan et al., 1985) indicate a strong relationship between blood
-------
1-12
lead and blood pressure in adult males. Because of the well-
established relationships between gasoline lead and blood
lead, and between blood pressure and cardiovascular disease,
these studies have potentially important implications for public
health. Preliminary estimates based on Pirkle et al. indicate
that lead in gasoline may be responsible for well over one million
cases of hypertension per year and for over 5,000 deaths from
heart attacks, strokes, and other diseases related to blood pres-
sure. Moreover, these estimates cover only males aged 40 to 59
and, in the case of heart attacks, strokes, and deaths, only white
males in that age group. (The Pirkle et al. study did not find a
statistically significant association between lead and blood pres-
sure in females, and available studies of the effects of blood
pressure on cardiovascular risks provide the most reliable dose-
response estimates for whites in that age range.) The benefits
may extend to older males and to nonwhites as well.
The Pirkle et al. and Harlan et al. studies have b'een published
in peer-reviewed journals. A summary of the Pirkle et al. study
also was placed in the docket of the phasedown rulemaking (Schwartz,
1984c). Because the papers have only recently been published and
have not yet received widespread review by the scientific community,
EPA has not relied on estimates of health effects related to blood
pressure in deciding on the phasedown rule being promulgated now.
These effects will be considered, however, in reaching a decision
on a ban. We have included a chapter in this RIA on the potential
impacts of the phasedown on blood pressure and cardiovascular
disease.
-------
1-13
I-B. Need for Government Intervention
Lead in gasoline is a classic example of what economists
call a negative externality; individual users of leaded gasoline
do not bear all of the costs it imposes on society as a whole.
Users of leaded fuel reap short-term savings in the form of lower
fuel costs; they also bear higher maintenance costs. Individually,
however, they bear only an infinitesimal fraction of the costs of
the health and environmental damages caused by their vehicles'
emissions of lead and, in the case of misfuelers, other pollutants
from poisoned catalysts. This disparity between private and social
costs generates an overuse of the good, in this case lead in gaso-
line, and increases the damages imposed on society.
The need for government intervention to rectify significant
negative externalities is well recognized and provides the pri-
mary intellectual basis for virtually all EPA regulations. The
predominant approach in the United States has been to impose
standards that limit the level of the externality (in this case,
the amount of lead permitted in gasoline). Market-oriented
approaches — such as pollution taxes or charges on emissions and
marketable permits — are rarely used, although they may be highly
efficient means for reducing negative externalities. These alter-
native approaches are discussed more fully in later sections of
this chapter.
Section 211(c)(l) of the Clean Air Act [42 U.S.C. §7545(c)
(1)] gives the Administrator of EPA broad authority to "control or
prohibit the manufacture ... or sale of any fuel additive" if its
emission products (1) cause or contribute to "air pollution
-------
1-14
which may be reasonably anticipated to endanger the public
health or welfare", or (2) "will impair to a significant degree
the performance of any emission control device or system ...
in general use ..."
EPA believes that further reductions in the lead content of
gasoline are justified by both of the tests under Section 211(c)(l)
Lead in gasoline has been shown to raise blood lead levels, which
endangers public health, and the misuse of leaded fuel damages
pollution control devices, substantially increasing emissions of
HC, NOX, and CO.
I.C. Alternatives to New Regulations
Regulation of lead in gasoline is amply supported both by
statutory authority and the significant negative externalities
associated with leaded gasoline. In this instance, however, EPA
already has regulations in place, so the issue is not whether
there should be any regulation at all, but whether the regulations
should be tightened. Before proposing stricter rules, the Agency
carefully considered alternative approaches that would not re-
quire new rules. Three such alternatives — no change in policy,
public education efforts, and stepped-up enforcement against
misfueling — are discussed in this section.
I.C.I. No Change in Policy
Under this approach, the Agency would not change existing
regulations and policies regarding lead in gasoline. Lead use
would decline over time, as existing vehicles designed to use
-------
1-15
leaded gasoline retired and the number of vehicles designed to
use unleaded increased. As discussed earlier, however, lead use
would decline less rapidly than expected at the time of the
previous rulemaking, and the number of misfuelers would grow.
Our best estimate is that leaded fuel would not decline below 20
percent of total gasoline fuel use under this scenario.
All of the benefit-cost calculations in later chapters
implicitly evaluate this alternative, because it is the baseline
from which costs and benefits are measured; by definition, it has
zero costs and benefits, and thus zero net benefits in each year.
Policies with positive net benefits yield higher net benefits than
this alternative. Conversely, policies with negative net benefits
are less efficient, in economic terms, than the status quo.
I.C.2. Public Education
Efforts to educate the driving public about the problems
caused by misfueling offer potentially useful supplements to
current policy. As discussed more fully in Chapter VII, the use
of leaded gasoline increases vehicle maintenance costs. Indeed,
EPA estimates that the maintenance savings of reducing lead in
gasoline would exceed the higher costs of manufacturing unleaded
fuel. EPA is undertaking a variety of initiatives to inform
vehicle owners of these maintenance costs, as well as the
adverse environmental effects of misfueling.
Public education efforts, while useful, are unlikely to
significantly reduce the use of leaded gasoline, in large part
because retail price differentials between leaded and unleaded
-------
1-16
gasoline are high, roughly three to four times higher than the
manufacturing cost differentials. (See Chapter II for further
discussion of this issue.) In addition, the health and environ-
mental costs of using leaded gasoline are externalities that are
not borne by individual users of leaded gasoline. Thus, even if
the social benefits of reducing lead in gasoline exceed the social
costs by a large margin, the strictly private benefits (reduced
maintenance costs) to most individual purchasers may be less than
their private cost (the retail price differential).
I.C.3. Stepped-Up Enforcement
Federal law makes it illegal for service stations and other
gasoline dispensing outlets to put leaded fuel in vehicles
requiring unleaded. Federal law does not apply to individual
vehicle owners who misfuel, although many states have such laws.
In theory, therefore, the misfueling problem could be mitigated
by stepped-up enforcement of existing state and federal laws.
Unfortunately, however, massive enforcement efforts would
be very expensive, and only partially successful in eliminating
misfueling. Moreover, even if it were possible to eliminate
misfueling, the serious health effects associated with legal use
of leaded gasoline would continue unabated.
Misfueling presents a serious enforcement problem.
Currently, almost 100 million vehicles require unleaded fuel, and
there are over 100,000 retail gasoline outlets, plus even more
private outlets (e.g., private tanks used to fill commercial
-------
1-17
fleets). Inspection and Maintenance (I/M) mobile source enforce-
ment programs test the emissions of individual cars, requiring
owners to make repairs if their cars exceed emission standards.
I/M programs appear to offer the most practical means of enforce-
ment, but less than one-fifth of all vehicles are in areas with
such programs.
I/M programs are used in areas to help achieve compliance
with National Ambient Air Quality Standards, primarily by encour-
aging the improved maintenance of vehicles and their emission
control equipment. They are an important part of strategies to
attain air quality standards, and they produce substantial emis-
sion reductions independent of their effects on misfueling.
Extending I/M programs to areas already attaining standards would
be expensive; a typical program costs about $6.50 per vehicle
inspection (U.S. EPA, 1981). By the mid-to-late 1980s, more than
100 million vehicles requiring unleaded gasoline will be registered
in areas that do not currently have I/M programs. If we assume
annual inspections, the cost of extending I/M programs to cover
all unleaded vehicles would be about $650 million per year, which
is higher than the estimated cost of reducing lead to 0.10 gplg
in all leaded gasoline.
This estimate understates the cost of an I/M approach because
it does not include the costs of required repairs, the time spent
by owners to have their vehicles inspected, or the higher costs
of manufacturing additional unleaded gasoline to meet the demand
of existing and potential misfuelers deterred by inspections.
These categories of costs are likely to be substantial.
-------
1-18
A national I/M program also is likely to miss many misfuelers.
Surveys show that while misfueling rates are lower in areas with
I/M programs than in those without, they are still significant;
the average misfueling rate in I/M areas appears to be about two
thirds of the overall national average, but the difference has
been shrinking (see Chapter VI). An I/M program aimed speci-
fically at misfueling (e.g., including tests for lead in tailpipes)
probably would be more successful, but even an extensive I/M
program is unlikely to solve the misfueling problem.
More importantly, I/M programs and other aids to enforcement
have no impact on lead emissions from vehicles legally permitted
to use leaded gasoline. EPA estimates that in 1983, legal users
accounted for about 85 percent of all leaded gasoline sales. Over
time, with a decrease in the number of vehicles allowed to use
leaded and a rise in the number of vehicles designed for unleaded,
that proportion will fall. Even in 1988, however, EPA projects
that, in the absence of new rules, roughly two-thirds of the
demand for leaded gasoline will be legal. Thus, we estimate that
even a fully effective enforcement program targeted on misfueling
could not reduce lead emissions from vehicles by more than 30
percent for the period from 1986 to 1990.
I.D. Market-Oriented Alternatives
Market-oriented approaches to environmental protection offer
the advantage of allocating control efforts on the basis of mar-
ginal control costs, yielding minimum costs for any given level
of overall protection. The two major alternatives are marketable
permits and emission charges.
-------
1-19
I.D.I. Marketable Permits
EPA's current lead-in-gasoline regulations, by allowing
"constructive averaging," in effect set up a system of marketable
permits for lead in gasoline. Refineries with lower-than-average
costs for producing octane without lead are allowed to reduce
their lead content below the limit of 1.10 gplg and to sell the
excess lead rights to refiners with higher-than-average costs, who
may then produce leaded gasoline with lead content above 1.10 gplg.
This approach offers clear advantages over a uniform standard of
1.10 gplg. Lead is reduced by the same amount, but costs are
lower. For those refineries that lower the lead content of their
gasoline, the sale of excess lead rights more than offsets higher
manufacturing costs. Conversely, for those refineries that buy
lead permits, the savings in manufacturing costs more than offset
the purchase price of the permits. Consumers presumably also
benefit when these lower costs translate into lower gasoline
prices.
The current system is working well. Based on confidential
reports filed with EPA by the industry, about 73 percent of
refiners participated in lead trading during the second quarter of
1984. During that same period, about 14 percent of the total
amount of lead was traded. Thus, the lead trading system initi-
ated in 1982 appears to be a major success story for introducing
market principles into environmental regulation.
Under the new rule being promulgated, refineries will be
allowed to continue constructive averaging through 1985, until
the 0.10 gplg standard takes effect. Once the 0.10 gplg standard
-------
1-20
applies, however, inter-refinery averaging no longer will be
allowed, because that standard is intended to protect those few
engines that may need a small amount of lead (or some other
additive) to protect against premature valve-seat recession.
Although EPA believes that the risk is small and that minor
fluctuations around this level over time will not damage engines
(hence, the rule permits quarterly averaging by refiners), con-
stant use of fuel with a lower lead content might damage a few
engines. For a trading system to work, some refiners must produce
gasoline with a lower lead content than permitted by the rules,
but that could cause damage to some engines because many vehicle
owners consistently purchase fuel produced by the same refinery.
Moreover, the refinery cost savings from allowing trading at 0.10
gplg would be trivial, as the amount of lead involved will be so
small.
EPA has proposed a rule change that will allow refineries
greater flexibility in meeting lead-phasedown goals without any
detrimental effects on health, the environment, or those engines
that may need protection against valve seat recession. Under
this proposal, refiners would be able to save some of the lead
that they are allowed to use during 1985 for use in 1986 or
1987. Thus, for example, if a refinery produced five million
gallons of leaded gasoline at 0.90 gplg during the second quarter
of 1985 (when the limit will still be 1.10 gplg), and did not
sell the extra lead rights to other refineries, it could "bank"
one million grams of lead ([1.10 - 0.90] x [5 million] = 1
million) for use in 1986 or 1987. Similarly, a refinery that
-------
1-21
produced five million gallons at 0.40 gplg during each of the
last two quarters of 1985 (when the limit will be 0.50 gplg)
also would have one million grams available for supplemental use
in 1986 or 1987. These banked lead rights could be used as
desired by the refinery in 1986 or 1987; it could, for example,
produce 10 million gallons at 0.20 gplg (rather than the 0.10
gplg standard that will apply then), or sell the million grams
to another refinery with higher marginal production costs. This
change will increase refineries' flexibility (thus lowering
production costs) without increasing the amount of lead released
to the environment.
I.D.2. Pollution Charges
The second market-oriented alternative to mandatory lead
content reductions would be to impose a charge on gasoline based
on its lead content. Ideally, the charge per gram of lead should
equal the marginal external damage caused by a gram of lead in
gasoline, thereby internalizing the health and environmental
damages caused by lead use. With such a charge, the price of
leaded gasoline would rise, encouraging consumers to switch to
unleaded. Refiners would have an incentive to reduce the lead
content of gasoline to the point where the marginal cost of
additional reductions would equal the charge. As with marketable
permits, refiners would end up producing leaded gasoline with
different lead levels, depending on the refiners' marginal costs.
Despite these attractive features, a charge has several
drawbacks and faces some obstacles. The primary drawback is the
same as identified earlier with regard to permits: leaded gaso-
-------
1-22
line that contains much less than 0.10 gplg will not protect those
few engines that may need valve lubrication. The charge might be
modified in one of two ways to address this problem: it could
be levied only on lead in excess of 0.10 gplg, or EPA could set a
minimum lead-content rule for leaded gasoline. In either case,
EPA believes that an appropriate charge would be high enough to
drive virtually all leaded gasoline to the minimum level of 0.10
gplg. If that happened, the cost-minimizing advantage of this
approach would be lost; the effect would not be materially
different from a uniform standard of 0.10 gplg.
Even with these limitations, a charge would still offer some
advantages. It would give refiners additional flexibility over
the short run while they made needed adjustments in equipment and
operating practices. If the charge were levied on all lead in
gasoline (not just that above 0.10 gplg), it also would provide
an additional disincentive for misfueling with 0.10 gplg gasoline
by raising its price. In addition, it would discourage the use
of leaded gasoline in other vehicles unless required to prevent
valve-seat recession.
One potential obstacle to levying a charge to reduce lead in
gasoline is that EPA may not have the necessary statutory author-
ity. If new legislation, were needed, it almost certainly would
be accompanied by extensive debate about both the general principle
of using charges or taxes to control environmental hazards and the
specific tax rate to be set for lead, which could cause significant
delays in solving a serious environmental problem. EPA believes
that the foregone benefits caused by such a delay would outweigh
any potential efficiency gains.
-------
1-23
I«E« Alternative Standards
EPA has considered a wide range of alternative phasedown
schedules. These alternatives included both different lead levels
and different effective dates. Chapters II through VII present
the methods used to estimate the costs and benefits of these
alternatives. In each chapter, we survey the effects in question
and discuss the methods used to derive monetized estimates. Each
chapter contains year-by-year estimates of the relevant effects
for three sets of rules, as summarized in Table 1-1: (1) the
primary Proposal made in August (0.10 gplg in 1986), (2) the
sample Alternative schedule discussed in the August Notice of
Proposed Rulemaking, and (3) the Final Rule. (Because the 1985
standard only applies to the second half of the year, our
estimates for that year are for half a year.) In all cases, the
costs and benefits depend on the amount of misfueling eliminated.
The sample benefit estimates presented in Chapters III-VII assume
that all misfueling is eliminated. The cost estimates in Chapter
II assume that the amount of misfueling declines linearly with
the level of the standard, from 100 percent of current levels at
0.50 gplg or above, to zero with a ban. A range of alternative
assumptions about misfueling, and their impacts on the costs and
benefits, is explored in Chapter VIII.
Throughout this document, we use a 10 percent real discount
rate to compute present values, and costs and benefits are expressed
in 1983 dollars. The choice of a discount rate is controversial,
the subject of much debate among economists. Generally, the
higher the discount rate, the lower the net benefits, because
costs usually are incurred sooner than benefits.
-------
1-24
TABLE 1-1. Alternative Phasedown Schedules
Phasedown Schedules
Proposed
Alternative
Final
Lead Limit
1985* 1986
N.A. 0.10
0.50 0.30
0.50 0.10
(gpig)
1987
0.10
0.20
0.10
1988
0.10
0.10
0.10
* 1985 limit applies as of July 1.
-------
1-25
The rate we use, 10 percent, is at the high end of the ranges
typically discussed; many economists would regard a much lower
rate as more appropriate. We have used 10 percent to be consist-
ent with guidelines from the Office of Management and Budget
(U.S. OMB, 1981). We also considered presenting sensitivity
analyses using lower rates, but decided not to do so in the
interests of simplifying the presentation; lower discount rates
would increase our estimates of net benefits.
Chapter VIII summarizes the costs and benefits of the final
phasedown rule and compares them to a wide range of alternative
schedules. It also examines the impact of different levels of
misfueling under each of the alternatives. Regardless of the
assumption about misfueling, the schedule contained in the final
rule appears to offer the highest net benefits of the options
considered. Chapter VIII also examines the costs and benefits
of the recent proposal to allow banking of lead rights.
-------
CHAPTER II
COSTS OF REDUCING LEAD IN GASOLINE
Since the 1920s, petroleum refiners have added lead to gaso-
line as a relatively inexpensive way to boost octane. To meet
octane demands with reduced amounts of lead, refiners have a
variety of options, the most important of which is to perform
additional processing of gasoline components in reforming and
isomerization units. In addition, refiners also may employ
additives other than lead, such as alcohols or MMT, or they may
purchase additional high-octane components, such as aromatics or
butane. At the margin, however, all of these alternatives are
more expensive than lead for making octane, so the cost of produc-
ing gasoline will rise under the rule being promulgated.
These higher refining costs will comprise virtually all of
the costs of the rule. (The one potential exception, damage to
the valve seats of some engines with unleaded gasoline, applies
only to the possible ban, not to the final phasedown rule; the
valve-seat issue is discussed in Chapter VII.) To estimate those
costs, we used a model of the refining industry orignally developed
for the Department of Energy (DOE).
Section A of this chapter explains why we estimated the social
cost of the rule using predicted changes in manufacturing costs,
rather than changes in retail prices. Section B provides an over-
view of the DOE model and refinery processes generally. Section C
discusses the input data and assumptions used for our projections
and presents year-by-year cost estimates for reducing lead in
gasoline. Section D reports the results of numerous sensitivity
-------
II-2
analyses conducted to test the robustness of the cost projections.
Finally, Section E, for illustrative purposes, examines the impact
on costs of the proposal to allow "banking" of lead rights during
1985 for use in 1986 and 1987.
II.A. Price versus Cost Differences
The costs we are measuring are the social costs of reducing
lead in gasoline. Our estimates of these costs are based on
estimates of changes in the costs of manufacturing gasoline (and
other petroleum products). In the long run in a competitive
market, the change in manufacturing costs is likely to be fully
reflected in changes in the amounts paid by consumers. In the
short run, however, the total amount paid by consumers may be
less than or greater than the change in manufacturing costs,
depending on supply and demand elasticities and other factors.
The divergence between the change in manufacturing costs and the
change in the amount paid by consumers, though, will be comprised
mostly of transfers from one segment of U.S. society to another,
rather than real social costs (or savings).
By real social costs, we mean the costs of real resources
that are used to comply with the rule (i.e., the extra energy,
capital, labor, etc. that is needed to meet the tighter lead
standard). If retail prices rise less than costs in the short
run, then the losses to refiners (and their shareholders) will
be offset by gains to consumers. If prices rise more than costs,
then the losses to consumers will be offset by gains to producers.
In either case, the net cost to all citizens, including sharehold-
ers, is just the real resources used in meeting the rule. Thus,
-------
II-3
so long as the rule does not affect the costs of distributing or
selling the product, the change in manufacturing costs is a good
approximation of the social cost of the rule.
The DOE model estimates that at current lead levels (1.10
gplg) the marginal manufacturing cost differential between
unleaded and leaded regular grades of gasoline is less than two
cents per gallon. Retail prices, however, diverge by an average
of about seven cents per gallon (Weekly Petroleum Status Report,
1984, various issues). Some commenters suggested that the gap
between the model's estimate of the cost differential and current
retail price differentials indicates either an error in the model,
or that it would be more appropriate to use the retail price
difference as the measure of social cost. Both arguments have
some superficial plausibility, but fail to withstand careful
analysis.
The much larger retail price differential appears to reflect
marketing strategies within the industry rather than errors in the
model's cost estimates. Price differentials at the wholesale level
are far smaller than the retail price spread, and discussions be-
tween EPA and petroleum refiners confirmed that manufacturing cost
differences between unleaded and leaded gasolines were about two
cents per gallon, and that inter-refinery trades occurred at a
differential of two to two-and-a-half cents per gallon. Also, the
differential between the spot prices of unleaded and leaded regular
gasolines in barge quantities has been between one and four cents
per gallon over the last several years. At the end of 1983,
wholesale price differentials in the Gulf of Mexico were about
-------
II-4
three cents per gallon (Platt's Oilgram).
The additional price gap between unleaded and leaded gasolines
at the retail level does not appear to reflect incremental distri-
bution or other costs. Both types of gasoline are distributed
through the same network of pipelines, terminals, barges, and tank
trucks, and similarities in sales volumes suggest that differential
inventory carrying costs do not apply-
Two possible explanations of the price differential have been
offered. The first is that service stations use their lowest cost
gasoline as the "fighting grade," prominently advertising it at a
low price to attract customers (including those purchasing another
grade). If this explanation is correct, unleaded regular may
become the new fighting grade, as it will be less expensive to
manufacture 87 octane unleaded than 89 octane leaded with 0.10
gplg. (To produce 89 octane leaded gasoline at 0.10 gplg, the
octane level of the gasoline must be boosted to over 88 before
the lead is added.) In that case, the price of unleaded gasoline
would be lower than leaded. Service stations may continue to use
leaded gasoline as the fighting grade, however, because it is
the lower-volume product.
The second possible explanation is that retailers are engaging
in price discrimination, taking a larger margin on the grade
(unleaded) with less elastic demand. This explanation presupposes
that individual retail outlets do not face the flat demand curves
of a perfectly competitive market, but rather that gasoline sales
are characterized by monopolistic competition. Under this scenario,
-------
II-5
the individual station owner is a geographic monopolist, i.e., the
only retailer at that location. Thus, the demand curve facing the
operator is not perfectly flat, hut if the price is too high,
customers will drive to a competitor.
Under such a market structure, prices deviate from marginal
costs, roughly in inverse proportion to each product's demand
elasticity, but firms do not reap economic profits (i.e., profits
beyond a normal return on capital) because of free entry of new
competitors. If leaded customers are more price sensitive (per-
haps because they may have lower incomes and are more willing to
search out lower-price stations, or because the leaded price is
displayed more prominently), their demand elasticity (at the level
of the individual service station) will be higher than the elastic-
ity of demand for unleaded customers. As a result, price discrim-
ination will cause the operating margin (price minus wholesale
cost) to be lower on leaded than on unleaded. If this explanation
is correct, the 0.10 gplg standard should greatly reduce the retail
price differential between unleaded and leaded gasoline, though
it would not necessarily reverse it.
Whichever of these explanations is correct, the difference in
manufacturing costs, rather than in retail prices, wot'iv". represent
the social cost (as conventionally defined in cost-benefit
analysis) of reducing lead in gasoline. This conclusion would
fail to hold only if it were possible to show that reducing the
lead content of gasoline would increase the real costs of distrib-
uting gasoline. No such evidence is available.
-------
II-6
II.B. The Refinery Model
This section discusses the DOE refinery model used to estimate
costs. Section II.B.I provides a brief introduction to the model,
while Section II.B.2 gives an overview of refining processes and
how they are represented in the model.
II.B.I. Introduction to the DOE Model
The DOE refinery yield model uses a linear programming frame-
work to simulate refining operations (U.S. DOE, 1984b). The model
represents individual refinery units and their inter-relationships
using a series of about 350 equations. Given a set of input
assumptions and constraints, it finds the least-cost method of
producing any specified set of final products. (The model also
can be run to maximize profit given final-product prices.)
In addition to estimating the total cost for a set of product
demands, inputs, and constraints, the model generates detailed and
useful information on important aspects of industry operations,
including the rates at which costs change (the shadow prices of
the constraints) and the extent to which various types of equipment
are utilized under different scenarios.
The model is based on many similar models developed and used
widely by the petroleum refining industry for its own planning
purposes. The refining industry was among the first to make
extensive use of linear programming. The DOE model has been used
by EPA for several years in its analyses of the impacts of regu-
lations on the petroleum industry and on petroleum product pur-
chasers. In the 1982 lead-in-gasoline rulemaking, EPA used the
-------
II-7
model to predict the costs of the rule and to predict the value
of lead rights under the trading program. Although it is impos-
sible to verify directly the cost estimates made then, it is
important to note that the actual prices of traded lead rights
were slightly lower than predicted by the model, indicating that
it did not underestimate the value of lead to refiners as an
octane booster.
The model also has been used by DOE for several purposes,
including: evaluating crude mixes for the selection of storage
sites for the Strategic Petroleum Reserve, assessing the impacts
of petroleum disruptions on product supplies, and evaluating the
industry's capability to respond to changes in feedstock quality
or product demands.
DOE subjected the model to two verification checks, and found
that for 1982 it generated a product slate very close to the actual
one. It also closely predicted refinery-gate price differentials.
It is also important to note that these verifications covered a
period before EPA initiated interrefinery averaging of lead use,
indicating that the accuracy of the model's predictions is not
dependent on the trading of lead rights. More recently, DOE
verified the model using 1983 data (U.S. DOE, 1985).
II.B.2. Overview of Refining Processes
So the reader can better understand how the model works, we
provide a brief description of refining processes. Figure II-l
is a schematic diagram of a very simple refinery, often called a
topping plant, which processes low-sulfur crude oils. A complex
-------
II-8
refinery contains distillation units and other types of processing
units; Figures II-2, II-3, and II-4 present schematics of such
refineries. (The model contains considerably more detail than even
these exhibits indicate.) In any given refinery, these different
process "units" are assembled into final structures that accomplish
different but related purposes. The basic similarity of process
units makes it possible to model refineries.
Basically, the model is a system in which the various units
that make up all types of refineries are represented by the boxes
in the schematics. Each unit uses inputs (crude oil or an inter-
mediate product) to make one or more intermediate or final products.
The exact types and quantities of the product(s) made are functions
of the properties of the inputs of each unit and the process that
each performs. Fuel and utilities (e.g., electricity and steam)
are consumed, and an operating cost is incurred for each operation.
A capital cost may or may not be charged, as appropriate to the
particular analysis being performed. Table II-l provides a summary
of the basic types of refinery processes. Appendix A contains a
more detailed description of processing operations.
Because all refineries are made up of these building blocks,
the smallest structure in the model is a process unit rather than
a plant. The individual functions that are modeled are the inputs
and outputs from each type of unit. The model is made up of re-
finery units, each of which has an output (or a series of products),
the quantity of which is a function of the material that the unit
is "fed." Some refinery process units incur some costs that vary
with how intensely they are run — called "severity". In the case
-------
II-9
Figure n-i
Schematic Diagram of a Bimple Oil Refinery
(Topping Plant)
FLOW DIAGRAM OF TOPPING REFINERY
PROCESSING- LOW SULFUR CRUDE OIL
GAS
CRUDE
OIL
DISTILLATION
RESID
GASOLINE (LOW OCTANE)
NAPHTHA
LIGHT GAS
HEAVY GAS
UE
OIL ^
OIL H
«-
e
L
E
N
D
1
N
G
8
L
E
N
D
1
N
G
REFINERY
GAS FUEL
(CONSUMED
INTERNALLY)
PETCMEM FEED
B MILITARY
JET FUEL
KEROSENE,
DISTILLATE
'FUEL OIL ft
DIESEL FUEL
YIELD. VOLUME
PERCENT OF
RAW MATERIALS
3.1
33.1
25.5
RESIDUAL
J
REFINERY
LIQUID FUEL
(CONSUMED
INTERNALLY)
37.2
*
^''Included with gas fuel
-------
11-10
Figure II-2
Schematic Diagram of a Hydroskimming Refinery
FLOW DIAGRAM OF HYDROSKIMMING REFINERY
PROCESSING LOW SULFUR CRUDE OIL
GAS
BUTANE (HIGH OCTANE)
GASOLINE (LOW OCTANE)
GASOLINE (HIGH OCTANE)
RESIDUE
REFINERY GAS
FUEL (CONSUMED
INTERNALLY)
GASOLINES
YIELD. VOLUME
PERCENT OF
RAW MATERIALS
4.9
2.2
29.0
KEROSENE,
JET FUEL.
DISTILLATE
FUEL OIL 8
DIESEL FUEL
RESIDUAL
FUEL OIL
REFINERY LIQUID
FUEL(CONSUMED
INTERNALLY)
25.5
38.3
*
* Included with gas fuel
-------
11-11
Figure II-3
Schematic Diagram of a Fuels Refinery
FLOW DIAGRAM OF FUELS REFINERY
PROCESSING HIGH SULFUR CRUDE OIL
r
NATURAL CAS LIQUIDS
HYDROGEN
RECOVERY 8
MANUFACTURE
HYDROGEN
GASOLINE (LOW OCTANE)
GASOLINE (HIGH OCTANE
((OPTIONAL!
I
L_ — _> (ASPHALT)
SULFUR
FOR HYDROGEN
TREATING
REFINERY GAS
FUEL (CONSUMED
INTERNALLY)
•+LPG
GASOLINES
KEROSENE,
JET FUEL,
DISTILLATE
FUEL OIL 8
DIESEL FUEL
RESIDUAL
FUEL OIL
REFINERY
LIQUID FUEL
(CONSUMED
INTERNALLY)
YIELD. VOLUME
PERCENT OF
RAW MATERIALS
1.5*
11.0
2.3
53.8
27.4
8.1
•' Percent by weight
"« Included with pas fuel
-------
11-12
Figure II-4
Schematic Diagram of a High Conversion Refinerv
FLOW DIAGRAM OF HIGH CONVERSION REFINERY *
PROCESSING HJGH SULFUR CRUDE OIL
NATURAL GAS LIQUIDS
(HIGH OCTANE)
GASOLINE (LOW OCTANE)
GASOLINE (HIGH OCTANE)
LIGHT GAS OIL
DISTILLATION
HEAVY GAS
OIL
GASOLINE (MEDIUM OCTANE)
DISTILLATION
VACUUM
PITCH
I I GASOLINE (LOW OCTANE)
!—(NAPHTHA
GAS OIL
SULFUR
FOR HYDROGEN
TREATING
REFINERY
GAS FUEL
(CONSUMED
INTERNALLY)
GASOLINES
KEROSENE.
JET FUEL.
DISTILLATE
FUEL OIL 8
DIESEL FUEL
REFINERY
LIQUID FUEL
(CONSUMED
INTERNALLY)
HIGH SULFUR
COKE
YIELD. VOLUME
PERCENT OF
RAW MATERIALS
1.4*
13.0
77.5
10.8
4.6*
* Percent by weight
** Included with gas fuel
-------
11-13
TABLE II-l. Functional Characterization of Refinery Processes
SEPARATION
Separation on the Basis of
molecular Weight
Distillation (atmospheric and
vacuum fractionation of crude
oil, naphtha splitting,
depropanizing, stabilization)
Absorption (recovery of olefins
frcm catalytic cracked gas,
recovery of propane from
natural gas or hydrocracked
gas)
Extraction (deasphalting of feed-
stock for lubricating oil manu-
facture or for catalytic
cracking)
Separation on the Basis of
molecular Structure
Extraction (recovery of
benzene, toluene, and zylenes
from catalytic reformate,
removal of aromatics from
lubricating oil feedstock)
Crystallization (dewaxing of
lubricating oils, recovery of
paraxylene from mixed xylenes)
ALTERATION (CONVERSION)
A. Conversion on the Basis of
Molecular Weight
Thermal cracking (visbreaking,
coking)
Catalytic cracking
Hydrocracking
Alkylation
Polymerization
B. Conversion on the Basis of
Molecular Structure
Catalytic reforming (benzene,
toluene, and xylene manufac-
ture and octane improvement)
Isomerization (normal butane to
iso for alkylation, normal
pentane and hexane to iso
for octane improvement)
TREATMENT TO REMOVE IMPURITIES
Hydrogen treatment (hydrotreating)
Caustic treatment (Merox, Bender)
Clay treatment (of lubricating oils)
Acid treatment
-------
11-14
of some important octane-improvement processes (particularly
reformers), the higher the severity, the higher the octane pro-
duced, but also the lower the "yield" (the amount of gasoline
material produced per unit of input). In general, the more the
process unit is used, the higher its marginal cost of processing
the next increment of feed will be.
In the model, all processes consist of a series of linear
relationships that describe the process output and operating cost,
given specified inputs and a set of operating conditions. The
relationships are stored in the model in the form of a process
data table. (Table II-2 shows an example of such a table.) Each
column in this process table represents the processing of a
specific type of crude oil and each row represents a specific input
or output stream, fuel, utility consumption, etc. For example, the
model specifies that as one barrel of Saudi light crude is pro-
cessed, a mix of 15 intermediate streams is created. The operation
consumes fuel, power, steam, and capacity, and incurs variable
operating costs of 9.3 cents per barrel.
For producing high-octane components for low-lead or unleaded
gasoline, four types of "downstream" units (i.e., units that pro-
cess outputs from the crude distillation process) are particularly
important: fluid catalytic cracking (FCC) units, isomerization
units, alkylation units, and reformers. Each of these units
enhances the octane of a different stream of intermediate products.
Finally, after all processing is complete, the refinery ends
up with numerous process output streams that are blended together
to produce final, salable refined products. This activity is
-------
11-15
TABLE II-2.
Sample Process Data Table from Refinery Model:
Yields and Operating Cost Coefficients for Crude
Distillation Unit
Saudi Arab light
Saudi Arab heavy
Mexican maya
Capacity factor
Still gas
Propane
Isobutane
Normal butane
Lt. St. run (C5-175) low oct.
Lt. st. run (C5-175) int. oct.
Lt. naph. (175-250) parf.
Lt. naph. (175-250) intm.
Naph. (250-325) parf.
Naph. (250-325) intm.
Hvy naph. (325-375) parf.
Hvy naph. (325-375) intm.
Kero. (375-500) jet fuel quality
Kero. (375-500) other
Dist. (500-620) hi sulfur
Hy gas oil (800-BTMS) (2.0% S)
Asph. , very hi sul.(4.3% S)
Crude Oil
Saudi Arab
Light Heavy
-1
-1
1.0 1.0
Yields (Fraction
0.001 0.001
0.003 0.003
0.002 0.002
0.013 0.015
0.051
0.040 0.035
0.070 0.060
0.050 0.044
0.020 0.011
0.020 0.020
0.020 0.014
0.115 0.090
0.015 0.005
0.130 0.090
0.180 0.180
0.143 0.300
Type
Mexican
Maya
-1
1.0
of Intake)
0.001
0.003
0.002
0.009
0.025
0.025
0.010
0.050
0.005
0.030
0.070
0.040
0.100
0.105
0.350
Operating Cost Coefficients
Fuel, fuel oil equivalent
Power, KWH
Steam, LB
Other var. op. cost, $
Capital charge
(Per Barrel of
-0.021 -0.022
-0.6 -0.6
-60.7 -63.4
-0.092 -0.093
varies
Throughput)
-0.020
-0.6
-57.9
-0.092
i I1^ i i v^ ^ *•* w * » ** •**• "•^•~"~ » '
fuel oil, power, steam, etc.
-------
11-16
represented in the model by product blending units. The blending
units contain quality data for all refinery streams and quality
specifications for final products. The components are then com-
bined by the model such that the qualities of the blended mixes
meet the minimum requirements of product specifications.
The refinery model can be operated in either of two modes —
minimum cost or maximum profit. It can constrain product quanti-
ties and compute a minimum cost solution. (This is useful for
analyzing the country as a whole or large refining regions in
which aggregate demands can be forecast.) Alternatively, the
simulation can vary product quantities to maximize profits at
preselected prices.
The principal reason to use computer models to simulate petro-
leum operations is to measure differences between alternative
scenarios, and thereby estimate the changes in petroleum activities
when some conditions change.
It is important to note that the procedure used does not simply
find the optimal way to make up the difference between the current
lead level and the lower level. Rather, subject to constraints
discussed later, we first run the model at the current lead
standard and find the costs of meeting that standard; then we
repeat the process, finding the cost of making gasoline at the
0.10 gplg standard. The difference between the cost of producing
gasoline under the current standard and the new standard is the
cost estimate of the rule change.
This approach minimizes the risk of error in the cost estimate,
because if the model mistakenly assumes some flexibility that the
-------
11-17
industry does not have, it would reduce the estimated cost of
meeting both the current and the new standard. The difference
between the costs is only affected if the overoptimistic assump-
tion generates greater savings in one case than in the other case,
and even then the size of the error is only the difference in
overoptimization, not the full impact of the overoptimistic assump-
tion. EPA believes that this approach to using the model adds
considerable confidence to the results.
The refinery linear programming model assumes (when uncon-
strained) that the least expensive refinery equipment is used to
meet both the current and the proposed gasoline lead standard. In
a competitive market, this is a reasonable assumption, because if
one refiner's marginal costs are lower than competitors', that
refiner would be expected to increase market share. As its market
share increased, its utilization of refinery equipment would
increase, particularly of octane-making equipment, and therefore
the refiner's marginal cost of octane would increase to the level
of its competitors. A new equilibrium would be reached with some
market share having shifted, but not necessarily a large share.
The fact that a linear program gives the same results as a competi-
tive industry had long been known and used in economics (see, for
example, Baumol, Economic Theories and Operations Analysis,
Prentice Hall, 1961).
The gasoline marketing system, which already moves millions
of barrels of gasoline per day in pipelines around the country,
helps assure that the more efficient producers can get their
product to market. In addition, the industry has a long tradition
-------
11-18
of trading components and products, which also results in a more
efficient allocation of feedstock-to-equipment. A competitive
market and an increasing marginal cost of octane for all refiners
means that the model simulates the industry reasonably accurately.
The verification tests for 1982 showed the model accurately predic-
ting refinery products and prices.
II.C. Base-Case Assumptions and Cost Estimates
Refineries are complex operations with multiple inputs and
outputs. The costs of producing any product depend in part on
the quantities of other products produced and their prices. For
example, the cost of producing gasoline depends partially on how
much residual oil is sold and its selling price. The more residual
oil is sold and the higher its price, the lower the cost of
producing gasoline. As a result of such joint costs, cost esti-
mates for a change in any one product must consider the full
range of refinery operations.
To estimate the costs of reducing lead in each year, we first
ran the DOE model in its cost-minimization mode with the constraint
that lead use not exceed 1.10 gplg (the current limit), and calcu-
lated the costs of the resulting solution. We then reran the
model at lower lead limits (varying from 0.50 gplg to zero,
depending on the year) and recalculated the costs; our cost esti-
mates are the differences between the two runs. At both 1.10
gplg and the lower levels, the model had to meet specified product
demands.
In using the refining model, it is necessary to supply it
with the values of many parameters. Section II.C.I describes the
-------
11-19
key input assumptions used in making our base-case estimates.
Section II.C.2 presents the resulting year-by-year cost estimates
based on those inputs. Section II.D describes the results of
multiple sensitivity analyses conducted to test the robustness
of our conclusions.
II.C.I. Base-Case Parameter Values
The key assumptions input to the model can be divided into
two general categories: gasoline demand and quality constraints,
and refinery operations. Each category is discussed below.
II.C.I.a. Gasoline Demand
Demand for gasoline in 1983 was about 6.6 million barrels
per day, or a little over 100 billion gallons per year. To project
gasoline demand in future years, we examined a 1983 study by Data
Resources Incorporated (DRI) and projections from DOE's Energy
Information Administration (1983), and adjusted them upwards to
reflect smaller projected gains in fuel economy.* We project that
gasoline demand will fall to about 6.5 million barrels per day in
1988, because of expected improvements in the average fuel effi-
ciency of vehicles in use, but then will level off in 1990 as
increases in vehicle miles travelled balance improvements in
miles per gallon. Since we made our original projections, the
DOE projections have been revised upward to about our level
(U.S. DOE, 1984 Annual Report to Congress).
* Compliance with the rule is harder and more costly with
higher gasoline demand.
-------
11-20
We projected the split in demand between leaded and unleaded
gasoline in two ways. First, we fit linear and logistic regres-
sions to the monthly leaded-unleaded split documented in the
Monthly Energy Review (from the Department of Energy), using time
as the explanatory variable. We also regressed the thirteen-month
moving average, to remove seasonal and random variation. These
all suggested an unleaded share of 67.5 percent (plus or minus 0.6
percent at a 95 percent confidence level) of the gasoline market
in 1988. Our vehicle fleet model (described in Appendix B), using
historic scrappage rates from DRI, predicted essentially the same
unleaded share.
We also had to estimate what portion of leaded demand is due
to "misfueling", as we expect the rule to have a major impact on
the misuse of leaded gasoline in vehicles equipped with catalysts.
Deliberate misfueling (i.e., the use of leaded gasoline in vehicles
designed for unleaded) appears to occur because leaded gasoline is
both cheaper and higher in octane than unleaded gasoline. Several
EPA and private studies have showed that widespread misfueling has
slowed the decline in lead emissions significantly, and challenged
the assumption that leaded gasoline would soon be eliminated
because of lack of demand. According to a 1983 survey by EPA
(U.S. EPA, 1984e), the current misfueling rate of light-duty
vehicles designed to use unleaded gasoline is about 15.5 percent.
(This figure was obtained after making certain adjustments; see
Chapter VI for more details.) The 1982 survey found a 13.5 percent
misfueling rate (U.S. EPA, 1983a). A panel study (Energy and
Environmental Analysis, 1984) in which consumers kept diaries of
-------
11-21
their gasoline purchases indicates that raisfueling is even more
widespread, though many people misfuel only occasionally.
We used the 1983 survey data to estimate misfueling rates by
age of vehicle (the older the vehicle, the higher the misfueling
rate) and by whether or not the vehicle was in an area with an
Inspection and Maintenance (I/M) program (vehicles in I/M areas
misfuel less). We also adjusted for the fact that older vehicles
travel fewer miles per year, on average. These adjusted rates were
then applied to our fleet model (described in Appendix B) to esti-
mate the demand for leaded gasoline by misfuelers. Our estimates
of misfueling did not influence our estimate of total demand for
leaded gasoline (which was based on a statistical analysis of
historical trends, as discussed earlier), only the split in
leaded demand between misfuelers and legitimate users.
The model requires further disaggregation of unleaded demand
into the demand for premium and regular. In our base case we
assumed that the demand for premium unleaded would comprise 25
percent of the total demand for unleaded gasoline. This is up
slightly from the current level of 24 percent.
Table II-3 presents our year-by-year estimates of gasoline
demand, both with and without misfueling. Note that we project
that total demand for leaded gasoline will decrease, but misfuel-
ers1 demand (in the absence of further regulation) will increase
in both absolute and proportional terms.
Refining costs depend on the octane level of gasoline, as well
as the split between leaded and unleaded. For each of the three
-------
11-22
grades of gasoline (leaded, unleaded regular, and unleaded pre-
mium) , we assumed that octane levels would remain at their current
averages, which are slightly higher than 87 octane for unleaded
regular, slightly higher than 91 octane for unleaded premium,
and a little above 89 octane for leaded gasoline. These octane
ratings reflect the average of two measures, "motor octane" and
"research octane." We constrained the model to meet or exceed
all three octane measures in current gasoline.
For 1986 and subsequent years, we assumed that the regulations
would reduce misfueling. Tn these runs we assumed that half of
the misfuelers do so strictly for price, but that the others
also want the higher octane provided by leaded regular. That
is, we assumed that of the potential misfuelers deterred by the
price change resulting from the rule, 50 percent would switch
to regular unleaded and the other 50 percent would require unleaded
gasoline averaging 89 octane. (The latter could be either an
intermediate grade of unleaded, or a mixture of regular and
premium unleaded.) Thus, we assumed that deterred misfuelers
would demand 88 octane unleaded, on average. Because 0.10 grams
of lead only adds about one octane number to a gallon of gasoline,
89 octane leaded gasoline must start with an unleaded gasoline
with a "clear-pool" octane rating rating slightly above 88.
Therefore, our cost estimates at 0.10 gplg are not very sensitive
to assumptions about the amount of misfueling eliminated;
indeed, the costs rise slightly with the misfueling rate.
We also had to specify the split between domestic and im-
ported gasoline supply. Over 4 percent of gasoline is now imported
-------
11-23
TABLE II-3. Year-by-Year Estimates of Gasoline Demand
Rule
With Misfuelina
Leaded
Unleaded
Without Misfuelinq
Leaded
Unleaded
1985
40
60
32
68
.2
.3
.1
.5
1986
37.5
62.8
28.8
71.5
1987
34
65
25
74
.9
.1
.6
.4
1988
32
67
22
77
.4
.3
.4
.2
1989
29.8
69.5
19.2
80.1
1990
27.6
71.4
16.4
82.6
1991
25.2
73.8
14.9
84.1
1992
24.3
74.7
13.4
85.6
Total 100.6 100.3 100.0 99.6 99.3 99.0 99.0 99.0
-------
11-24
and over 60 percent of imports are unleaded, but our base case
constrained the model to meet all demand for gasoline using
domestic refineries. (For modeling purposes, "domestic refineries"
include those in Puerto Rico and the Virgin Islands.)
We also constrained the model to meet other important charac-
teristics of current gasoline, including its Reid Vapor Pressure
(RVP). Because certain high-octane blending components (such as
butane) and additives (such as alcohols) have high RVPs, this
constraint raised the estimated cost of meeting the rulo.
In addition to gasoline demand, we also had to specify
demands for other products. These were based on a study done
for DOE's Energy Information Administration (Decision Analysis
Corporation, 1983). In most cases, we constrained the model
to produce these quantities domestically. In a few cases, where
there are wellestablished and large international markets (e.g.,
for residual oil), we allowed the model to slightly increase
\
or decrease imports at fixed prices if that was the lowest-cost
solution.
Il.C.l.b. Assumptions About Refinery Operations
The DOE model requires that many individual parameters be
specified for refinery operations. Here we discuss only the most
important of those assumptions.
We based our estimates of available capacity on published
tabulations of existing equipment and announced projects that will
be completed in the relevant years. Table II-4 presents the
estimates of available "stream-day" capacity used in the model.
For such capacity, we did not include a capital charge for its use
-------
11-25
in meeting tightened lead rules because the cost of that equipment
is already "sunk"; it does not vary with the standard. In our
estimates for 1985, 1986, and 1987, we constrained the model to
use this existing capacity because it takes two to three years for
refiners to plan, obtain permits for, and complete major construc-
tion projects. (There are "debottlenecking" and upgrading actions
that refiners can take to improve their ability to make octane in
as little as six months, other steps that might take a year, etc.
We have ignored such changes and, therefore, somewhat underesti-
mated industry capacity.) For the later years, however, we
allowed the model to determine whether construction of additional
capital equipment was economical. In those later years, the
model predicted that costs would be slightly lower if new equip-
ment (primariy isomerization units) were added. This new equip-
ment, added in response to the rule, was included in the cost
estimates using a real capital charge rate of 15 percent (which
corresponds to a discount rate of 10 percent and an equipment
lifetime of just under 10 years).
Because of maintenance and breakdowns, equipment cannot be
used to its full daily capacity, called "stream-day" capacity in
the refining industry. To account for this fact, we limited capac-
ity utilization to 90 percent. This figure reflects tight, but
feasible, operating conditions, based on industry practice and the
judgment of EPA's expert consultants on the refining industry
(Sobotka and Company). This judgment was also reinforced in a
letter to the National Petroleum Refiners Association from one of
its consultants, who suggested that reformers (a key piece of
-------
11-26
TABLE II-4. Estimated U.S. Refinery Processing Unit Capacities
for 1988 (thousands of barrels per day)
Processing Unit Capacity
Crude distillation 15,900
Vacuum distillation 6,880
Coking 1,340
Visbreaking 170
Low pressure reforming (150 psig) 940
Medium pressure reforming (250 psig) 2,280
High pressure reforming (450 psig) 750
Catalytic cracking 5,340
Hydrocracking 980
Alkylation 960
Polymerization 78
Butane isomerization 56
Pen./Hex. Isomerization 206
-------
11-27 ^
downstream equipment in meeting low-lead requirements) could be
operated at 92 percent of capacity (Soloman and Associates, 1984,
in Docket EN-84-05). Although no data are publicly available on
the utilization of individual pieces of equipment, it is important
to note that during the last four years of the 1960s, the refining
industry used more than 90 percent of its crude capacity; in 1966,
the utilization rate was 91.8 percent.
In addition to specifying maximum utilization rates for all
equipment, we also specified minimum utilization rates for certain
relatively inefficient reformers. Earlier versions of the DOE
model included two types of reformers. In response to criticisms
of that assumption, we modified the model to distinguish among
three types of reformers: low-pressure (the most modern and
efficient), medium-pressure, and high-pressure (the oldest and
least efficient). When unconstrained, the model uses the available
capacity of the low-pressure reformers first, then the medium-
pressure reformers, and then, only if more reforming is needed,
does it use any high-pressure reformers.
Some refineries, however, do not have any medium- or low-
pressure reformers. To account for this fact, we determined how
many refineries have only high-pressure reformers and required
that the model use 75 percent of their capacity. This is a very
strong assumption, as it is tantamount to saying that those
refiners neither lose market share nor purchase such components
from other refiners. Together with that requirement, we also
assumed that refineries with both high-pressure and better-quality
reformers would use the less-efficient reformers to some limited
-------
11-28
extent, and constrained the model to direct at least 270,000
barrels per day to high-pressure reformers; that works out to a
minimum capacity utilization rate for such reformers of 36 percent.
Because the model is more tightly constrained by the availability
of reformer feed than by reformer capacity, this minimum utiliza-
tion constraint increases the costs and makes the low lead levels
more difficult to achieve. In most runs, for both 1.10 gplg and
lower levels, the model used the maximum amount of low-pressure
capacity, the minimum amount of high-pressure capacity, and all
of the swing took place in the utilization of medium-pressure
reformers.
Several more detailed input assumptions also are worthy of
note. In the Preliminary RIA (U.S. EPA, 1984f) issued at proposal,
we assumed that refiners would continue to use the current mix of
catalyst grades in their FCC units. During the comment period, EPA
received information from commenters about some new, more efficient
catalysts now in use at some refineries. After checking with some
catalyst manufacturers (W.R. Grace, 1984, and Union Carbide,
personal communication) to assure that capacity is available to
produce sufficient quantities of these new catalysts, we assumed
that 25 percent of FCC capacity could use the new catalysts by
the second half of 1985, and that 50 percent could use them by
1986; to be conservative, we doubled the selling price to account
for the cost of replacing existing catalysts early.
The DOE model distinguishes among three types of naphtha
produced by the crude distillation process; different types of
crude produce different proportions of these naphthas. In the
-------
11-29
Preliminary RIA, we allowed the model to allocate those naphthas
optimally for further processing in isomerization units or for
direct blending into gasoline. In practice, such optimal alloca-
tion is likely to be difficult, although it can be approached by
segregating the storage of different naphthas, and through careful
purchases of different types of crude by refineries with different
isomerization units. To limit this optimization, our base case
now forces the model to process a preselected mix of these naphthas
that achieves only 20 percent of the gain possible with optimal
segregation. In response to several comments, we also reduced the
reformer yields with parafinic naphthas at higher severity levels.
In some runs, the model found that the most economical method
of meeting demand with less lead included altering the mix of
crude inputs, switching from heavy crudes (such as those from
Saudi Arabia) to lighter, high-quality crudes (such as those from
Nigeria), despite the fact that we include a price differential
between such crudes. To account for possible rigidities in long-
term contracts and the like, we constrained the model not to
reduce its use of heavy crude in going from 1.10 gplg to a lower
level.
The model also found it more economical to increase the use
of alcohols in gasoline in some cases, despite the fact that we
forced the model to increase gasoline production when alcohol was
used (alcohol has a lower BTU content than other gasoline compo-
nents and, therefore, delivers fewer miles per gallon). Gasoline
containing alcohol, however, faces uncertain consumer acceptance,
and many pipelines will not accept it because of possible moisture
-------
11-30
problems. Thus, we constrained the model to use no more alcohol
at low lead levels than it used at 1.10 gplg. Because of scheduled
expansion in ethanol capacity we allowed the model to use somewhat
more ethanol at either level than is currently used, but oxinol
levels were kept at current usage, and no MTBE use was allowed,
despite current usage of about 13,000 barrels per day for octane
enhancement of unleaded gasoline.
In addition to limiting the model's use of alcohol to boost
octane at low lead levels, we did not allow it to use any MMT,
another octane enhancer. During the mid-1970s, refiners used MMT
to increase octane in unleaded gasoline. After tests showed that
MMT harmed catalysts, such use was banned. MMT is still legal in
leaded gasoline, however, and some unknown amount currently is
used. Nonetheless, to be conservative, our base case did not
include any possible use of MMT to increase the octane of low-lead
gasoline.
One additional change has been made in our cost estimates.
When EPA began analyzing potential changes in its lead regulation
in late 1983, it assumed that oil prices would remain constant in
1983 dollars. That is, they were assumed to increase with infla-
tion. These price assumptions were kept in the March 1984 Cost-
Benefit Study, Prelimary Regulatory Impact Analysis, and subse-
quent docket submissions. In fact, however, oil prices, far from
increasing by a few dollars per barrel since 1983, have fallen by
several dollars per barrel, even in nominal terms. Table II-5
shows the original price estimates for Nigerian light and Saudi
heavy crude used in our model runs, the prices at the end of
-------
11-31
TABLE II-5. Prices of Crude Oil and Petroleum Products in 1983
and 1985 (dollars per barrel)
Product
Crude Oil
Nigerian Crude
Saudi Heavy
Residuel Oil
0.3% Sulfur
0.5% Sulfur
1% Sulfur
2% Suffur
Other Products
Isobutane
Normal Butane
LPG
1983
Price
31.50
28.00
28.00
27.00
26.00
24.75
26.50
25.70
18.20
1984
Current
Dollars
29.18
28.60
28.00
27.43
26.76
26.18
24.50
23.75
17.78
Price
1983
Dollars
27.40
26.88
26.29
25.76
25.13
24.58
23.00
22.30
16.70
-------
11-32
1984, and what those prices are equivalent to in constant 1983
dollars. It also shows those same prices for residual oil,
butanes, and propane.
Product prices are based on about a six-month period in
1984 to avoid seasonal distortions. The differences between the
first and last columns show that this assumption has seriously
distorted our cost estimates.
Rather than rerun all of the refinery analyses to find the
costs under these lower prices, we have adopted the following
approach. We have assumed that oil prices stopped falling in
December 1984 and will henceforth increase with inflation. We
have repriced the cost estimates of our model runs using the new
prices, assuming (to be conservative) that the costs of crude
fell by the same percentage drop as the heavier crudes, not the
lighter crudes. To test the reliability of this approach, we
reran the model for 1986 and 1988 with the new prices in, and
compared it to the results of our repricing. These costs should
be somewhat lower, because, in response to the change in the
relative prices of crudes and products, refiners may find it less
expensive to change some of their operating procedures. As Table
II-6 shows, the costs using our procedure are quite close, but
slightly higher than the costs given by completely rerunning the
model. We also checked the marginal cost estimates for some of
the more highly constrained sensitivity analyses, and achieved
close agreement as well.
Because the British coal strike may have been keeping up the
price of high sulfur residual oil on the world market, we also
-------
11-33
TABLE II-6. Cost of 0.10 gplg Standard with New Oil Prices
New Model Run versus Repricing, Assuming No
Misfueling (millions of dollars)
Year
1986
1988
New Model Run
573
502
Repricing
608
531
-------
11-34
recalculated the costs assuming that high sulfur residual oil was
selling for $1 per barrel less than current prices. This only
increased the cost of our 1986 case by $14 million, so these esti-
mates are not very sensitive to that potential factor.
II.C.2. Base-Case Results
Table II-7 presents information on the model's estimates at
1.10 gplg and at lower lead levels in 1985, 1986, 1987, and 1988.
(The figures for 1985 apply only to the second half of the year.)
These estimates assume that 0.50 gplg in 1985 does not eliminate
any misfueling and that 0.10 gplg in the later years reduces mis-
fueling by 80 percent. (See Chapter VITI for a discussion of
alternative assumptions about misfueling and the costs associated
with those assumptions.) In addition to the cost estimates, the
table reports several other important pieces of information: the
capacity utilization rates for various types of reformers (low-,
medium-, and high-pressure, plus the aggregate figure) and the
marginal cost of producing an octane-barrel of gasoline. (An
octane-barrel is defined as raising the octane of a barrel of
gasoline by 1 point.) Note that the 0.50 gplg standard for the
second half of 1985 has relatively little impact. The cost is
only $96 million; the overall utilization of reformers rises from
50 percent at 1.10 gplg to 59 percent at 0.50 gplg. The marginal
cost of producing an octane-barrel rises from 15.8 cents at 1.10
gplg to 20.4 cents at 0.50 gplg. Thus, refiners should be able to
meet the 0.50 gplg standard with relative ease and at a relatively
moderate marginal cost by using excess capacity.
-------
11-35
With regard to these estimates of marginal costs, it is
interesting to note the comment of one refiner (Amoco submission
to Public Docket EN-84-05) critical of EPA's cost estimates:
"Current wholesale spot market price differentials between leaded
and unleaded imply a marginal cost of about 20 cents/BON [barrel
of octane number]. The trading value of lead rights reflects a
similar cost of octane..." As shown in Table II-7, the model
with EPA's base-case assumptions projects a marginal cost of 15.8
cents per octane-barrel at 1.10 gplg in 1983 dollars, or 16.7
cents in 1984 dollars, very similar to the Amoco estimate for
current (late 1984) conditions.
The base-case results for 0.10 gplg in 1986 and 1987 show
somewhat tighter operating conditions, but no feasibility problems.
.In both years, overall reformer utilization rises to 66 or 67
percent, and the marginal cost of an octane-barrel rises to a
little more than 29 cents.
The results for 1988 are of interest because that is the
first year in which we assumed that additional capacity could be
added. The model found it most economical to add about 98,000
barrels per stream-day of isomerization capacity. That allows
slightly lower use of reforming capacity than in 1986 or 1987,
and reduces the marginal cost of octane to 28.4 cents per barrel
at 0.10 gplg. The cost of that extra isomerization capacity
($322 million), annualized at 15 percent, is included in the
$535 million cost estimate for 1988, and also in the cost estimates
for subsequent years.
-------
11-36
TABLE II-7. Base-Case Results for 1985-1988, with Partial Misfueling
Year
Reformer utilization
(%) by reformer type
Marginal
Cost of
Octane
Total
Cost
Lead
1985
1.10
0.50
1986
1.10
0.10
1987
1.10
0.10
1987
1.10
0.10
limit
gplg
gpig
gplg
gplg
gplg
gplg
gplg
gplg
Low
90
90
90
90
90
90
90
90
Med.
40
53
40
67
41
66
42
62
High
36
36
36
36
36
36
36
36
Agg.
50
59
50
67
52
66
52
64
(^/barrel)
15.
20.
16.
29.
16.
30.
17.
28.
8
4
3
2
4
0
2
4
(million $)
N.A.
96*
N.A.
608
N.A.
558
N.A.
532
rCost for quarters III and IV.
-------
11-37
The 1988 results also are interesting because they reflect
the long-run marginal cost of producing octane (including capital
charges). The reason that the marginal costs for 1.10 gplg are
lower is that the refining industry has excess octane-producing
equipment, and we have not included the costs of that sunk capital
in estimating either total or marginal costs. Note also that
the marginal cost of 0.50 gplg in 1985 (20.4 cents per octane-
barrel) is far below the long-run marginal cost, which suggests
that refiners would not find it cost-effective to build new capa-
city to meet that standard even if it were possible; the minimum-
cost solution for 1985 would not include any new equipment. In
1986 and 1987, however, at 0.10 gplg, the marginal costs slightly
exceed the long-run figure, indicating that it would be cheaper
to build new capacity, if it were possible (although it is still
possible to meet the 0.10 gplg standard without new capacity).
Table II-8 presents the year-by-year cost estimates under
the base-case assumptions. Estimates are shown for three differ-
ent rules: the original proposal (0.10 starting 1/1/86), the
illustrative alternative discussed in the Notice of Proposed
Rulemaking (0.50 gplg on 7/1/85, 0.30 on 1/1/86, 0.20 on 1/1/87,
and 0.10 on 1/1/88), and the Final Rule (0.50 gplg on 7/1/85 and
0.10 on 1/1/86). Again, these estimates assume full misfueling
in 1985 and 80 percent reductions in misfueling in subsequent
years, when the 0.10 gplg standard applies. The costs fall over
time because of projected declines in the demand for leaded
gasoline.
-------
11-38
TABLE II-8. Year-by-Year Estimates of Costs of Meeting
Alternative Rules, with Partial Misfueling
(millions of 1983 dollars)
Rule
Proposed
Alternative
Final
1985
0
96
96
1986
608
364
608
1987
558
448
558
1988
532
532
532
1989
504
504
504
1990
471
471
471
1991
444
444
444
1992
441
441
441
-------
11-39
JI.D. Sensitivity Analyses
The base-case results shown in Table II-8 represent EPA's
"best estimates" of the cost of complying with the phasedown
rule; i.e., they are based on what the Agency believes are the
most realistic assumptions about refinery capabilities and gaso-
line demands. The Agency also conducted numerous sensitivity
analyses to test the robustness of the results. Our efforts
focused on 1986, because that is the first year in which the 0.10
gplg standard will apply. As shown in Table II-7, the 0.50 gplg
standard in 1985 is much easier to meet, as measured either by
reformer utilization or by the marginal cost of an octane-barrel.
The 1987 results are similar to those for 1986, but in practice
the 0.10 gplg standard should be easier to achieve in that year
because refiners will have two years to adjust their operations.
Most of the sensitivity runs focused on changes unfavorable
to the rule; i.e., on alternative parameter values that would
increase the estimated costs. We did so not because we believe
that such values are more likely to occur than those that would
reduce the cost of the rule, but rather to probe the hypothetical
conditions under which the rule would become extremely costly or,
possibly, infeasible with existing refinery equipment. Many of
the sensitivity analyses, particularly those varying several
parameters simultaneously, responded to comments received on the
August 1984 proposal.
We performed two types of sensitivity analyses. The first
dealt with the issue of aggregation: to what extent does a national
model underestimate costs by failing to account for regional
-------
11-40
differences? We concluded that the base-case cost estimates are
not artificially low because we used a national model. The
disaggregated runs also indicated, however, that the rule would
cause tighter operating conditions in Petroleum Allocation for
Defense Districts (PADDs) I-IV and VI than in PADD V (the West
Coast, Alaska, and Hawaii), so our second set of sensitivity
analyses, designed to probe the limits of feasibility, focused
on PADDs I-IV/VI (the rest of the country).
II.D.I. Level of Aggregation
Several comments received before and after the August 1984
proposal criticized EPA's cost analysis for relying on a national
model. To examine that issue, we conducted two sets of sensitivity
analyses.
The first compared EPA's estimates to those based on an
analysis performed by Turner-Mason Associates (TMA) (1984) for
the Lead Industries Association. TMA disaggregated the refining
industry into six groups, based on geographic location, size and
type of refinery, and other factors. It estimated the cost of
meeting the 0.10 gplg in 1988 to be $995 million, compared to
EPA's estimate at the time of proposal of $503 million. In
addition to disaggregating, however, TMA changed several other
assumptions. To see to what extent those changes, rather than
the level of aggregation, explained the difference in results,
we incrementally made each of those changes in the inputs to the
national DOE model. The end result was that our national model
with all of the TMA assumptions predicted slightly higher costs
-------
11-41
than the TMA analysis, $1,016 million vs. $995 million (September
19 r "Supplemental Analysis of Refining Costs," submission to Public
Docket EN-84-05). This result indicated that different levels of
aggregation were not the source of discrepancy between EPA's and
TMA's estimates, and that EPA's model accurately reproduced the
results obtained with TMA's substantially finer level of disaggre-
gation.
Note that the base-case estimates in this Final RIA incor-
porate several of the more pessimistic assumptions used by TMA,
in particular a reduction in reformer yields at high severities.
Other assumptions made in the TMA analysis, however, do not
appear reasonable, in particular the capital charge rate (27
percent real) and a reduction in butane use of 50 percent from
its current level.
The second approach to the aggregation issue divided the
refining industry into two units, PADD V (the West Coast) and
PADDs I-IV (the rest of the continental U.S.) plus PADD VI (the
Virgin Islands and Puerto Rico). The rationale for this analysis
was that, while PADDs I-IV/VI are tightly interconnected by pipe-
lines and water shipping, PADD V is relatively isolated; only a
few small pipelines connect PADD V to the rest of the country,
and transport by sea between the Gulf and the West Coast is
relatively expensive, though by no means impossible.
Table II-9 compares the national base-case estimates with
the separate estimates for the two parts of the country for 0.10
gplg in 1986. Note that the sum of the two regional estimates
is insignificantly different than the national estimate, $611
-------
11-42
million vs. $608 million, which indicates that using the national
model did not bias downward our base-case estimates. Table II-9
also reveals, however, that the rule generates somewhat higher
operating rates and marginal costs for producing octane in
refineries in PADDs I-IV/VI, which suggests that any problems
from tighter operating constraints would show up in that part of
the country before they occurred in PADD V.
It is also important to note that very fine levels of dis-
aggregation can lend a strong upward bias to the cost estimates
by failing to account for the flexibility in the refining industry
to trade products and shift production in the face of changes in
market conditions. Suppose, for example, that each refinery were
modeled separately, with the constraint that it produce exactly
the same slate of products that it produces now. That approach
would overestimate total costs, because it would not allow for
adjustments in product mixes across refineries. Such adjustments
may be quite substantial, as refineries with better octane-making
equipment would increase their total production of gasoline or
increase the proportion of production that is premium gasoline,
while other refineries would reduce their total gasoline output
(or premium gasoline share), possibly shifting to the production
of other petroleum products that place less burden on octane
enhancement.
II.D.2. Other Parameters
We conducted extensive sensitivity analyses of other key
parameter values in PADDs I-IV/VI. Table 11-10 summarizes the
-------
11-43
TABLE II-9. Costs of Meeting the 0.10 gplg Standard:
Comparison of National and Regional Results for 1986
Model
National Model
Base case
Reformer Utilization
(%) by Reformer Type
Low Med. High Agg.
Marginal
Cost of
Octane
(j/barrel)
Total
Cost
(million $)
90 67 36 67
29.2
608
Regional Models
PADDs I-IV/VI
PADD V
Total
90 71 36 69
90 60 36 60
31.6
21.0
531
_8_0
611
-------
11-44
base-case and alternative assumptions explored. As noted earlier,
we focused on adverse changes; i.e., those that would make it more
difficult to meet the standard. We also ran some sensitivity
analyses of changes that would make it easier to meet the rule
than our base case predicts.
II.D.2.a. Assumptions Varied
The first item listed in Table 11-10 is the demand for un-
leaded premium gasoline. In our base case, we assumed that the
demand for unleaded regular would be 25 percent of the total demand
for unleaded gasoline, up slightly from its current level of 24
percent. As shown in Table 11-10, we also ran sensitivity analyses
that assumed that the fraction of premium would grow. Case Al
assumes that the demand for premium would grow by about 1.5 per-
centage points per year, reaching 27 percent of total unleaded
demand in 1986 and 30 percent of the total in 1988. Case A2
assumes that premium demand grows at an even higher rate, about
5 percentage points a year, reaching 34 percent of total unleaded
demand in 1986. This high rate of growth is out of line with
historical trends and appears implausible, particularly if the
marginal cost of producing octane rises, as it will with the
phasedown.
Case B tightens the constraint on capacity utilization of
all downstream processing units (including, most importantly, FCC
and reforming units) from 90 percent of stream-day capacity to 85
percent. Several commenters suggested that EPA's assumption of
90 percent was too optimistic. Although the Agency continues to
-------
11-45
Case
Al
A2
B
C
D
El
E2
E3
Parameter/Assumption
Premium share of unleaded demand
Premium share of unleaded demand
Maximum capacity utilization of
downstream processing units
Share of FCC units using new
catalyst
Maximum ethanol use, '000 barrels/day
Summer RVP limit
Summer RVP limit and 5% higher demand
Summer RVP, summer gasoline demand, and
Base-Case
Value
25%
25%
90%
50%
60
No
No
Alternative
Value
27%
34%
85%
0%
30
Yes
Yes
summer distillate demand No
F Substitution of light for heavy crudes
permitted No
G MMT use allowed in leaded gasoline No
H Premium share of unleaded demand 25%
I Percent of naphtha segregated
for gasoline processing 20%
Yes
Yes
Yes
23%
100%
-------
11-46
believe that 90 percent utilization can be achieved, this sensi-
tivity analysis examined the impact of a lower utilization rate.
Case C assumes that all FCC units continue to use the same
catalysts that they do now; the base case, as discussed earlier,
assumes that by 1986 half of the FCC capacity would have switched
to one of the newer, more efficient catalysts now available.
Case D examines the impact of reduced alcohol use. In the
base case, we constrained the model to use no more alcohol
(ethanol) at 0.10 gplg than it used at 1.10 gplg, which was
60,000 barrels per day- For this sensitivity analysis, we limited
ethanol use to 30,000 barrels per day.
Case E analyzes only the summer quarter, when vapor pressure
constraints are tighter. This case makes one extreme assumption
about the summer quarter: that the refining industry's normal
reliance on seasonal storage of gasoline components to meet
summer demand does not occur, and that all summer demand must be
met by current production. This seems unlikely, however, because
the industry has 300 million barrels of gasoline storage capacity,
in part because it normally produces extra gasoline in the winter
and spring and stores it to meet summer demand peaks (Schwartz,
memo in Docket EN-84-05, September 11, 1984e). Case El looks at
the RVP change in isolation, while case E2 combines it with 5
percent higher daily demand for gasoline, which is typical of the
summer quarter. In both cases, distillate production is kept at
the annual average (whereas it normally is reduced in the summer
to allow for greater gasoline production), and no account is taken
of imports. Case E3 combines case E2 with a 50,000 bpd reduction
-------
11-47
in distillate. In all three cases, the cost estimates presented
later are only for one quarter, and are not comparable to cost
estimates for the other (annual-cost) runs. The other measures
of "tightness," however, can be compared.
Case F allowed the model more flexibility in choosing the
crude slate. The base case constrained the solution to use as
much heavy crude at 0.10 gplg as it did at 1.10 gplg. This sensi-
tivity analysis allowed the solution to include some substitution
of high-quality light crudes for heavy crudes if that was the
most economical approach; but even in this sensitivity analysis,
the swing was not allowed to exceed 250,000 barrels per day.
Case G allowed the model to use an alternative octane booster,
MMT, at up to 0.05 grams per gallon, but only in leaded gasoline.
That compares to the average of 0.0625 grams per gallon of MMT
used in unleaded gasoline until its use in unleaded was banned
because of its adverse effects on tailpipe hydrocarbon emissions.
The use of MMT is still permitted in leaded gasoline, although
data on current levels of use are not publicly available. The
base case makes the pessimistic assumption that no MMT will be
used.
Case H assumed lower octane demand, with the premium share
of demand for unleaded gasoline falling to 23 percent. This
could reflect several possibilities. First, if the price of
leaded rises above that of unleaded, some current legal users of
leaded gasoline (as well as misfuelers) may switch to unleaded
regular; as discussed in Chapter VII, the experiences of the
U.S. Armed Services and the U.S. Postal Service suggest that
-------
11-48
many vehicles designed to operate on leaded regular (89 octane)
will perform satisfactorily on 87 octane unleaded. During the
public hearings, several refiners also mentioned that regular
leaded gasoline in California has been reduced to 88 octane.
Second, this drop in octane could occur if sharp increases in
the marginal cost of producing octane led some refiners to slightly
reduce the octane levels of their products.
Case I altered the assumption about the ability of refineries
to segregate different types of naphthas for further processing.
This sensitivity analysis allows for greater segregation of
naphthas, as was assumed in the estimates made for the Preliminary
RIA.
II.D.2.b. Results of Sensitivity Analyses
Table 11-11 presents the results of varying the parameter
values individually. In all cases, the estimates are for 0.10
gplg in 1986. Note that these results cover only PADDs I-IV/VI,
the part of the country where difficulties, if they occur, are
likely to show up first. The first line shows the results for
PADDs I-IV/VI under the base case assumptions.
As the other lines in the table show, the impacts of varying
the individual assumptions are modest. The maximum cost increase
is less than 19 percent, and that occurs only under the higher of
the two high-octane scenarios. That scenario assumes that the
premium share of unleaded demand increases at 5 percentage points
each year (from a base of 24 percent in 1984), which seems highly
unlikely. Indeed, that rate of growth in the premium share is so
-------
11-49
TABLE ll-ll. Effects of Varying Individual Parameters/Assumptions:
Case
Base
Al
A2
B
C
D
El
E2
E3
F
G
H
I
I
1
Changes I
None
27% premium
35% premium
85% utilization
Old catalysts
Reduced alcohol
Summer RVP
El , summer demand
E2, summer distillate
Crude flexibility
Use of MMT
23% premium
Naphtha segregation
Reformer utilization
[%) by reformer type
jOW
90
90
90
85
90
90
90
90
90
90
90
90
90
Med.
71
72
82
84
87
77
77
90
90
70
69
70
66
High
36
36
36
36
36
36
36
36
36
36
36
36
36
Agg.
69
69
74
74
77
72
72
78
78
68
68
68
66
Marginal
Cost of Total
Octane Cost
(jzf/barrel) ($ million)
31.6
32.8
33.0
29.2
32.5
33.0
33.2
34.4
31.6
31.6
29.0
30.3
25.2
531
541
629
486
617
554
135*
136*
127*
513
489
496
441
*Cost for summer quarter only
-------
11-50
high that it implies that the absolute amount of unleaded regular
gasoline would fall 6 percent from 1984 to 1986, a highly improb-
able occurence.
Looking at the other measures of tightness, the most severe
impacts occur under case E2, which assumes summer RVP and 5
percent higher demand with no distillate adjustment, seasonal
storage, or imports. Even in that case, however, the marginal
cost of producing octane rises less than 9 percent (to 34 cents
per octane-barrel), and utilization of high-pressure reformers
does not rise above the minimum level that we forced the model
to use. Overall reformer utilization was only 78 percent of
capacity, up from 68 percent in the base case. These increases,
of course, would be balanced by lower-than-average operating
costs, marginal costs, and reformer severities in the other three
quarters when RVP standards are higher than the annual average.
Case E3 shows that adding even small distillate flexibility brings
the marginal cost of octane back down to the base case value.
Case B, lowering maximum downstream utilization from 90
percent to 85 percent, yielded counter-intuitive results; the cost
is slightly lower than in the base case, as is the marginal cost
of octane. The utilization of the medium-pressure reformers
increased in this case, from 71 percent of stream-day capacity to
84 percent. Closer examination of the output of the model revealed
that 85 percent utilization raised the cost of both the 1.10 gplg
and the 0.10 gplg cases, as expected, but reduced the difference
between the two. Tightening that constraint appears to have raised
the total cost of producing octane, but lowered its incremental
-------
11-51
cost over the relevant range.
On the positive side of the ledger, the most important
parameter change was case I, which allowed more optimal segrega-
tion of naphthas, as in the runs made for the Preliminary RIA;
that reduced the costs in PADDs I-IV/VI from the base case by 17
percent, to $441 million. It also reduced overall reformer
utilization from 68 to 66 percent of capacity, and had a more
dramatic impact on the marginal cost of producing octane, reduc-
ing it by 20 percent, to 25.2 cents/barrel. Allowing the use of
MMT in leaded gasoline, case G, also had a significant impact on
costs.
In addition, we ran sensitivity analyses that varied several
parameter values simultaneously to probe the limits of feasibility;
the results are shown in Table 11-12. They should be interpreted
carefully, for while each parameter change may be plausible alone,
increasing the number of simultaneous negative changes generates
increasingly implausible circumstances. Furthermore, market
forces of supply and demand create feedback loops that make some
of the extreme cases exceedingly unlikely. For example, if the
marginal cost of producing octane rises (due, say, to lower
utilization of downstream capacity), the price of high-octane
unleaded premium is likely to rise, reducing the demand for that
product (or at least forestalling the increases assumed in cases
Al and A2) and the demand for high-octane alcohol is likely to
increase, not decrease.
Lowering the utilization rate for downstream equipment and
simultaneously reducing the allowable amount of alcohol, run Ml
-------
11-52
TABLE 11-12. Effects of Varying Multiple Parameters/Assumptions:
PADDs I-IV/VI
Run
Base
Ml
M2
M2a
M3
M3a
M3b
M4
M4a
M5
M5a
M6
M6a
Changes**
None
Utilization (B) ,
less alcohol (D)
(B), (D), plus
old catalysts (C)
(B),(D),(C), plus
lower premium (I)
(B) ,(C) ,(D), plus
higher premium (Al
(B),(C),(D),(A1),
plus MMT (G)
(B)f(C),(D),(Al),
plus MTBE and
oxinol
Old catalysts (C) ,
higher premium (Al
less alcohol (D)
(C),(A1) ,(D),
plus MMT (G)
Old catalysts (C) ,
less alcohol (D) ,
summer RVP (El)
(C) ,(D) ,(E1), plus
lower premium (I)
(C) ,(D) ,(E1), plus
higher premium (Al
(C) ,(D),(E1) ,(A1),
plus MMT (H)
Reformer Utilization
(%) by Reformer Type
Low
90
85
85
85
) 85
85
85
),
90
90
90
90
) 90
90
Med.
71
85
85
85
85
85
85
90
90
90
90
90
90
High
36
38
78
70
78
68
61
40
36
58
58
62
64
Agg.
69
75
83
82
83
81
80
79
78
83
83
84
83
Marginal
Cost of
Octane
(jd/barrel)
31.6
33.4
59.4
52.0
61.0
49.4
48.3
49.4
41.4
54.2
51.3
85.4
48.3
Total
Cost
($ million)
531
524
775
754
830
717
773
763
655
217*
202*
231*
197*
(Table 11-12 continues on next page)
-------
11-53
TABLE 11-12. fP»nfrin..o,l)
Reformer Utilization
(%) by Reformer Type
Marginal
Cost of Total
Octane Cost
Run Chanaes** T.^w M-3^ - Hinh AqqT (//barrel)
M7
M7a
M7b
M7c
M7d
(C) ,(D) ,(A1), plus
summer RVP and
demand (E2) 90 90 89 90 87.0
(C),
plus
(C) ,
plus
(C),
(G),
(C),
MTBE
(D),(A1),(E2),
MMT (G) 90 90 86 89 62.8
(D) ,(A1) ,(E2),
storage 90 90 52 82 54.5
(D),(A1) ,(E2),
plus imports 90 90 55 83 48.2
(D),(A1),(E2),
, oxinol, and
summer distillate 90 90 76 87 48.2
($ million)
267*
220*
215*
193*
218*
*Cost for summer quarter only
**Letters refer to parameters listed in Table 11-10
-------
11-54
in Table 11-12, has little impact on any of the measures. Also
eliminating the use of any new FCC catalysts, as in run M2, makes
for substantially tighter, though still feasible, operating
conditions, raising the total cost to $775 million and the mar-
ginal cost to 59.4 cents/octane-barrel. Run M2a made the same
negative changes as run M2, but lowered the demand for premium
unleaded to 23 percent (2 percentage points below the base case,
but only 1 percentage point below current demand); compared to
run M2, that cut the cost about 3 percent, and reduced the marginal
cost of octane by about 12 percent.
Run M3 added higher demand for premium to the changes made
in run M2. That further tightened the predicted operating condi-
tions, but did not make the attainment of the 0.10 gplg standard
infeasible. Run M3a made those same changes, but allowed the
use of MMT, which cut total and marginal costs significantly.
Run M3b also duplicated M3, but allowed additional purchases of
oxinol (10,000 bpd) and MTBE (27,000 bpd) at prices that reflect-
ed both their cost and their adverse impact on fuel economy.
It also cut total and marginal costs substantially. Both of
these increases are within existing capacity for those products.
Run M4 probed a slightly different combination of sensitivi-
ties, combining low ethanol use, high octane demand, and old
catalysts (as well as no imports, no MTBE, etc.). Again, costs
increased but the industry would still be able to comply. Once
more, adding some MMT (run M4a) substantially reduced costs.
Runs M5-M7 were summer sensitivity runs. Again, to probe the
-------
11-55
limits of feasibility, we began by assuming no seasonal storage
was available. Run M5 assumed that no new catalysts would be
used, that alcohol use could not exceed 30,000 barrels per day,
and that all gasoline would have to meet summertime RVP specifica-
tions with no distillate flexibility, imports, or MTBE use.
This is a very demanding, and unlikely, set of events, but 0.10
gplg remained feasible. Lowering the demand for premium (run
M5a) brought the total and marginal costs down significantly;
storage of high octane components produced during the first half
of the year (as is done by most refiners now) would be one way
to reduce the amount of premium gasoline that had to be produced
in the summer. The cost estimates for these two runs (and the
subsequent ones in the table) are only for the summer quarter.
Again, the higher-than-average costs in the summer (when RVP
standards are tighter than average) are balanced by lower-than-
average costs in the winter (when RVP standards are looser than
average) .
Run M6, which added higher demand for unleaded premium to
run M5, generated a model solution that must be characterized as
infeasible. The marginal cost of producing octane was extraordi-
narily high, and the model ran so much extra crude oil that some
of the products of crude distillation had no outlet; they were
simply dumped. The conditions assumed in that run, however, are
collectively highly implausible. If the marginal cost of octane
were to rise to the level shown in run M6, the price of high-octane
unleaded premium would rise sharply, making the surge in unleaded
premium demand and lack of use of existing MTBE and oxinol capacity
-------
11-56
contemplated in run M6 impossible. Moreover, because run M6
only applies to summer conditions, normal use of existing storage
capacity could alleviate the summertime crunch, since the same
case (M4) with annual average RVP levels was feasible. It is
also interesting to note that modest use of MMT, as allowed in
run M6a, restored feasibility and brought the estimated marginal
cost down to a much more comfortable level.
Run M7 added higher overall gasoline demand to the changes
in run M6; again, the result was infeasibility, though for an
even more implausible set of conditions. Again, storage of
gasoline or high-octane gasoline components produced in the
winter or spring could alleviate this summertime problem since M4
is the same run for the full year. Relaxing some of our least-
likely conservative assumptions in other areas also restored
feasibility. For example, run M7a shows that MMT restored fea-
sibility, bringing the marginal cost of an octane-barrel back
down to 62.8 cents. Run M7b shows that 400,000 barrels per day
of seasonal storage also restored feasibility, reducing the
marginal cost of octane to 54.5 cents, and run M7c shows that
taking account of imports and MMT (but not storage) reduced
costs even further. Run M7d shows the results of allowing the
normal seasonal distillate flexibility, use of existing MTBE
capacity, and use of 10,000 bpd more of existing oxinol capacity,
but neither imports nor seasonal storage. Again, feasibility is
restored.
To summarize the summer sensitivity runs, it takes an
extremely unlikely combination of high-cost sensitivities (e.g.,
-------
11-57
higher premium demand, no use of the newer catalysts, lower use
of ethanol, summer RVP standards, high total demand, no imports,
no MTBE, no MMT, no increase in oxinol, no seasonal storage, no
distillate flexibility, and no trading with PADD V) to produce
an infeasible result. Relaxing the assumption about seasonal
storage, within the limits of existing storage capacity, always
restores feasibility (case M4), as does relaxing many of the
other assumptions singly (e.g., the assumption on MMT, on dis-
tillate flexibility, or on using existing MTBE capacity). Since
the industry currently manipulates to its advantage the use of
storage capacity, MMT, MTBE, swings in distillate production,
and imports simultaneously, EPA believes that concern about
summer infeasibility problems is unwarranted.
We also ran the two most extreme sensitivity analyses (runs
M3 and M7) on PADD V to check our assumption that problems, if
they occured at all, would show up first in PADDs I-IV/VI; Table
11-13 reports the results. As a comparison with the correspond-
ing runs in Table 11-12 shows, conditions in PADD V are projected
to be looser (as measured by reformer utilization and the marginal
cost of producing an octane barrel) than in PADDs I-IV/VI.
To assure ourselves that the 0.50 gplg standard for the
second half of 1985 would be feasible, even under worst-case
conditions, we ran the two most extreme sensitivity analyses,
shown in Table 11-14. Even in the worst case examined (M7),
overall reformer utilization did not exceed 75 percent and the
marginal cost of producing an octane-barrel did not rise above
24 cents. The results show that 0.50 gplg in 1985 remains quite
-------
11-58
TABLE 11-13. Sensitivity Analyses for 1986: PADD V
Run
Changes
**
Reformer Utilization
(%) by Reformer Type
Low Med. High Agg.
Marginal
Cost of Total
Octane Cost
(^/barrel) ($ million)
Base None 90
M3 Utilization (B),
less alcohol (D),
catalysts (C),
higher premium (Al) 85
M7 Old catalysts (C),
less alcohol (D),
higher premium (M),
summer RVP and
demand (E2) 90
58 36 59
70 34 68
77 36 76
21.5
29.8
32.2
*Cost for summer quarter only
**Letters refer to parameters listed in Table 11-10.
80
107
31'
-------
11-59
TABLE 11-14. Sensitivity Analyses for 1985; PADDs I-IV/VI
Run Charges**
Base None
M3 Utilization (B),
higher premium (Al)
less alcohol (D)
old catalysts (C)
M7 Old catalysts (C),
less alcohol (D),
higher premium (Al)
summer RVP and
demand (E2)
Marginal
Reformer Utilization Cost of Total
(%) by Reformer Type Octane Cost
Low Med. High Agg. (^/barrel) ($million)
90
90
55
85 78
84
36
36
36
60
71
75
21
22.3
23.4
79
89
48'
*Cost for summer quarter only
**Letters refer to parameters listed in Table 11-10,
-------
11-60
feasible and not excessively costly under all circumstances.
These sensitivity analyses clearly indicate that the phase-
down schedule contained in the rule is feasible with existing
equipment under expected conditions. Even if several conditions
are more adverse, simultaneously, than in our base case, the
0.10 gplg standard remains feasible, although it may require
very efficient utilization of refineries and careful attention
to the full array of methods available for meeting octane demands
with reduced lead. Only in the very worst cases, which combine
many adverse conditions simultaneously, does it appear that the
refining industry would experience great difficulty in complying
with the rule and then only for three months. Such multiple
worst-case scenarios are useful for probing the limits of feasi-
bility, but they are too implausible to deserve much weight, and
certainly too implausible to affect the outcome of this rule-
making. Moreover, as discussed in the next section, the proposed
rule to allow banking would render even those circumstances
feasible.
II.E. Impact of Banking on Costs
EPA has proposed to allow the "banking" of lead in 1985 for
use in 1986 and 1987 (50 FR 718; January 4, 1985), when the 0.10
gplg standard will apply. Under this proposal, refiners will
have the option of reducing lead use before it is required, and
then applying those early reductions to increase the amounts of
lead they are allowed to use in the two succeeding years. The
purpose of this provision is to give individual refiners lower
-------
11-61
costs and extra flexibility in reducing lead (even though the
standards are both feasible and reasonable without it) without
increasing the total amount of lead used in gasoline between
1985 and 1987. Although refiners would be under no obligation
to avail themselves of the right to bank, EPA expects that the
majority would, for at least three reasons: (1) lead provides a
greater octane boost at low levels, so it is more valuable in
1986 and 1987 at 0.10 gplg than at the higher levels permitted
in 1985; (2) in the short run, the marginal cost of producing
octane rises with the amount of octane produced, again making
lead more valuable in 1986 and 1987 than in 1985; and (3) banking
lead rights in 1985 will give refiners extra flexibility to deal
with unexpected problems (such as equipment breakdowns) in the
later years.
II.E.I. Base-Case Banking Results
It is difficult to predict precisely how banking will be
used. Some refiners are likely to cut their lead use vory quickly
so as to bank a large amount, either for later use at that refinery
or for sale to other refineries less well-equipped to produce
octane without lead, while others will bank relatively little.
Whatever the specific pattern followed, however, the experience
with interrefinery averaging ("trading") over the last few years
suggests that the refining industry will make effective use of
this mechanism to reduce costs by reallocating lead use to those
refineries (and, with banking, to those times) where lead has
the highest marginal value in reducing production costs. (As
-------
11-62
discussed in Chapter I, refiners representing about three-fourths
of all refining capacity now trade in any given quarter.)
To explore the potential impacts of banking, we examined two
possible phasedown patterns. For this analysis, we assumed (as
above) that misfueling would continue unabated during 1985, when
the standard will be 1.10 gplg during the first half of the year
and 0.50 gplg during the second half of the year. For 1986 and
1987, when the limit will be 0.10 gplg, we assumed that misfueling
would fall to 20 percent of its current level. We also assumed
that misfueling would follow this same pattern with banking. If
refiners use less than 1.10 gplg in the first half of 1985, or
less than 0.50 gplg in the second half, the marginal cost of
producing leaded gasoline will increase. Each gallon of leaded
gasoline produced below the limit will produce banked rights
that have value, so the net marginal cost of producing a gallon
of leaded gasoline will be little affected. Similarly, in 1986
and 1987, refiners who use banked rights to produce leaded gaso-
line with more than 0.10 gplg will face lower marginal costs of
production than they would have otherwise, but they also will
consume rights that have a market value, so the net marginal
cost will be higher, roughly the same as it would have been at
0.10 gplg. As a result, it seems reasonable to make the same
misfueling assumptions with and without banking.
Table 11-15 compares the schedule without banking to two
possible alternatives with banking. Alternative 1 assumes that
refiners do not start banking until the second quarter of 1985,
at which point they use an average of 0.60 gplg, thus banking
-------
11-63
TABLE 11-15. Alternative Phasedown Patterns with Banking (gplg)
1985 (by quarter) 1986 1987
Alternative I II 111-IV
Without Banking 1.10 1.10 0.50 0.10 0.10
With Banking
Alternative 1 1.10 0.60 0.40 0.25 0.19
Alternative 2 0.80 0.60 0.45 0.30 0.21
-------
11-64
0.50 grams, on average, for each gallon of leaded produced.
(Note that this industry average could reflect wide variations
across refiners; some might not bank at all, while others saved
large amounts.) Under alternative 1, we also assume some banking
in the last half of the year, with leaded gasoline averaging 0.40
gplg, slightly below the limit of 0.50 gplg. A total of 7.0
billion grams of lead (about 22 percent of the total allowed in
1985) would be banked during 1985, allowing refiners to average
0.25 gplg in 1986 and 0.19 gplg in 1987. Shaving 0.10 gplg from
the annual average in 1985 translates into a larger per-gallon
increase in 1986 or 1987, because the amount of leaded gasoline
produced in the later years is smaller; the total amount of lead
use over the three years, however, is the same as without banking.
Alternative 2 assumes that some refiners are able to reduce
lead more quickly, so that banking begins in the first quarter of
1985. Those extra banked rights from the first quarter are then
used to reduce banking slightly in the last half of 1985 and to
achieve slightly higher lead levels in 1986 and 1987. The amount
banked is 9.1 billion grams. Again, the total amount of lead
used is the same as without banking.
Table 11-16 compares the year-by-year costs and the present
values of the costs with banking to those without. The estimated
savings are substantial; $173 million for alternative 1 and $226
million for alternative 2. These estimates probably understate
the actual savings that will be realized with banking, because
they do not account for the extra flexibility it allows in meeting
unexpected problems (e.g., equipment breakdowns or a sudden
-------
11-65
TABLE 11-16. Refining Costs Under Alternative Phasedown Patterns,
with
Alternative
Without Banking
With Banking
Alternative 1
Alternative 2
Partial
1985
96
176
170
Misfueling (m
1986
608
420
378
illions or 1
1987
558
463
452
983 dollars)
Present
Value
1,105
932
879
-------
11-66
surge in summertime demand) and in reallocating lead use to
those refiners with higher marginal costs of producing octane.
Table 11-17 reports the marginal costs of an octane-barrel
for 1985 (by quarter), 1986, and 1987 under the alternative
banking scenarios. Under either alternative, the marginal cost
remains under 22 cents per octane-barrel in 1985 and rises to
only 26.7 cents in 1986. These estimates should be compared to
the marginal costs without banking, shown in Table II-7. It is
particularly interesting to note that with banking, the marginal
cost of octane never rises above the level shown for 1988, when
new equipment will first be available. This result has several
important implications. First, it indicates that, with banking,
the phasedown schedule is no more expensive at the margin than
it would be if the Agency delayed the phasedown until 1988, when
refiners can build new equipment. Second, it suggests that
refiners would not find it cost-effective to build new capacity
before 1988, even if it were possible. (It would be cheaper to
buy banked lead rights.) Finally, it suggests that the 1985-to-
1987 phasedown (again, with banking) should not cause financial
difficulties for refiners who could profitably operate at 0.10
gplg once they had time to add equipment. Indeed, to the extent
that some refiners have trouble obtaining loans for capital equip-
ment, these results suggest that they will have an easier time
from 1985 to 1987 than they will in later years, because during
the earlier years they can buy lead rights, the cost of which is
an operating expense, not a large capital outlay requiring loan
or equity financing.
-------
11-67
TABLE 11-17. Impact of Banking on Marginal Costs of Octane
(cents per barrel)
Year
Quarter
1985
I
II
III-IV
1986
I-IV
1987
I-IV
Without
Banking
N.A.
N.A.
20.4
29.2
30.0
With
Alt. 1
N.A.
21.4
21.1
26.7
25.0
Banking
Alt. 2
17.6
21.4
21.1
22.5
25.0
-------
11-68
II.E.2. Sensitivity Analyses with Banking
We also reran several of the sensitivity analyses for 1986
with banking, with the results shown in Tables 11-18 (for alter-
native 1) and 11-19 (for alternative 2). As a comparison of those
tables with Table 11-12 shows, banking greatly reduces the diffi-
culty of meeting the rule, even in the unlikely scenarios that
make several adverse changes simultaneously. For example, without
banking, as shown in Table 11-12, cases M6 and M7 both drive the
marginal cost of an octane-barrel in PADDs I-IV/VI in 1986 over
85 cents and are essentially infeasible. If refiners follow
alternative 2 with banking, however, they can use 0.30 gplg in
1986. As Table 11-19 shows, reformer utilization can then fall
and the marginal cost of octane decreases by more than 50 percent,
to about 40 cents per octane-barrel. This greater comfort in
1986, of course, is partly offset by higher costs in 1985, but
the reductions needed in 1985 should not strain the capacity of
the refining industry. Moreover, this analysis of PADDs I-IV/VI
understates the benefits of banking because rights banked in
PADD V could be sold to refineries in PADDs I-IV/VI. (Based
on a separate analysis of PADD V, we expect such transfers to
occur, as the model shows a lower marginal value of lead in PADD
V than in PADDs I-IV/VI, at any given level of lead use.)
These analyses suggest that while banking is not necessary
to meet the phasedown schedule, it does yield significant cost
savings and, perhaps more importantly, provides an extra margin
of safety against unexpectedly adverse conditions (e.g., higher
octane demands or lower-than-expected ability to utilize down-
-------
11-69
TABLE 11-18. Sensitivity Analyses for 1986 with Banking:
Alternative 1, PADDs I-IV/VI
Reformer Utilization
(%) by Reformer Type
Marginal
Cost of Total
Octane Cost
Run
Base
Base
Ml
M2
M3
M4
Changes**
None (no banking)
None (with banking)
Utilization (B),
less alcohol (D)
(B), (E), plus
old catalysts (C)
(B),(C),(E), plus
higher premium (Al)
Old catalysts (C) ,
Low
90
90
85
85
85
Med.
71
60
81
85
85
High
36
36
36
49
55
Agg.
69
62
72
77
79
(jz< /barrel)
31.6
22.7
23.7
40.3
40.5
($ million)
531
420
365
489
531
higher premium (Al),
less alcohol (D) 90
M5 Old catalysts (C),
less alcohol (D),
summer RVP (El) 90
M6 (C),(D),(E1), plus
higher premium (Al) 90
M7 (C),(D),(A1), plus
summer RVP and 90
demand (E2)
87
90
90
90
36
40
46
72
77
79
81
86
32.1
39.7
46.6
51.5
*Cost for summer quarter only
**Letters refer to parameters listed in Table 11-10
514
140*
149*
169*
-------
11-70
TABLE 11-19. Sensitivity Analyses for 1986 with Banking:
Alternative 2, PADDs I-IV/VI
Reformer Utilization
(%) by Reformer Type
Marginal
Cost of Total
Octane Cost
Run
Base
Bank
Ml
M2
M3
M4
Changes**
None (no banking)
None (with banking)
Utilization (B) ,
less alcohol (D)
(B) ,(E), plus
old catalysts (C)
(B) ,(C) ,(E), plus
higher premium (Al)
Old catalysts (C) ,
Low
90
90
85
85
85
Med.
71
58
78
85
85
High
36
36
36
43
46
Agg.
69
61
71
76
77
(i /barrel)
31.6
22.7
23.3
34.2
36.2
($ million)
531
378
324
411
440
higher premium (Al),
less alcohol (D) 90
M5 Old catalysts (C),
less alcohol (D),
summer RVP (El) 90
M6 (C),(D)f(El). plus
higher premium (Al) 90
M7 (C),(D),(A1), plus
summer RVP and
demand (E2) 90
84
89
90
90
36
36
36
61
75
78
78
84
29.8
30.5
40.0
42.6
*Cost for summer quarter only
**Letters refer to parameters listed in Table 11-10.
453
121'
127'
-------
11-71
stream refinery units, such as FCC units and reformers). Moreover,
these gains are achieved without any increase in the amount of
lead allowed.
-------
CHAPTER III
HUMAN EXPOSURE TO LEAD FROM GASOLINE
Estimating the health benefits of an environmental regulation
requires predicting how the regulation will affect human exposure
levels, in most cases, exposure estimates require extensive model-
ing of emissions, dispersion patterns, population distributions,
and the amounts of inhaled or ingested material that are absorbed
by the human body. Such modeling requires that many parameters be
estimated, often on the basis of very limited information.
In the case of lead in gasoline, however, exposure can be
assessed directly using several large data sets that make it
possible to relate lead in gasoline directly to lead in individ-
uals' blood, without taking the intermediate steps of dispersion
modeling, etc. Analyses of these different data sets have
shown a strong and consistent relationship between the amount of
lead in gasoline and the amount of lead in blood, a relationship
confirmed by experimental data as well.
This chapter presents the methods used to estimate the impact
of reducing lead in gasoline upon levels of lead in the blood of
children and adults. These projections are used in Chapters IV
and V to estimate health benefits for children and adults, respe-
ctively. Section A of this chapter addressess the basic issues,
including the data and the statistical methods used. Section B
discusses the question of causality. Section C presents estimates
of the numbers of children whose blood lead levels would be reduced
below various levels as a result of reducing lead in gasoline.
-------
III-2
III.A. The Relationship Between Lead in Gasoline and Lead in Blood
Individuals are exposed to lead from gasoline through many
pathways. When leaded gasoline is burned in an engine, small
amounts are deposited in the engine and exhaust system, but most
of it is emitted from the tailpipe to the air, where it remains
suspended for a time before settling to the ground. Some
exposure occurs through direct inhalation of the emitted lead.
Additional exposure occurs from ingestion of lead-contaminated
dust, or inhalation of such dust that has been stirred up. Lead
from gasoline also deposits on food crops and is then ingested.
These multiple routes make it very difficult to model individual
exposure pathways. It is possible, however, to estimate the
total amount of lead exposure from gasoline using statistical
methods, as discussed below.
III.A.I. Recent Studies
Several recent articles have shown that blood lead levels
for all age groups will fall as gasoline lead content falls.
The first important statistical studies were done by Billick et
al. (1979), who showed a strong relationship between the blood
lead levels of several hundred thousand children screened in New
York City's lead screening program and local gasoline lead use.
Figure III-l on the next page shows this relationship graphically.
In 1982, Billick et al. presented additional regression
analyses on data from New York City's lead screening program (with
data on several additional years); a Chicago screening program
(800,000 children over more than ten years); and a Louisville,
Kentucky program, all of which confirmed the earlier results.
-------
III-3
FIGURE III-l
Relationship Between Gasoline
Lead and Blood Lead in New York City
35 -
^^M
8
IU
BLACK
HISPANIC
GASOLINE LEAD
30 -
O
III
HA , y,\
•u I/ \ .X\ / \ / \
* K \ / \ •' v \ -
1 7 v \/
1 ,5 - \y
IU
o
10 -
. . t 1 1 . 1 1 1 1 1 1 t 1 t 1 t
1 Vv-V
V ^ A
/\ v>-
vv
V
i _i_ J. j i j ! • j i
z
m
5-° s
AD (Billion olgr
o
<*
K
3.0 2.
6.0
1970
1971 1972 1973 1974 1975
QUARTERLY SAMPLING DATE
1976
-------
III-4
An EPA report (Janney, 1982) found a strong association be-
tween gasoline lead and blood lead in the data from the second
National Health and Nutrition Examination Survey (NHANES II),
after controlling for age, race, sex, degree of urbanization, and
income. Analyses of the NHANES data by the Centers for Disease
Control (CDC) and the National Center for Health Statistics perform-
ed concurrently showed similar results.
Annest et al. (1983) published a paper analyzing the nation-
wide downward trend of blood lead levels from 1976 to 1980 that
was demonstrated in the NHANES II data. Blind quality-control
data were used to ensure that there was no drift in laboratory
measurements. This downward trend was present after controlling
for age, race, sex, region of the country, season, income, and
degree of urbanization, and was present in each age-sex-race sub-
group. Gasoline lead was a significant predictor (p < 0.001) of
blood lead after controlling for age, race, sex, degree of urbani-
zation, income, season, and region of the country, both in all
groups and separately for blacks, whites, white males, white
females, white children, white teenagers, and white adults.
A recent paper by Schwartz et al. (1984b) of EPA's Office of
Policy Analysis presented the results of a study of the relation-
ship between blood lead levels and gasoline lead. Several data
sets were employed for this analysis, including the NHANES II
and the CDC lead poisoning screening program. The statistical
results indicated a highly significant regression coefficient
for gasoline lead levels, which was consistent across all of
the data sets. Estimates of environmental lead from sources
-------
III-5
other that gasoline indicated that paint and other dietary lead
were not the primary sources of the observed decline in blood
lead levels during the 1970s.
Gasoline lead was strongly associated with both the level
of lead in human blood and with the prevalence of elevated blood
lead levels. The association appeared to be causal because other
factors, such as changes in dietary lead and paint lead, did not
account for the changes in blood lead levels that have been asso-
ciated with gasoline lead; the results suggested that more than
one-half of the lead in the average American's blood in the second
half of the 1970s was due to gasoline. Since gasoline lead usage
in that period was restricted by regulation to about 60 percent
of what otherwise would have occurred, those regulations appeared
to have reduced substantially the average blood lead level in the
U.S. and the number of children with lead toxicity-
III.A.2. Available Data Sets
Because gasoline sales and the use of various gasoline
additives are regulated by federal law, information on them is
available from the Department of Energy and the Environmental
Protection Agency. Several data sets contain information on
individual blood lead measurements. Most of these data sets
target children in high-risk groups who have been screened for
lead poisoning. One data set, however, NHANES II, is a large
representative sample of both adults and children in the U.S.
This section describes the data sets used in Section III.A.3. to
estimate the relationship between gasoline lead levels and
blood lead levels.
-------
III-6
III.A.2.a. Gasoline-Use Data
We combined monthly data on national leaded gasoline sales
(from the U.S. Department of Energy) with quarterly average lead
concentrations in grams per gallon (reported to EPA by refiners)
to compute national monthly gasoline lead usage. For the Chicago,
New York, and Louisville areas, we used gasoline sales data from
the Ethyl Corporation's monthly survey of area gasoline sales,
combined with national grams per gallon of lead, to obtain metro-
politan gasoline-lead usage.
III.A.2.b. The NHANES II
The data base for the regressions used to estimate the
coefficients in our prediction models was the health and demo-
graphic information collected in the NHANES II survey. The U.S.
Bureau of the Census selected the NHANES II sample according to
rigorous specifications from the National Center for Health
Statistics so that the probability of selection for each person
in the sample could be determined. The survey used subjects
chosen through a random multi-stage sampling scheme, designed
to utilize the variance minimization features of a stratified
random sample. A total of 27,801 persons from 64 sampling areas
was chosen as representative of the U.S. non-institutionalized
civilian population, aged six months through 74 years. Of those
27,801 persons, 16,563 were asked to provide blood samples,
including all children six months through six years and half of
those aged seven through 74 years. The nonrespondent rate for
blood samples was 39 percent and did not correlate with race,
-------
III-7
sex, annual family income, or degree of urbanization.* A study
of the potential nonresponse bias indicated that this was not
a significant problem (Forthofer, 1983).
Lead concentrations in the blood of sampled persons and con-
trol groups were determined by atomic absorption spectrophotometry
using a modified Delves Cup micro-method. Specimens were analyzed
in duplicate, with the average of the two measurements being used
for the statistical analysis. Bench quality control samples were
inserted and measured two to four times in each analytical run to
calibrate the system. In addition, at least one blind quality-
control sample was incorporated with each 20 NHANES II blood
samples. No temporal trend was evident in the blind quality-
control measurements (National Center for Health Statistics, 1981).
The NHANES II data did, however, display a marked relation-
ship between blood lead and gasoline lead. Figure III-2 plots
gasoline lead and blood lead over time. Note how closely the
changes in blood lead track the changes in gasoline lead, follow-
ing both seasonal variations and the long-term downward trend.
Figure III-3 plots blood lead as a function of gasoline lead after
controlling for age, race, sex, income, degree of urbanization,
region of the country, educational level, smoking, alcohol consump-
tion, occupational exposure, dietary factors, and interactions
among those factors; again, note the strong relationship.
* Because children were less likely to respond (51 percent of
the children did not provide blood for lead determinations in
the NHANES II data set), they were double-sampled. The weights
used to adjust the data to the national population accounted
for both the oversampling and under-response of the children.
-------
FIGURE III-2
110
I100
90
G
OT
w
a
S
i
2 70
•O
w
O
O
O
Ul
50 H
LEAD USED IN GASOLINE PRODUCTION AND
AVERAGE NHANES II BLOOD LEAD LEVELS
(FEB. 1976 - FEB. 1980)
LEAD USED IN
GASOUNE
AVERAGE
BLOOD
LEAD LEVELS
'/
1976
1977
1978
YEAR
1979
1980
16
n
w
15 1
14
» *
Up
in
9
-11
-10
a
ID
O
I
ao
-------
19
18
17
T3
3 16
o
3 .
Q
m 14
o
UJ
I 13
12
11
10
III-9
FIGURE III-3
NHANES II Data: Blood Lead versus Gasoline Lead
10 2.2 2.4 2.6 Z8 aO 32 3.4 a6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 56 5.8 &0
ADJUSTED GASOLINE LEAD (100 METRIC TONS PER DAY)
Note- Each point represents the average of approximately 310 observations. However
the regression line shown is the true regression line for all 9987 individuals.
-------
ni-io
III.A.2.c. The CDC Screening Program for Lead Poisoning
During the 1970s, CDC provided funds for community-wide
screening programs in many cities. Approximately 125,000 children
were screened each quarter of the year. Results reported back
to CDC included the number of children screened, the number with
lead toxicity (defined as free erythrocyte protoporphyrin levels
above 50 ug/dl and blood lead levels over 30 ug/dl), and the
number with severe lead toxicity- Figure III-4 shows a plot of
gasoline lead versus the percent of children with lead toxicity.
III.A.2.d. Chicago, New York, and Louisville, Kentucky Data
Billick analyzed detailed data from the screening programs
in New York, Chicago, and Louisville, including the average
blood lead levels as well as the percent of children with blood
lead levels over 30 ug/dl. We have analyzed these data further,
focusing particularly on the Chicago data, which include a
l-in-30 sample of the over 200,000 blood lead screening tests
performed in Chicago between 1976 and 1980. Figures III-5 and
III-6 show the relationship between gasoline lead and blood lead
in Chicago and New York, respectively.
III.A.3. Statistical Analyses of Exposure
We used multiple regression analysis to examine both the
relationship between gasoline lead and individual blood lead,
and the relationship between gasoline lead and the probability
of undue lead exposure (above 30 ug/dl) or lead toxicity (in the
CDC data). For data from the NHANES II survey, we performed the
analyses on individual blood lead measurements, with the explanatory
-------
8
~ 7
#
i.
i
Q 5
4
£
2 o
1
5 2
1
in-ii
FIGURE III-4
CDC Data: Gasoline Lead versus Percent of Children
with Lead Toxiclty
10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50
GASOLINE LEAD (BILLIONS OF GRAMS PER CALENDAR-QUARTER)
-------
25
24
111-12
FIGURE III-5
Chicago Data-. Gasoline Lead versus Blood Lead
23
O)
^D
Q
S
i20
m
19
18
17
16
0.15 0.18 0.21 0.24 0.27 0.30 0.33 0.36 0.39 0.42 0.45 0.48 0.51 0.54
GASOLINE LEAD (BILLIONS OF GRAMS PER CALENDAR-QUARTER)
-------
24
23
22
21
I
-2 20
Q
n 19
18
17
16
15
CO
111-13
FIGURE III-6
New York City Data: Gasoline Lead versus Blood Lead
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80
GASOLINE LEAD (BILLIONS OF GRAMS PER CALENDAR-QUARTER)
-------
111-14
variables being gasoline lead, age, race, sex, degree of urbaniza-
tion, alcohol consumption, smoking, occupational exposure, dietary
factors, lead in canned food, region of the country, educational
attainment, income, season of the year, and interactions among
these variables. We also ran separate regressions for various
age, race, and geographic groups. We weighted the results by the
the inverse of the probability of selection to produce nationally
representative estimates. The regressions were run using a pro-
gram (SURREGR) designed to account for the stratified, clustered
sampling procedure in the NHANES II survey, and many different
specifications were considered.
Table III-l presents the results for the linear regression
of gasoline lead and other explanatory variables on blood lead
levels. (See Appendix C.I for definitions of the other variables.)
Table III-2 presents the results for the logistic regression on
the probability of blood lead level over 30 ug/dl for children
aged six months to 7 years. Table III-l shows that gasoline lead
was a highly statistically significant predictor of blood lead
levels (p < 0.0001; t-statistic > 60) even after accounting for
potential confounding variables. The average value of gasoline
lead usage was 426 metric tons per day, suggesting that an average
of about 9 ug/dl of blood lead was due to gasoline lead during
that period. Additional regression results on the NHANES II
data, including results for specific subgroups used in later
analyses, are in Appendix C.
For the CDC data, only quarterly data on the percent of
children screened with and without lead toxicity were available.
-------
I11-15
Effect
Intercept
Gasoline Lead
Low Income
Moderate Income
Child (under 8)
Number of Cigarettes
Occupational ly Exposed
Vitamin C
Teenager
Male
Male Teenager
Male Adult
Small City
Rural
Phosphorus
Drinker
Heavy Drinker
Northeast
South
Midwest
Educational Level
Riboflavin
Vitamin A
^ ~ —
Coefficient
6.15
2.14
0.79
0.32
3.47
0.08
1.74
-0.04
-0.30
0.50
1.67
3.40
-0.91
-1.29
-0.001
0.67
1.53
-1.09
-1.44
-1.35
-0.60
-.188
0.018
Standard Error
0.037
0.059
0.034
0.125
0.000
0.063
0.000
0.05
0.19
0.26
0.26
0.085
0.10
0.00
0.03
0.10
0.11
0.14
0.25
0.02
0.005
0.000
P-Value
0.0000
0.0025
0.0897
0.0000
0.0000
0.0000
0.0010
0.1841
0.2538
0.0026
0.0000
0.0039
0.0003
0.0009
0.0007
0.0000
0.0028
0.0005
0.0115
0.0000
0.0186
0.0355
* The coefficients of the dummy variables show how much blood lead (in ug/dl)
is, on average, attributable to a specific effect. The coefficient of gaso-
line lead shows the number of ug/dl of blood lead attributable to each 100
metric tons per day of gasoline lead use.
-------
111-16
TABLE III-2. Logistic Regression on Probability of Blood Lead
> 30 ug/dl for Children 6 months to 7 years
Logistic Regression Results*
Black children = under 8 years old, 479 observations
Dependent variable: 1 if blood lead is over 30 ug/dl; 0 otherwise
Model Chi square = 39.63 with 5 D. F.
Variable Beta Std. Error Chi square P
30.13 U.OOOO
12.40 0.0004
12.26 0.0005
3.33 0.0679
4.39 0.0361
0.90 0.3433
Fraction of concordant pairs of predicted probabilities
and responses = 0.718
White children = under 8 years old, 2225 observations
Dependent variable: 1 if blood lead is over 30 ug/dl; 0 otherwise
Model Chi square = 33.58 with 5 D.F.
Variable Beta Std. Error Chi square P
43.93 0.0000
8.59 0.0034
17.21 0.0000
3.23 0.0724
5.36 0.0206
2.31 0.1285
Fraction of concordant pairs of predicted probabilities and
responses = 0.637
Intercept
Gaslead
Poor
Age 1
Age 2
Age 3
-6.9468
0.8633
0.9815
1.1404
1.1938
0.5428
1.2656
0.2452
0.2803
0.6246
0.5696
0.5728
Intercept
Gaslead
Poor
Age 1
Age 2
Age 3
-8.1667
0.6331
1.2174
1.4332
1.7168
1.1405
1.2322
0.2160
0.2935
0.7978
0.7415
0.7503
*All logistic regression results were run using PROC LOGISTIC
within the Statistical Analysis System (SAS). This procedure
uses individual data where the dependent variable is one if
the individual is above the threshold, and ze*-o otherwise.
-------
111-17
We performed both weighted linear and logistic regressions on
the probability of lead toxicity as a function of gasoline
lead.
Tables III-3a and III-3b show the results of regressing the
percent of children with lead toxicity against gasoline lead for
the 20 quarters between 1977 and 1981. "Dummy" represents the
period before CDC modified its definition of lead toxicity in
1978. The correlation coefficients between gasoline lead and the
percent of children with lead toxicity (0.8027) and between the
change in gasoline lead and the change in the percent with lead
toxicity (0.817) were both very large, and the regression
coefficient was highly significant (p < 0.0001). (The magnitude
of the coefficient is not directly comparable to the NHANES II
both because of the difference in outcome variable -- percent
toxic versus blood lead level — and the difference in the units
of gasoline lead.) The regressions predicted that if there had
been no lead in gasoline at all, there would have been approxi-
mately 80 percent fewer cases of lead toxicity. This does not
mean that 80 percent of the cases were due solely to gasoline
lead, but rather that in 80 percent of the cases both gasoline
lead and other exposures were required to bring children above
the CDC definition of lead toxicity.
For the Chicago, New York, and Louisville data, we performed
logistic and linear regressions of the probability of blood lead
levels being over 30 ug/dl for each race separately, with
explanatory variables being age, season, and gasoline lead.
Gasoline lead was always significant, and explained the seasonal
-------
111-18
TABLE III-3a. Regression of CDC Screening Data: Percent of Children
with Lead Toxicity on Gasoline Lead
Variable
Constant
Gas Lead*
Dummy
Coefficient
1.363
0.1601
-1.036
T-Statistic
1.722
5.322
-1.524
Incremental R?_
0
0.5215
0.0427
P-Value
0.05
<0.001
0.07
R2 = 0.687
Durbin Watson statistic = 1.786
Simple correlation coefficients;
Gas Lead Dummy
Percent Toxic 0.8027 0.4068
Gas Lead - 0.6927
* Gasoline lead is in billions of grams per quarter. In the last quarter
of 1978, gasoline lead use was 40 billion grams. By the second quarter
of 1980, regulatory action had reduced this to 20 billion grams.
TABLE III-3b. Regression of CDC Screening Data: Change in Lead Toxicity
on Change in Gasoline Lead
Variable
Constant
Delta Gas Lead
Coefficient
0.3675
0.3938
T-Statistic
1.43
5.84
P-Value
0.09
<0.0001
R2 = .6670
Durbin Watson statistic = 1.887
Simple correlation coefficient; 0.8167
-------
111-19
variation in elevated blood lead levels. We also regressed quar-
terly average blood lead levels in Chicago and New York against
gasoline lead, race, age, and season. Table III-4 shows the
results of regressing the Chicago data for black children for the
18 quarters from 1976 until mid-1980. (Our results differ from
Billick's in that he only had lead concentration values for two
quarters of the year and had to interpolate the others. We had
lead concentrations for all four quarters and, not surprisingly,
using this better gasoline-lead data gave a stronger relationship.)
Again, gasoline lead was an excellent predictor of children's
blood lead levels (p < 0.0001). Autocorrelation corrections were
run, and showed no significant correlation, nor did they produce
a noticeable change in gasoline lead coefficients. Since the
gasoline lead coefficient in Table III-4 was in billions of grams
per quarter for the Chicago area, we had to adjust from local to
national units to compare it to the NHANES II results. Scaling
by the ratio of Chicago's gasoline use to the nation's, and
converting to the units used in the NHANES II regressions (100
metric tons per day), the Chicago coefficient would correspond
to a national coefficient of 2.08, which is essentially identical
to the NHANES II results (a coefficient of 2.14, as shown in
Table III-l).
Regressions on all of the data sets showed that gasoline
lead was an extremely significant explanatory variable both for
individual blood lead levels and for the percent of children with
undue lead exposure or lead toxicity. Gasoline lead appeared to
have accounted for 60 percent of the lead in Americans in the
-------
111-20
TABLE III-4. Black Children in Chicago: Regression of Average Blood~Lead
Levels on Gasoline Lead Levels
Variable
Constant
Gas Lead*
Age 1
Age 2
Age 3
Age 4
R2 = 0.6194
Durbin Watson
Coefficient T-Statistic Incremental P.2
15.48 22.33 0
17.02 11.27 0.5757
0.85 1.53 0.0106
1.68 3.03 0.0418
1.01 1.84 0.0153
0.66 1.20 0.0065
statistic =2.01
P-Value
-
<0.0001
0.07
0.005
0.05
>0.10
Simple correlation coefficients:
Average Blood
Gas Lead
Gas Lead Age 1 Age 2 Age 3
Lead 0.7587 0.0009 0.01615 0.0337
000
Age 4
0.0342
0
*Gas Lead is lagged one month behind quarterly gasoline lead in billions
of grams. The average for this period in Chicago was 0.379 billion grams.
-------
111-21
second half of the 1970s, and to have explained both the seasonal
increases in blood lead levels from winter to summer and the
long-term drop in blood lead levels during the late 1970s. EPA
refinery reports show that the rate of decline in gasoline lead
accelerated after late 1978, and this was paralleled by an accel-
erated decline in blood lead levels.
Gasoline lead had the same coefficient in rural areas, in
urban areas, and in urban areas with populations over one million,
In the NHANES II regressions, a dummy variable for residence in
central cities versus suburban areas was not statistically signi-
ficant. The gasoline lead variable that best correlated with
blood lead was gasoline sales for the preceding month.
III.B. The Question of Causality
While not crucial to a rule designed to be precautionary in
nature, we did examine whether there is a causal relationship
between gasoline lead and blood lead. In epidemiology there
are several general criteria for determining whether association
represents causality (Kleinbaum et al., 1982; Lilienfeld and
Lilienfeld, 1980). The most useful criteria are:
Is there experimental evidence to support the findings?
Do several studies replicate the results?
Does a dose-response relationship exist?
Are there consistent effects in different types of studies?
Does cause precede effect?
Is the model biologically plausible?
Is it unlikely that other factors not included in the
analysis would change the results?
-------
111-22
We will address these issues in turn, with particular emphasis on
the last one, as no study can ever measure all possible confounding
factors.
III.B.I. Experimental Evidence
Facchetti and Geiss (1982) investigated the contribution of
gasoline lead to blood lead in Turin, Italy during the late 1970s
by changing the isotopic composition of the lead added to gasoline,
and monitoring the isotopic composition of blood lead. This iso-
topic lead experiment indicated that changes in the isotopic com-
position of air lead followed closely and rapidly changes in the
isotopic composition of gasoline lead. Changes in the isotopic
composition of blood lead also paralleled changes in gasoline lead.
Based on modeling of the results, Facchetti and Geiss estimated
that at least 25 percent of the (high) blood lead levels in Turin
were due to gasoline lead; this was at least 6 ug/dl. Since
blood lead isotopic ratios were still changing when the gasoline
isotopes were switched back, the actual impact of gasoline
lead is probably higher.
Manton (1977) analyzed isotopic changes in blood lead in the
United States and found the contribution of airborne lead (predomi-
nantly gasoline) was between 5 and 10 ug/dl in most of his subjects.
Tera et al. (1985) recently analyzed the isotopic ratios
of blood lead in children in Washington, B.C. as the isotopic
ratio of air lead changed. Their data showed that, as late as
1983, at least 38 percent of the lead in children's blood still
came from gasoline lead, despite the 50 percent reduction in
gasoline lead since 1978.
-------
111-23
Our analysis of the impact of gasoline lead indicated that
in the late 1970s, about 9 ug/dl of blood lead resulted from the
lead in gasoline. This magnitude of effect is similar to that
found in the isotope studies.
11I.E.2. Does Cause Precede Effect?
One way to determine whether the trend in blood lead was
caused by gasoline lead or was due to another variable is to
examine the lag structure in light of our biomedical knowledge.
The half-life of lead levels in the blood is about 30 days
(Rabinowitz et al., 1976). Because the average blood test oc-
curred on the 15th of the month, the current month's gasoline lead
would have had only 15 days to affect blood lead levels, and so,
though significant, should have a lesser impact. The previous
month's gasoline lead, on the other hand, represents emissions
on average 15 to 45 days prior to examination, and since direct
inhalation and even dust exposure shows up rapidly in the blood,
we would expect this one-month lagged gas lead to be more signifi-
cant, with a noticeably higher coefficient. Similarly, we expected
gasoline sold two months previously to be less significant and
of lesser magnitude. If gasoline sales were merely a proxy for
time, however, all three months should be equally good predictors,
since t, t-1, and t-2 (where t is the month since commencing the
survey) equally represent the passage of time.
We regressed the individual blood lead levels in the NHANES
II survey against current, one-month lagged, and two-month lagged
gasoline lead simultaneously, and found one-month lagged gasoline
lead was most significant and two-month lagged gasoline lead was
-------
111-24
least significant. This suggested that the causal model was
correct. The coefficient of two-month lagged gasoline lead was
one-half that of one-month lagged gasoline lead, which matches
the one month half-life of lead in the blood.
III.B.3. Replicability and Consistency
We replicated our analysis of the national NHANES II data
with our analysis of the site-specific Chicago and New York
screening data and with Billick's analysis of the screening data
from Chicago, New York, and Louisville. In addition, Rabinowitz
and Needleman (1983) have examined umbilical cord blood from over
11,000 consecutive births at Boston Women's Hospital between 1979
and 1981. They found a strong association (p < 0.001) with gaso-
line lead used in the Boston area, and also that one-month lagged
gasoline lead had the highest correlation. No significant monthly
variation was noted in the mothers' education levels, smoking, or
drinking, and water lead levels increased somewhat over the period,
while blood lead and gasoline lead levels fell. Thus, both local
and national data from different studies, collected by different
investigators, show the same pattern of gasoline lead being signi-
ficantly related to blood lead.
The results of studies of both individual blood lead levels
and average blood lead levels were consistent with analyses of
elevated blood lead levels. Billick examined the probability of
blood lead levels above 30 ug/dl in Chicago, New York, and
Louisville, and found a strong relationship to gasoline lead.
We repeated that analysis using logistic regressions, with the
same results. Our investigation of the national CDC screening
-------
111-25
program data, using both linear and logistic regressions, found
the probability of lead toxicity was strongly dependent on the
amount of lead in gasoline. Finally, we performed logistic
regressions on the NHANES II data of the probability of both black
and white children (under 8 years) and preteenagers (ages 8-14)
being over 30 ug/dl of blood lead, and again found a strong
relationship to gasoline lead levels.
III.E.4. Does a Dose-Response Relationship Exist?
To assure ourselves that the linear relationship we found
was the true form of the dose-response relationship, we divided
the NHANES sample in half. We repeated the regression for whites
in the second half of the survey period, when the average gasoline
lead levels were roughly 50 percent of those in the first half.
The gasoline lead coefficient was essentially unchanged at
1.94 (compared to 2.14 for the full sample). This indicated a
stable linear relationship.
For blacks, the sample size was too small to divide the
sample, so we used an alternate procedure. We regressed log
(blood lead) against demographic variables and log(gasoline
lead), for a range of zero gasoline intercepts. This model
chose the best exponent for the relationship blood lead = (gasoline
lead)B. For the intercept (8 ug/dl) with the highest R2, the
exponent was 0.98, indicating a linear dose-response relationship.
A square-root regression was also tried, and gave an inferior
fit to the linear regression. The Chicago data, as noted before,
gave the same magnitude of gasoline lead's effect on blood lead,
-------
111-26
and visual examination of the regression plots in Figures III-3
to III-6 confirm the linearity in the relationship.
III.B.5. Biological Plausibility
Gasoline lead produced 90 percent of the emissions of lead
into the air in the 1970s, and it was the major source of lead
contamination of the environment. Lead is emitted as predominantly
respirable particulates (less than 1 micron) from auto exhausts.
Respirable particulates reach into the lung and show a high absorp-
tion rate. Gasoline lead is a major source of lead in street and
household dust and in soil contamination and, therefore, in addition
to direct inhalation, results in secondary exposure through the
inhalation and absorption of dust. Roels (1976) has shown a high
correlation between blood lead and lead on the hands, presumably
from air and dust contamination. Lead is known to be absorbed
from both the lung and gut. Thus, the fact that gasoline lead
is related to blood lead is biologically plausible.
III.B.6. Control for Confounding Factors
Billick's analysis of children controlled for age and race,
and our reanalysis controlled for season as well. The Annest
et al. analysis of the NHANES II data controlled for age, race,
sex, income, degree of urbanization, region of the country, and
season.
We used several approaches to control for confounding factors.
Where we had sufficient information, we included potential con-
founding factors, or surrogates for them, in our regressions.
In addition, we examined external data to check for changes in
-------
111-27
confounding factors that might bias our results, and performed
several internal statistical checks to examine the likelihood of
misspecification error.
III.B.6.a. External Validation
A great deal is known about the sources of lead exposure.
The major general sources of body lead are food, water, and paint.
While we have limited data on the specific sources of exposure
of the individuals in our data sets, if all other general exposures
were, on average, constant during the period, the effects of these
will be part of the constant term in our regressions. Bias would
occur only if these sources changed over the period.
The Food and Drug Administration's estimates of lead in the
diet (based on market basket surveys) during the NHANES II period
are shown on Table III-5. As there was no downward trend in
dietary lead intake, this was unlikely to have been a potential
source of bias in the model. Lead in canned food did change
during the period, and we included a variable for this in our
regression model. (See Appendix C.) Dietary variables were used
to represent differences in food consumption between individuals,
which were also accounted for by age, sex, race, income, education,
and regional variables.
-------
111-28
TABLE III-5. Lead in the Diet (micrograms/day)
Fiscal Year
1976
1977
1978
1979
1980
Infants
21
22
25
36
— —
Toddlers
30
28
35
46
—
Males aqed 15-20
71.1
79.3
95.1
81.7
82.9
The exposure to lead from water is predominantly a function
of pH levels. The pH of drinking water supplies appeared to have
been constant over the four to five-year period of our data.
Change in paint lead exposure was also an unlikely source
of bias for several reasons. First, adult blood lead levels
decreased by almost as much (37 percent vs. 42 percent) as child-
ren's blood lead, and adults generally do not eat paint. Second,
because ingestion of paint lead usually results in the absorption
of enough lead to produce large increases in blood lead, we
would expect a drop in paint lead exposure to reduce blood lead
levels only in people whose levels are above the mean. However,
the drop in blood lead recorded in the NHANES II data shifted the
entire distribution dramatically. Indeed, even low blood-lead
groups showed major declines in blood lead. This would not have
occurred if ingestion of paint lead were the determinant.
Third, the drop occurred across geographical boundaries, in
suburbs as well as central cities. (Suburbs have a lower fraction
of pre-1950 housing stock, and, therefore, inherently less expo-
-------
111-29
sure to lead paint, yet they showed the same drop and the same
gas lead coefficient.*) Finally, lead paint removal programs
during this period reached only 50,000 of the 30 million housing
units with lead paint (less than 0.2 percent), so exposure was
unlikely to have changed in this period (Morbidity and Mortality
Weekly Reports, 1976-1980).
To lay these issues to rest, however, we did an additional
analysis of the Chicago data. For each year of the screening
program, the CDC reported the percent of lead toxic children
with a lead paint hazard in their home or the home of a close
relative. In 1978, however, the homes of all screened children
in Chicago (no matter what their blood lead level) were checked
for lead paint hazard. This survey of over 80,000 housing units
established the general prevalence of lead paint exposure in
the screened population. With the probability of paint lead in
the house given lead toxicity, and the probability of paint
lead in the house in general, we used Bayes Theorem to compute
the probability of lead toxicity given paint lead in the house,
and the probability of lead toxicity when lead paint was not in
the house. We then regressed these quarterly probabilities on
gasoline lead. The results are shown in Table III-6.
* Shier and Hall (1977) analyzed over 2,500 housing units in
Pittsburgh and found that the fraction that had lead paint
concentrations above 2 mg/cm^ decreased from over 70 percent
in pre-1940 housing to about 43 percent in 1940-59 housing,
and to 13 percent in post-1960 housing.
-------
111-30
TABLE III-6. Chicago Probability of Blood Lead > 30 ug/dl
With and Without Paint Lead Hazard in the Home
Variable
Without Lead
Paint Hazard
Constant
Gas Lead
With Paint
Lead Hazard
Constant
Gas Lead
Coefficient t-Statistic
-2.087 -0.6583
35.398 4.423
12.80 0.579
80.16 1.436
P-Value
0.26
0.0005
0.30
0.07
As we expected, gasoline lead was a more significant
explanatory variable for children not exposed to lead paint
than for the exposed group. Over 80 percent of the children
were not exposed to lead paint at home, and among them gasoline
lead was highly significant and explained most of the elevated
blood lead levels. It was also striking that even among children
with an identifiable lead paint hazard in their homes, gasoline
lead was still strongly related to the probability of lead
toxicity-
III.B.6.b. Seasonality
U.S. blood lead levels were strongly seasonal, with summer
levels substantially higher than winter levels. However, when
the NHANES II data were tested with seasonal variables, none of
them was statistically significant, or even close to significant,
when gasoline was included; the highest F-statistic for any
season was 0.82. This indicated that the same gasoline lead
-------
111-31
coefficient successfully explained both the short-term increases
in blood lead levels from winter to summer and the long-term
decrease in blood lead levels over the four-year period. The
six regressions on the Chicago, New York, and Louisville data,
as mentioned before, also indicated that seasonal!ty was insigni-
ficant once gasoline lead was in the model. This suggested that
other long-term trends in lead exposure cannot have biased the
gasoline coefficient, as the short-term and long-term gasoline
coefficients were the same.
III.B.6.C. Other Time Trends
To test the hypothesis that there was another unknown lead
factor that was decreasing over the period, and whose effects
might be attributed to gasoline in our regressions, we repeated
our analysis with time as a variable; time was entered as the
number of days after February 1, 1976 that a blood sample was
taken.
The results indicated that time was not significant when
gasoline lead was in the regression. Moreover, the effect of
gasoline lead on blood lead was reduced by only 23 percent if
we kept the insignificant time variable in. In addition, the
Chicago data for average blood lead were analyzed with time as a
variable. Here, again, gasoline lead was significant (p < 0.02)
while time was insignificant (p > 0.30). We also tested time-
squared on the CDC data, and it was insignificant.
III.B.6.d. Geographic Sampling Pattern
To be certain that the pattern of geographic sampling over
time in the NHANES II period did not produce changes in blood lead
-------
111-32
that the regression falsely attributed to gasoline, we included
dummy variables for four regions of the country in our analysis.
To check further, we inserted dummy variables for all 48 locations
identified by the National Center for Health Statistics (NCHS).
NCHS did not release locational data for 16 counties with popu-
lations below 100,000; these were all represented by one additional
dummy variable. Rerunning the regressions with these identifiers
changed the gasoline lead coefficient by only about 5 percent,
which was insignificant.
This result was extraordinary because the NHANES II sampled
different cities at different times, and the regression allowed
the differences in blood lead levels over time to be attributed
to changes in city location rather than gasoline lead. This meant
that the coefficient of gasoline lead was heavily determined by
the month-to-month changes in blood lead and gasoline lead within
each city during the two months or so spent at each site. Yet
the results did not change.
In addition, we performed a regression that had an interac-
tion term between gasoline lead and the city identifiers. This
procedure allowed a different slope in the gasoline lead/blood
lead relationship for each county. The regression still yielded
a coefficient for national gasoline lead of 1.83, with p < 0.0001.
Finally, the fact that analyses of many individual cities
across the country (Boston, Chicago, New York, and Louisville)
yielded similar results suggested geographic location was not an
important source of bias.
-------
111-33
III.B.6.e. Subgroup Analysis
To verify the robustness and stability of the relationship
between gasoline lead and blood lead, we ran regressions of the
NHANES II data for several demographic groups separately. This was
done both to verify that we properly controlled for these vari-
ables, and to investigate whether the observed drop in blood lead
levels was, in fact, due to a universal exposure such as gasoline
lead, or could have been due to a source primarily affecting
certain subgroups. The stability of the gasoline lead coefficient
across these subpopulations reduced the likelihood of a speci-
fication bias.
This was because the bias introduced into the gasoline
coefficient due to an omitted variable would be proportional to
the regression coefficient that variable would have had if
included. If the gasoline lead coefficient was insignificantly
different among the regressions run for different subgroups, the
omitted variable either must not have significantly biased the
gasoline lead coefficient, or must have been coincidentally
constant among all the demographic groups. The only known such
variables were gasoline lead and food lead.
We performed separate regressions for males, females, adults,
and children, for cities over one million, and for smaller urban
areas. The maximum pair-wise difference in the gasoline lead
coefficient among the six subgroups was less than 10 percent. In
addition, we changed the definitions of large city (from over
1,000,000 to over 250,000), of rural (to include rural areas or
cities under 10,000), and of the age categories. The coefficient
-------
111-34
of gasoline lead changed by only 2 percent when we performed this
regression. The stability of our findings, given the many addi-
tional tests we conducted, testifies to the robustness of the
relationship between blood lead and gasoline lead. In addition,
a study by Annest et al. (1983) of the U.S. Public Health Service,
also using data from the NHANES II, found that the only reasonable
explanation for the decline in blood lead levels was the decline in
the amount of lead in gasoline. The Centers for Disease Control,
the epidemiological branch of the Public Health Service, has also
endorsed this relationship in its comments on this proposed rule
(Public Docket EN 84-05) and the 1982 rule.
III.C. Impact of Rule on Numbers of Children
Above Various Blood-Lead Levels
In estimating the benefits of reducing lead in gasoline, we
relied on the regression coefficients estimated from the NHANES II
data set because it is the largest and most representative of those
available. Estimates of adult health benefits use the continuous
functional forms because the relationship between blood lead and
blood pressure is a continuous function of blood lead levels;
their application is discussed in Chapter V. The benefit esti-
mates for children, however, are functions of the numbers of
children brought below various blood lead levels. Section III.C.I
describes how we used the results of the NHANES II logistic
regressions to estimate changes in the numbers of children above
these various levels. Section III.C.2 discusses the interpreta-
tion of these results with respect to prevalence and incidence.
-------
111-35
III.C.I. Estimation Procedure
To estimate the numbers of children above different blood
lead levels, we relied on logistic regressions estimated from
the NHANES II data, of the type reported in Table III-2. These
regressions were estimated separately for blacks and whites, and
for each two-year age group from 6 months through 13 years. In
each regression, the dependent variable was the natural logarithm
of the odds of being above the level, while the independent
variables were various demographic factors and gasoline lead.
To predict how the number of children above each level
would change as the amount of lead in gasoline was reduced, a
mechanism was needed to forecast the distribution of blood lead
levels as a function of gasoline lead. In this analysis, we
assumed that the distribution of blood lead would remain log-
normal as gasoline lead levels declined. Then, estimates of the
mean and variance of the associated (transformed) normal distri-
bution could be used to determine the percentage of the popula-
tion above any specific blood lead level. The estimates of the
mean and standard deviation of the underlying normal distribution
were derived from logistic regression estimates of the percentage
of children with blood lead levels above 30 ug/dl and SURREGR
estimates of the mean of the log-normal distribution using the
Statistical Analysis System (SAS) procedure, SURREGR.
If the -distribution 'X1 is normal with mean 'u1 and standard
deviation 's1 (X:N (u,s)), then Y = exp (X) is log-normal with a
mean of 'a1 and a standard deviation of 'b1, where
-------
111-36
(III-l) a = exp (u -I- 1/2 s2) and
(III-2) b = exp (2u + s2) (exp (s2) -1)
Further, if eg and vg are the same percentiles of the log-normal
and its corresponding normal distribution, respectively, we have
(III-3) 6g = eXp (U + Vg S).
We used the logistic regressions to estimate eg in eguation
(III-3) and the SURREGR regressions to estimate a in eguation
(III-l), which yielded
(III-4) a = exp (u + 1/2 s2)
(III-5) eg = exp (u + vg s)
Solving these equations for u and s produced a quadratic equation:
(III-6) 0 = (In (eq) - In (a)) - v s + .5s2
which had the solution s = v + [v 2 - 2 (In (e ) - In (a))] °-5,
Only the smaller root yielded sensible values for u and s. Then
u = ln(a)-l/2 s2. Using the estimated values for u and s, we
determined percentages of the distribution above 10, 15, 20, and
30 ug/dl by looking up the results of (In (10) - u)/s, etc., in
the normal table.
We used a logistic regression eguation to estimate the
percentage of children over 30 ug/dl to control for problems of
multiple sources of exposure. If we had simply used the regres-
sions explaining the mean and assumed a constant standard devia-
-------
111-37
tion, we would have predicted that removing lead from gasoline
would have resulted in there being no children above 30 ug/dl.
This seemed unreasonable because paint and food are known
alternate sources of lead, and also are associated with high
blood lead levels. The logistic regressions confirmed that the
geometric standard deviation changes as the mean falls. Because
of the sensitivity of the blood lead distribution to age, we
estimated separate distributions for each two-year age interval.
The tabulated changes in the numbers of children above various
levels represented the sum of distributions for each age category.
The regression results are shown in Appendix C.
For children from six months to seven years old, we used
logistic regressions for the percent above 30 ug/dl of blood
lead. For children aged eight to thirteen, we used logistic
regressions for the percent above 20 ug/dl blood lead because
there were too few observations above 30 ug/dl for the logistic
procedure to yield accurate estimates.
Table III-7 presents the estimated reductions in the numbers
of children over various blood lead levels from 15 ug/dl to 30
ug/dl. The estimates are presented for three phasedown schedules:
the original proposal (0.10 gplg starting 1/1/86); the alternative
discussed in the NPRM (0.50 gplg on 7/1/85, 0.30 on 1/1/86, 0.20
on 1/1/87, and 0.10 on 1/1/88); and the final rule (0.50 gplg on
7/1/85 and 0.10 on 1/1/86). All of the estimates assume that
misfueling is totally eliminated. (The impacts of alternative
assumptions about misfueling are explored in Chapter VIII.)
-------
111-38
TABLE III-7. Estimated Reductions in Numbers of Children Over
Various Blood Lead Levels, Assuming No Misfueling
(thousands of children)
Blood Lead Level
Rule
30
25
20
15
10
ug/dl
Proposed
Alternative
Final
ug/dl
Proposed
Alternative
Final
ug/dl
Proposed
Alternative
Final
ug/dl
Proposed
Alternative
Final
ug/dl
Proposed
Alternative
Final
1985
0
22
22
0
72
72
0
232
232
0
696
696
0
1,972
1,972
1986
52
46
52
172
154
172
563
501
563
1,726
1,524
1,726
4,949
4,354
4,949
1987
47
45
47
157
149
157
518
491
518
1,597
1,508
1,597
4,595
4,333
4,595
1988
43
43
43
144
144
144
476
476
476
1,476
1,476
1,476
4,261
4,261
4,261
1989
39
39
39
130
130
130
434
434
434
1,353
1,353
1,353
3,918
3,918
3,918
1990
36
36
36
119
119
119
400
400
400
1,252
1,252
1,252
3,637
3,637
3,637
1991
32
32
32
106
106
106
357
357
357
1,125
1,125
1,125
3,283
3,283
3,283
1992
31
31
31
103
103
103
348
348
348
1,098
1,098
1,098
3,215
3,215
3,215
-------
111-39
III.C.2. Incidence Versus Prevalence
Our predicted decrease in the number of children above a
given threshold is for a specific point in time; our cost esti-
mates are for an entire year. If children remain above 30 ug/dl
for less than a year, there will be more children above 30 ug/dl
in a year than we estimated and our benefits will be understated.
Conversely, if children remain above 30 ug/dl for more than a
year, these cases may be counted twice and we will overstate
benefits.
This raises the difficult epidemiological issue of prevalence
versus incidence. Prevalence is the percent of people who have
the condition of interest at a particular time (e.g., the percent
of people with the flu on February 14). Incidence is the percent
of people who develop new cases of the flu in a given time period
(e.g., the month of February). Prevalence is incidence times
average duration.
This issue is important because the NHANES II survey,
upon which we based our regressions, measured the prevalence of
cases above 30 ug/dl blood lead or other thresholds, rather than
the incidence. Yet the benefits we want to estimate would, in
fact, be reduced numbers of cases in a time period, i.e.,
incidence.
Clearly an excursion of a child's blood lead level above
30 ug/dl for a day or two will produce less damage than a pro-
longed elevation. However, data indicate that such occurrences
are not very likely. Odenbro et al. (1983) in Chicago found
fairly stable blood lead levels in individual children with high
-------
111-40
levels. For these children, levels remained high for more than
a few days, usually for months or years. However, if the average
duration of elevated blood lead was six months, the actual number
of children affected in a year would be twice the average preva-
lence for the year. This obviously would affect our benefit
estimates.
Because, as discussed in Chapter IV, we only valued cognitive
losses for children in CDC categories III and IV,* and because
data from Odenbro et al. suggested that such children's blood lead
levels remain elevated for a long time unless treated, we believe
our prevalence estimate is reasonable for estimating cognitive
effects. Medical management costs, on the other hand, seem more
reasonably associated with incidence.
In any case, it was necessary to determine the duration of
elevated lead levels. To do this we evaluated several available
pieces of information. They all suggested that the average
duration was less than one year, so that our estimate of prevalence
(based on the NHANES data) understated annual incidence.
* CDC classifies children as "lead toxic" if they have blood lead
levels above 30 ug/dl and free erythrocyte protoporphyrin (FEP)
levels above 50 ug/dl. (FEP is a measure of the derangement of
the heme synthesis process caused by lead.) Children with blood
lead levels between 30-49 ug/dl but with FEP under 50 are in
CDC category Ib. Children between 30-49 ug/dl blood lead
and 50-109 ug/dl FEP are category II. Category III children
have blood lead levels between 30 and 49 ug/dl and FEP levels
between 110 and 249 ug/dl or blood lead levels between 50 and
69 ug/dl and FEP levels between 50 and 249 ug/dl. Category
IV includes all children with blood lead greater than 70 ug/dl,
and children with blood lead between 60 and 69 ug/dl and FEP
levels above 250 ug/dl.
-------
111-41
Our first source was the CDC screening program data. This
program screened about 100,000 to 125,000 children per quarter
of the year to detect lead toxicity. Approximately 6,000 to
7,000 cases were found each quarter; this established the general
prevalence of lead toxicity in the screening population. However,
this prevalence rate showed strong within-year variation, with
levels much higher in the third quarter, summer, when gasoline
consumption was also highest. This intra-year variation suggested
that the average duration was not so long that the effects of
quarterly changes in exposure were swamped by cases that origin-
ated in earlier quarters.
We also have used the CDC lead screening data in another
way. CDC reported, quarterly, the number of children under
pediatric management, which included all the new cases discovered
during that quarter plus the children remaining under pediatric
management who had been discovered with lead toxicity in the
previous quarters. We compared that number to the sum of the
cases detected in the same quarter plus the previous two quarters
and found the results were quite close. This suggested that
children remained under pediatric management for an average of
three quarters. However, children generally had several medical
visits after their blood levels returned to normal to ensure that
the decline was real. This implied that the average duration of
blood lead levels above 30 ug/dl was even shorter, closer to two
quarters. If this is true, then it is possible that we have
underestimated the annual incidence of cases of children above
30 ug/dl by as much as a factor of two.
-------
111-42
The amount of time it takes for lead toxicity percentages to
respond to fluctuations in gasoline lead levels also may help to
determine the duration of lead toxicity. If this time is rela-
tively short (e.g., a few months or less), it is unlikely that
duration would extend beyond a year. For lead toxicity to last
a year or more, one would expect lead toxicity levels to be
relatively insensitive to intra-annual variation in gasoline
lead. The CDC data, which show such variation, suggest that the
average duration is substantially less than one year.
Two other data sets supported the conclusion of a short lag
between gasoline lead and blood lead levels. First, in the
NHANES II data, we examined both the lag structure of blood lead's
relationship to gasoline lead, and whether any seasonal dummy var-
iables were significant in explaining the large observed seasonal
variations in blood lead. Schwartz et al. (1983) found that the
lag structure of average blood-lead levels' dependence on gasoline
lead could extend to three months.
In addition, Billick (1982) examined the results of the
screening programs for lead toxicity in Chicago (800,000 children
screened) and in New York City (450,000 children screened) over a
10-year period and found a strong seasonal pattern in the number of
children with lead toxicity. This pattern followed the seasonal
variation of gasoline use. When Schwartz et al. (1984b) analyzed
these data in a logistic regression, gasoline explained the
cyclical variation in blood lead levels, with no seasonal variable
obtaining a p-value of better than 0.38.
-------
111-43
All of this suggested that the average time a child spent
above 30 ug/dl was short enough so that quarterly prevalence
rates corresponded well to quarterly exposure incidence.
Therefore, our estimate of the annual incidence of children
above 30 ug/dl is likely to be low, as is our estimate of
avoided medical expenses.
-------
CHAPTER IV
BENEFITS OF REDUCING CHILDREN'S EXPOSURE TO LEAD
The scientific literature presents evidence of a wide range
of physiological effects associated with exposure to lead. These
range from relatively subtle changes in various biochemical
measurements at very low levels of exposure, with uncertain impli-
cations for health, to severe retardation and even death at very
high levels of exposure. Although such effects are found in indi-
viduals of all ages, particular concern has focused on children,
because thay appear to be at greater risk.
Because the body is a complex structure of interdependent
systems and processes, effects upon one component will have
cascading implications throughout the body. This interdependence
is well illustrated by multi-organ impacts resulting from the
inhibition of heme by lead, and consequent reduction in the body
heme pool. These effects are depicted graphically in Figure IV-1,
taken from EPA's most recent Draft Lead Criteria Document
(p. 13-31).
This chapter summarizes the available evidence of the
effects of lead on children, and develops rough estimates of the
benefits of reducing exposure to lead by reducing lead in gaso-
line. Section A deals with the pathophysiological effects of
lead, while Section B addresses the evidence on neuropsychologi-
cal effects (primarily reduced cognitive ability). The final
section discusses the methods used to monetize the benefits of
reducing children's exposure to lead. Although these estimates
in no way cover the complete range of potential benefits, they
total several hundred million dollars per year for the rule
being promulgated.
-------
IV-2
FIGUHE IV-1 Multi-Organ Impacts of Lead's
Effect on the Heine Pool
•IOIKI
«MH
IMOI
IT root
IDTTMIWOItTIC KDUCIO
«»<"' • 7r:r "
•IUML MOUCIOMIiOniOTIIIIS
iMicn IM tTiocmnwiti """"
4
i
MMiiHOcm.^ „„„„,, ,„„,,,„. tf
UIICTI vuuiii o f(
., . 1 \
V
IHICTt * Will •IlllltttO
iunie«MTI
iiricnMiiimns
MOM. MO
KM»« Clltl
*
•»*IMOIITILIII>IIO>
MOMUVI CKOUCIIO*
•TAIIIIIIOIVIIOWIIII
oiHuvoiniriTiii
•tMIHIOIMdlllL
t«UI MMfOITUIS
•rumoMLCiiw
•Oil UIICMO
MBIUCdl
•»>!•! OULCIW
MLC • CTCLIC
•uciiOTioi m»ioinii
•MIIIIOOI10IIIICATIO*
•1 l«VIIIOMII«T*l
•rumo
M10IIIIO1IOI
MTIIIIOMItlflW
Of TUmOMUII
•Mimo
•voioirimot
01 COH1OOI
UUtOVUCUlM
oiniimrroiiciiMcn
^ MPMIIIOIOIIMO
t*o»ioi«iiOHifiiT
-*" LIVUSOr TMV»TOM.M|
*
oniunoioiiioOLlAiniii
MMOKUUHIITIII
IMCTIW
Multi-organ impact of reductions of heme body pool by lead. Impairment of heme
synthesis by lead (see Section 12.3) results in disruption of a wide variety of important physio-
logical processes in many organs and tissues. Particularly well documented are erythropoietic,
neural, renal-endocrine, and hepatic effects indicated above by solid arrows (—»*•). Plausible
further consequences of heme synthesis interference by lead which remain to be more conclu-
sively established are indicated by dashed arrows ( •*-).
-------
IV-3
IV.A. Pathophysiological Effects
Elevated blood lead levels have long been associated with
neurotoxicological effects and many other pathological phenomena:
an article on lead's neurotoxicity was published as early as 1839,
on anemia in the early 1930s, on kidney damage in 1862, and on
impaired reproductive function in 1860. From an historical per-
spective, lead exposure levels considered acceptable for either
occupationally-exposed persons or the general population have been
revised downward steadily as more sophisticated biomedical tech-
niques have shown formerly unrecognized biological effects, and as
concern has increased regarding the medical and social signifi-
cance of such effects. In the most recent downward revision of
maximum safe levels, the Centers for Disease Control (CDC)
lowered its definition of lead toxicity from 30 ug/dl blood lead
and 50 ug/dl of free erythrocyte protoporphyrin (FEP) to 25 ug/dl
blood lead and 35 ug/dl FEP- The present literature shows biologi-
cal effects as low as 10 ug/dl (for heme biosynthesis) or 15 ug/dl
(for renal system effects and neurological alterations).
There is no convincing evidence that lead has any beneficial
biological effect in humans (Expert Committee on Trace Metal
Essentiality, 1983).
The finding of biological effects at blood lead levels as
low as 10 ug/dl potentially has important implications for public
health, as such levels are common in the U.S. population. As
Table IV-1 shows, between 1976 and 1980 over three-quarters of
children under the age of 18 had blood lead levels in excess of
10 ug/dl, and 15 percent exceeded 20 ug/dl. The rates among blacks
-------
IV-4
TABLE IV-1. Blood Lead Levels Of Children in the United States
1976-80 (percent in each cell; rows sum to 100 percent)
<10 ug/dl
10-19
ug/dl
20-29
ug/dl
30-39
ug/dl
40-69
ug/dl
All Races
all ages 22.1 62.9 13.0 1.6 0.3
6 months-5 years 12.2 63.3 20.5 3.5 0.4
6-17 years 27.6 64.8 7.1 0.5 0.0
White
all ages 23.3 62.8 12.2 1.5 0.3
6 months-5 years 14.5 67.5 16.1 1.8 0.2
6-17 years 30.4 63.4 5.8 0.4 0.0
Black
all ages 4.0 59.6 31.0 4.1 1.3
6 months-5 years 2.7 48.8 35.1 11.1 2.4
6-17 years 8.0 69.9 21.1 1.0 0.0
Source: Table 1, Advance Data #79, May 12, 1982, from Vital and
Health Statistics, National Center for Health Statistics (Supple-
mental Exhibit 4.) NOTE: These results were produced after
adjusting the data for age, race, sex, income, degree of urbani-
zation, probability of selection, and non-response to the NHANES
survey.
-------
IV-5
and among preschool children were even higher.
Lead's diverse biological effects on humans and animals are
seen at the subcellular level of organellar structures and pro-
cesses, and at the overall level of general functioning that
encompasses all of the bodily systems operating in a coordinated,
interdependent way. The biological basis of lead toxicity is its
ability, as a metallic cation, to bind to bio-molecular substances
crucial to normal physiological functions, thereby interfering
with these functions. Some specific biochemical mechanisms
involve lead's competition with essential metals for binding for
sites, inhibition of enzyme activity, and inhibition or alteration
of essential ion transport. The effects of lead on certain sub- <
cellular organelles, which result in biochemical derangements
common to and affecting many tissues and organ systems (e.g., the
disruption of heme synthesis processes), are the origin of many of
the diverse types of lead-based functional disruptions of organ
systems.
Lead is associated with a continuum of pathophysiological
effects across a broad range of exposures. In addition to the
high level effects mentioned above, there is evidence that low
blood-lead levels result in:
1. Inhibition of pyrimidine-5'-nucleotidase (Py-5-N)
and delta-aminolevulinic acid dehydrase (ALA-D)
activity, which appears to begin at 10 ug/dl of
blood lead or below (Angle et al., 1982).
Hernberg and Nikkanen (1970) found 50 percent of
ALA-D inhibited at about 16 ug/dl.
2. Elevated levels of zinc protoporphyrin (ZPP or FEP)
in erythrocytes (red blood cells) at about 15 ug/dl.
This probably indicates a general interference in
heme synthesis throughout the body, including inter-
-------
IV-6
ference in the functioning of mitochondria (Piomelli
et al. , 1977) .
3. Changes in the electrophysiological functioning of
the nervous system. This includes changes in slow-
wave electroencephlogram (EEC) patterns and increased
latencies in brainstem and auditory evoked potentials
(Otto et al., 1981, 1982, 1984) which begin to occur
at about 15 ug/dl. The changes in slow-wave EEC
patterns appear to persist over a two-year period.
Also, the relative amplitude of synchronized EEC
between left and right lobe shows effects starting at
about 15 ug/dl (Benignus et al., 1981). Finally,
there is a significant negative correlation between
blood lead and nerve conduction velocity in children
whose blood lead levels range from 15 ug/dl to about
90 ug/dl (Landrigan et al., 1976).
4. Inhibition of globin synthesis, which begins to
appear at approximately 20 ug/dl (White and Harvey,
1972; Dresner et al., 1982).
5. Increased levels of aminolevulinic acid (ALA) in
blood and soft tissue, which appear to occur at
about 15 ug/dl and may occur at lower levels
(Meredith et al., 1978). Several studies indicated
that increases of ALA in the brain interfered with
the gamma-aminobutyric acid (GABA) neurotransmitter
system in several ways (Draft Criteria Document,
p. 12-128 ff).
6. Inhibition of vitamin D pathways, which has been
detected as low as 10 to 15 ug/dl (Rosen et al.,
1980a, 1980b; Mahaffey et al., 1982). Further, as
blood lead levels increased, the inhibition became
increasingly severe.
These data cite the lowest observed effect levels to date, and
do not necessarily represent affirmative findings of thresholds
below which exposures can be considered safe.
The types of specific effects listed above as occurring at
blood lead levels below 25 ug/dl indicate (a) a generalized lead
impact on erythrocytic pyrimidine metabolism, (b) a generalized
lead-induced inhibition of heme synthesis, (c) lead-induced
interference with vitamin D metabolism, and (d) lead-induced
perturbations in central and possibly peripheral nervous system
-------
IV-7
functioning. The medical significance of all these effects is
not yet fully understood. However, current knowledge regarding
the deleterious and vital nature of the affected physiological
functions both individually and in the aggregate suffices to
warrant both public health concern and efforts to minimize
their occurrence due to lead exposure.
As lead exposure increases, the effect on heme synthesis
becomes more pronounced and effects broaden to additional bio-
chemical and physiological mechanisms in various tissues, causing
more severe disruptions of the normal functioning of many organ
systems. At very high lead exposures, the neurotoxicity and
other pathophysiological changes can result in death or, in some
cases of non-fatal lead poisoning, long-lasting sequelae such
as mental retardation and severe kidney disease.
This chapter discusses the known pathophysiological effects
of lead that occur in children, particularly the hematological
and subcellular neurotoxic effects, and the expected change in
the number of children at potential risk of those effects under
EPA's regulation. EPA is considering these effects in its
current National Ambient Air Quality Standard process.
IV.A.I. Effects of Lead on Pyrimidine Metabolism
The best-known effect of lead on erythrocytic pyrimidine
metabolism is the pronounced inhibition of Py-5-N activity. This
enzyme plays a role in the maturation of erythrocytes, as well as
erythrocyte function and survival; it controls the degradation
and removal of nucleic acid from the maturing cell (reticulocyte),
-------
IV-8
Interference with this process can increase red cell membrane
fragility. As noted earlier, the disruption of this function by
lead has been noted at exposure levels beginning from 10 ug/dl.
At blood lead levels of 30-40 ug/dl, this disturbance is suffi-
cient to materially contribute to red blood cell destruction and,
possibly, decreased hemoglobin production contributing to anemia
(World Health Organization, 1977; National Academy of Sciences,
1972; Lin-Fu, 1973; Betts et al., 1973). The significance of
this interference with pyrimidine metabolism transcends the red
cell; the mechanism of this inhibition suggests a wide-spread
impact on all organs and tissues.
IV.A.2. Effects on Heme Synthesis and Related Hematological
Processes
IV.A.2.a. Mitochondrial Effects
The mitochondrion is the critical target organelle for
lead toxicity in a variety of cell and tissue types, followed
probably by cellular and intracellular membranes. The scientific
literature shows evidence of both structural injury to the mito-
chondrion (e.g., Goyer and Rhyne, 1973; Fowler, 1978; Fowler et
al. , 1980; Bull, 1980) and impairment of basic cellular energetics
and other mitochondrial functions (e.g., Bull et al., 1975; Bull,
1977, 1980; Holtzman et al., 1981; Silbergeld et al., 1980).
IV.A.2.b. Heme Synthesis Effects
The best-documented effects of lead are upon heme
biosynthesis. Heme, in addition to being a constituent of hemo-
globin, is an obligatory constituent for diverse hemoproteins in
all tissues, both neural and non-neural. Hemoproteins play
-------
IV-9
important roles in generalized functions, such as cellular ener-
getics, as well as in more specific functions such as oxygen
transport and detoxification of toxic foreign substances (e.g.,
the mixed-function oxidase system in the liver). Available data
on elevated arainolevulinic acid (ALA) and free erythrocyte proto-
porphyrin (FEP) levels, inhibited ALA-D, and the like show inhi-
bition in the heme biosynthetic pathway at low blood-lead levels,
with statistically significant effects detectable at 10-15 ug/dl
(Meredith et al., 1978; Piomelli et al., 1982; Angle et al., 1982).
This heme biosynthetic disturbance may result in the impairment
of many normal physiological processes in a host of organ systems
and/or the reduced reserve capacity of many cells or organs to
deal with other types of stress (e.g., infectious diseases).
The interference of lead with heme synthesis in liver
mitochondria appears to result in the reduced capacity of the
liver to break down tryptophan, which, in turn, appears to
increase levels of tryptophan and serotonin in the brain (Litman
and Correia, 1983). Such elevation of neurotransmitter levels
may be responsible for some of the neurotoxic effects of lead,
since elevated tryptophan levels have been associated with encepha-
lopathy, and elevated serotonin levels produce neurologic symptoms
similar to acute porphyria attacks.
The elevation of ALA levels is another indication of lead's
interference in heme synthesis and mitochondrial functioning.
Such elevations can have serious neurotoxic implications. In
vitro studies have shown that ALA can interfere with several
physiological processes involved in the GABA-ergic neurotrans-
-------
IV-10
mitter system, including a possible role as a GABA-agonist
(Brennan and Cantrill , 1979). There appears to be no threshold
concentration for ALA at the neuronal synapse below which
presynaptic inhibition of GABA release ceases.
Since ALA passes the blood brain barrier and is taken up by
brain tissue, it seems likely that elevated ALA levels in the
blood correspond to elevated ALA levels in the brain (Moore and
Meredith, 1976). Furthermore, lead in the brain is likely to
enhance brain ALA concentrations because neurons are rich in mito-
chondria, the subcellular site of ALA production. As mentioned
earlier, blood ALA elevations begin to be detectable at 18 ug/dl
of blood lead (Meredith et al., 1978).
IV.A.2.C. Impact of Lead on Red Blood Cell Abnormalities
High levels of blood lead ( > 40 ug/dl) are known to produce
anemia. To examine the association between lead and red cell
abnormalities in individuals below 30 ug/dl, we performed two
analyses. First, we examined the relationship between blood
lead levels and various red cell indices. Second, because FEP
is a more stable indicator of a person's lead exposure over
several months than a single blood lead measurement, as well as
their sensitivity to lead, we also analyzed the relationship
between elevated FEP levels and anemia. We found that blood
lead levels below 30 ug/dl were associated with increased risks
of microcytosis and hypochromia, and that FEP levels were
associated with increased risks of anemia in children, even
below 35 ug/dl of FEP.
-------
IV-11
For our analysis we used data from the NHANES II survey.
Among the hematological information collected were mean corpus-
cular volume (MCV), mean corpuscular hemoglobin (MCH), serum
iron, hematocrit, FEP, and percent transferrin saturation. We
used regression analysis of these data for 1,967 children under
the age of eight to determine whether there was a relationship
between blood lead levels and the presence of hematological
abnormalities.
IV.A.2.c.l. Effects of Lead Exposure on Blood Cell Volume
and Hemoglobin Content
We found that blood lead was inversely related to both mean
cell volume (MCV) and mean cell hemoglobin (MCH), even for blood
lead levels below the current CDC 25 ug/dl guideline for deter-
mining lead toxicity. Linear regressions were performed of MCV
and MCH on blood lead levels in children, controlling for race,
age, family-income, iron status (i.e., the level of iron in their
blood), degree of urbanization, and other nutritional factors.
(As previous work led us to expect, percent transferrin saturation
was a superior control for iron status compared to serum iron
and was used throughout our analysis.)
Table IV-2 lists the variables considered in the regressions.
Income was found not to be a significant confounding variable
once we controlled for iron status, and was dropped from the
analysis. Race also had no bearing on MCV once iron status was
controlled for, although it was a significant explanatory variable
for MCH. This suggested that there may be additional dietary or
biochemical factors predisposing black children to lower erythro-
cyte hemoglobin levels.
-------
IV-12
TABLE IV-2. Variables Considered in Regressions of FEP, MCV,
MCH, and Anemia
Age under 2
Age 2-4
Age 4-6
Race
Sex
Degree of Urbanization
Family Income
Serum Albumin
Dietary Calcium
Dietary Calories
Serum Copper
Dietary Carbohydrates
Dietary Fat
Serum Iron
Blood Lead
Dietary Phosphorus
Dietary Protein
Transferrin Saturation
Dietary Vitamin C
Serum Zinc
-------
IV-13
The regressions for both MCV and MCH found blood lead to be
a significant explanatory variable (p < 0.0001 and 0.0033, respec-
tively) for the decreases in each. "Hockey stick" regressions
on the MCV relationship indicated a threshold at 10 ug/dl, the
same level at which heme synthesis disturbance by lead has been
reported to begin. Figures IV-2 and IV-3 show these relationships
for MCV and MCH respectively.
We also analyzed the probability of children having
abnormally low MCV and MCH levels as a function of blood lead,
since this is a clearer sign of physiological derangement. For
this analysis, we used logistic regressions. Once again, blood
lead was a significant explanatory factor (p < 0.0001), both in
mean cell volume being low (MCV < 80 femptoliters [fl]) and in
mean cell volume being seriously low (MCV < 74 fl). Blood lead
levels were also significantly associated (p < 0.023) with the
percent of children having MCH less than 25 picograms (pg), but
only for children under six.
To test the hypothesis that the relationship with MCV held
at low blood-lead levels as well as high blood-lead levels, we
repeated the regression for MCV < 74 fl using only those child-
ren whose blood lead levels were less than 25 ug/dl. The regres-
sion coefficient for blood lead was unchanged and significant
(p < 0.014). Thus, blood lead levels under 25 ug/dl were also
associated with increased risks of microcytosis.
Because blacks have a higher incidence of thalessemia, we
repeated the MCV regressions for white children only. Lead was
-------
IV-lH
FIGURE IV-2
The Relationship Between MCV and Blood Lead After Adjusting
for all Other Significant Variables
82.5
79.5 -
ADJUSTED BLOOD LEAD
The regression line is the true regression for all 1,967 children.
For ease of display, each point represents the average of about
100 children with consecutive blood lead levels.
-------
IV-15
FIGURE IV-3
The Relationship between MCH and Blood Lead After Adjusting
for All Other Significant Variables
MCH (pg/cell)
28.0
27.5 -
27.0 -
26.5
15
20 25
LEAD (jjg/dl)
30
35
The regression line is the true regression line for all 1,967
children. For ease of display, the points shown represent the
average of about 100 children with consecutive blood lead levels.
-------
IV-16
still a significant predictor of MCV (p < 0.0001), after control-
ling for all other significant variables. We also used Mentzer's
index, which indicates that persons with thalessemia generally
have an MCV/red cell count of < 11.5. Even after we eliminated
all such people, lead was still a significant predictor of MCV,
with no change in the coefficient.
To further investigate the relationship between lead and
abnormal hematological variables, we used our regression to
predict the percentage of children with MCV < 74 fl as a function
of blood lead for two cases: children with average transferrin
saturation levels (22.4 percent saturated for children in the
NHANES II survey) and children with transferrin saturation levels
one standard deviation below average (13.6 percent). The results
are shown in Figure IV-4. Note that at 25 ug/dl of blood lead,
almost 10 percent of the children with average iron levels and 17
percent of the children with below average iron levels had MCVs
of less than 74 fl.
The relative risk of children having MCV levels less than
74 fl when their blood lead levels were 25 ug/dl compared to
10 ug/dl was 1.98 (the 95 percent confidence interval was 1.44-
2.71). Using the same 10 ug/dl reference point, the relative
risk at 20 ug/dl was 1.53 (1.27-1.95 at 95 percent). Since
logistic regressions gave the same results when we used only
children with blood lead levels under 25 ug/dl, and since the
95 percent confidence limits on the relative risk did not include
1.0, these results showed increased risks of hematological abnor-
malities in children at blood lead levels of 20 ug/dl and below.
-------
IV-17
FIGURE IV-4
PERCENT OF CHILDREN WITH MCV BELOW 74
(Age 6 Months to 8 Years)
25
20
15
u
"o
10
With Transferrin Saturation
One Standard Deviation Below Average
With Average
Transferrin Saturation
10 15 20 25
Blood Lead Level (pg/dl)
30
35
Prediction of Percent of Children with MCV
Below 74 fl as a Function of Blood Lead Levels
-------
IV-18
IV.A.2.c.2. The Relationship Between Blood Lead and FEP
The increased interference of lead in the formation of
hemoglobin, and consequent accretion of protoporphyrins in red
blood cells, has been well documented by Piomelli et al. (1982).
Our analysis of the NHANES II data confirmed that study's results.
Annest and Mahaffey (1985) have recently analyzed the relationship
between FEP levels and blood lead in the NHANES II data and found
a strong relationship after controlling for iron status. We also
analyzed the NHANES II data and found that, even after controlling
for iron status using transferrin saturation, the relationship
was very strong.
A considerable body of literature exists suggesting that FEP
levels are exponentially related to blood lead levels (Piomelli
et al., 1973; Kammholz et al., 1972; Sassa et al., 1973; Lamola
et al., 1975a, b; Roels et al., 1976). To test this relationship
in the NHANES II population, we used several alternative specifi-
cations. We considered a linear model, a model where FEP was
proportional to both exp(blood lead) and exp(percent transferrin
saturation), a model where FEP was proportional to exp(blood lead)
and (transferrin saturation)3, and a model where FEP was propor-
tional to (blood lead)Bl and (transferrin saturation)B2- The
model that that fit best was exp(blood lead) times (transferrin
saturation)6. We examined the possibility of different additive
intercepts in this model and found the highest correlation coeffi-
cient and F-statistic for a zero additive constant. This model
suggested an exponential relationship to blood lead and a power
law dependence on transferrin saturation. "Hockey stick" analysis
-------
IV-19
of the relationship between PEP and blood lead gave a threshold
of 18 ug/dl. Figure IV-5 shows the FEP-blood lead relationship.
We also investigated the relationship between the probability
of elevated FEP levels and blood lead, and verified previous find-
ings. Again using NHANES II data, we performed logistic regres-
sions on the probability of FEP levels being above 50 ug/dl* as a
function of blood lead, using both blood lead and log(blood lead)
as the independent variable; we obtained a better fit with blood
lead than with log(blood lead).
Again, we checked to see whether the relationship between
the risk of elevated FEP and blood lead held at lower blood-lead
levels, repeating the regression only for children with blood lead
levels under 30 ug/dl. Using multiple logistic regressions, blood
lead was again extremely significant (p < 0.0001). The coeffi-
cient of blood lead for the low group was 0.178 +_ 0.04 compared
to 0.175 + 0.018 for the regression with all blood-lead levels,
a trivial difference between the two cases. This indicated that
the risk of seriously elevated FEP levels was strongly related to
blood lead, even at blood lead levels below the previously defined
"safety level."
Piomelli and coworkers1 studies have suggested a threshold
for lead-induced increases in FEP levels of about 15 ug/dl blood
lead. Taking 17.5 ug/dl of blood lead as our reference level,
* The 50 ug/dl FEP level is considered by CDC to indicate
severe enough interference with heme processes that med-
ical attention is usually required, even when not coupled
with elevated blood-lead levels.
-------
IV-20
Figure IV-5
The Relationship Between PEP Level and Blood Lead
After Adjusting for All Other Significant Variables
k)Q(FEP)
(Mg/
3.8
3.6
3.4
3.0
-I L.
15
20
25
30
LEAD (pg/dl)
35
40
45
The regression line shows the true regression line for all I
individual observations. For ease of display, each point repre-
sents the mean of about 100 children with consecutive blood lead
levels.
-------
IV-21
our regression predicted that the relative risk of FEP levels over
50 ug/dl was 1.55 (1.42-1.70 at 95 percent confidence levels) at
20 ug/dl of blood lead, and was 3.73 (2.55-4.89 at 95 percent
confidence) at 25 ug/dl of blood lead. This was true across all
transferrin saturation levels.
IV.A.2.C.3. The Relationship Between FEP Levels and Anemia
To assess the implications of elevated FEP levels, we analyzed
the relationship between log(FEP) and hematocrit, hemoglobin, and
MCV. We performed linear regressions on all three outcomes as a
function of log(FEP), controlling for race, age, and transferrin
saturation. These analyses showed that log(FEP) was strongly
inversely related (p < 0.0001 in all cases) to hematocrit levels,
hemoglobin levels, and MCV. We then performed logistic regres-
sions on the probability of abnormal levels of hematocrit, hemo-
globin, and MCV as a function of log(FEP), with the same controls.
They also showed that FEP was an excellent predictor (p < 0.0001)
of the probability of abnormally low levels of all three indica-
tors. Again, we repeated our regressions using only children with
FEP values below 33 ug/dl, and FEP was still very significant
(p < 0.0001). (FEP levels of less than 33 ug/dl are generally
associated with blood lead levels under 30 ug/dl.) The coeffi-
cients differed by less than one standard deviation from those for
the full sample. Thus, the relationship appeared to hold for low
FEP levels as well as high ones.
Figure IV-6 shows the regression's prediction of the percent
of children with anemia as a function of FEP levels at normal
transferrin saturation levels for children. The data used in the
-------
IV-22
regression contained FEP levels as low as 9.6 ug/dl, but we have
shown the projections only for 18 ug/dl and above. For our
definition of anemia, we have used the hemoglobin and hematocrit
levels representing the minimum normal range levels recommended
by the Journal of Pediatrics (1977). These definitions are
supported by the work of Yip et al. (1981), and the use of hema-
tocrit and hemoglobin levels to define anemia is standard
(Harrison, Principles of Internal Medicine, 9th Edition).
Figure IV-6 shows that as FEP levels increase from 20 ug/dl
to 50 ug/dl, an additional 20 percent of children aged 2-6 years
would develop anemia at normal iron levels. Our earlier regres-
sions of blood lead levels on FEP suggested that blood lead levels
of less than 15 ug/dl were necessary to keep average FEP levels
below 20 ud/dl. Since elevated FEP is a symptom of an underlying
interference in heme synthesis, it cannot be viewed as the cause
of these abnormal hematocrits. The causal association must be
with whatever produced the excess FEP. As the data portrayed in
Figure IV-3 were for normal iron levels, the anemia appeared to
be the outcome of the lead exposure, for which FEP served as a
surrogate.
In summary, blood lead levels below 30 ug/dl seem to be
associated with increased microcytosis and hypochromia in children
and increased interference with heme synthesis producing elevated
levels of free erythrocyte protoporphyrin. Elevated FEP levels,
even below 33 ug/dl, are themselves associated with an increased
risk of anemia. In addition, the reduced mean cell volumes and
the lower hematocrit values again indicate that lead's effect on
-------
IV-2 3
FIGURE IV-6
Prediction of Percent of Children with Anemia as a Function
of FEP Level at Normal Transferrin Saturation Levels
PERCENT OF CHILDREN WITH ANEMIA
By Age and Race at Average Transferrin Saturation Levels
12-5
FEP Level (/ig/dl)
-------
IV-24
heme synthesis continues at levels below 30 ug/dl of blood lead.
This further strengthens the case for considering elevated FEP
levels themselves, which mark lead's interference with normal
body activity, as a pathophysiological effect.
IV.A.3. Lead's Interference with Vitamin D Metabolism and
Associated Physiological Processes
Another potentially serious consequence of lead exposure is
the impairment of the biosynthesis of the active vitamin D meta-
bolite, 1,25-(OH)2 vitamin D, which is detectable at blood lead
levels of 10-15 ug/dl. Further, an inverse dose-response rela-
tionship has been reported between blood lead and 1,25-(OH)2
vitamin D throughout the range of measured blood lead values up
to 120 ug/dl (Criteria Document, p. 12-49; Rosen et al., 1980a,
1980b; Mahaffey et al., 1982). Interference with vitamin D pro-
duction disrupts calcium, zinc, and phosphorous homeostasis,
partially resulting in the reduced absorption of these elements
from the gastro-intestinal tract. This alters the availability
of these elements for physiological processes crucial to the
normal functioning of many tissues, cell membranes, and organ
systems.
The reduced uptake and utilization of calcium has two
compounding conseguences. First, it interferes with calcium-
dependent processes that are essential to the functioning of
nerve cells, endocrine cells, muscle cells (including those in
the heart and other components of the cardiovascular system),
bone cells, and most other types of cells. The second concern
is possible increased lead absorption resulting from decreased
-------
IV-2 5
calcium availability. The latter can be expected to create a
feedback response, further exacerbating the inhibition of vitamin
D metabolism and reduced calcium availability, leading to even
greater lead absorption and greater vulnerability to increasingly
more severe lead-induced health effects (Rosen et al., 1980b;
Barton et al., 1978). These effects are especially dangerous
for young (preschool age) children who are developing rapidly.
These children, even in the absence of lead, generally are
susceptible to calcium deficiencies because of the large amount
of calcium used for the formation of the skeletal system, as
well as several other calcium-dependent physiological processes
important in young children.
Even moderate levels of lead exposure in children are
associated with vitamin D disturbances that parallel certain meta-
bolic disorders and other disease states, as well as severe kidney
dysfunction (Criteria Document, p. 12-37). At blood lead levels
of 33-55 ug/dl, 1,25-(OH)2 vitamin D is reduced to levels compar-
able to those observed in children who have severe renal insuffi-
ciency with the loss of about two-thirds of their normal kidney
function (Rosen et al., 1980a; Rosen and Chesney, 1983; Chesney
et al., 1983). Analogous vitamin D hormone depressions are found
in vitamin D-dependent rickets (type I), oxalosis, hormone-defi-
cient hypoparathyroidism, and aluminum intoxication in children
undergoing total parenteral nutrition.
Lead-induced interference of 1,25-(OH)2D biosynthesis
affects a wide range of physiological processes. The vitamin D-
endocrine system is responsible in large part for the maintenance
-------
IV-2 6
of extra- and intra-cellular calcium homeostasis (Rasmussen and
Waisman, 1983; Wong, 1983; Shlossman et al., 1982; Rosen and
Chesney, 1983). Thus, modulation in cellular calcium metabolism
induced by lead at relatively low concentrations may potentially
disturb multiple functions of different tissues that depend upon
calcium as a second messenger (Criteria Document, p. 12-40). It
also appears that 1,25-(OH)2D participates directly in bone
turnover by orchestrating the population of cells within the
bone (Criteria Document, p. 12-41). An immunoregulatory role
for the vitamin D hormone is evident through the widespread exist-
ence of 1,25-(OH)2^3 receptor sites on immunoregulatory cells,
such as monocytes and activated lymphocytes (Provvedini et al. ,
1983; Bhalla et al., 1983).
The negative correlation between blood lead and serum
1,25-(OH)2D, the active form of vitamin D, appears to be another
example of lead's disruption of mitochondrial activity at low
concentrations. While serum levels of 1,25-(OH)2 vitamin D
decreased continuously as blood lead levels increased from an
apparent threshold of 10 to 15 ug/dl, this was not true for its
precursor, 25-(OH) vitamin D. In fact, in lead-intoxicated
children after chelation therapy, vitamin D levels were restored,
but the precursor levels remained unchanged (Rosen et al., 1980a,
1980b; Mahaffey et al., 1982). This indicates that lead inhibits
renal 1-hydroxylase, the kidney enzyme that converts the precursor
to the active form of vitamin D. Renal 1-hydroxylase is a
mitochondrial enzyme system, which is mediated by the hemoprotein,
cytochrome P-450. This suggests that the damage to the mitochon-
-------
IV-2 7
drial systems detected at 15 ug/dl has uncompensated consequences.
If cytochrome P-450 is being inhibited at the low levels of
blood lead that the reduced renal 1-hydroxylase activity suggests,
we must consider the possibility that other physiological func-
tions related to cytochrome P-450 may also be disrupted. For
example, reduced P-450 content has been correlated with impaired
activity of the liver detoxifying enzymes, aniline hydroxylase
and aminopyrine demethylase, which help to detoxify medications,
hormones, and other chemicals (Goldberg et al., 1978).
While cytochrome P-450 inhibition has been found in animals,
and in humans at higher lead levels, this has not yet been examined
in children at low blood-lead levels (i.e., 10 to 15 ug/dl). But
the disruption of vitamin D biosynthetic pathways at these levels
is suggestive of an effect.
The reduction in heme caused by lead exposure probably
underlies the effects seen in vitamin D metabolism. This would
explain the similarity in apparent "thresholds" for the effect of
lead on both erythrocyte protoporphyrin accumulation and decreases
in levels of serum 1,25-(OH)2D- It would also indicate a cascade
of biological effects among many organ and physiological systems
of the body (depicted graphically in Figure IV-1). Together, the
interrelationships of calcium and lead metabolism, lead's effects
on 1,25-(OH)2D, and the apparent disruption of the cytochrome
P-450 enzyme system provide a single molecular and mechanistic
basis for Aub et al.'s observation in 1926 that "lead follows
the calcium stream."
-------
IV-28
IV.B. Neurotoxic Effects of Lead Exposure
Lead has been known to be a neurotoxicant since the early
1800s, and neurotoxicity is among the more severe consequences of
lead exposure. At very high-blood lead levels, encephalopathy
and severe neurotoxic effects are well documented; the neurotoxic
effects at lower blood lead levels, however, are less clearly
defined. Recent research has investigated the occurrence of
overt signs and symptoms of neurotoxicity and the manifestation
of more subtle indications of altered neurological functions in
individuals who do not show obvious signs of lead poisoning.
IV.B.I. Neurotoxicity at Elevated Blood-Lead Levels
Very high blood-lead levels (i.e., above 80 ug/dl in children)
are associated with massive neurotoxic effects that can include
severe, irreversible brain damage, ataxia (i.e., the inability to
coordinate voluntary muscular movements), persistent vomiting,
lethargy, stupor, convulsions, coma, and sometimes death. Once
encephalopathy occurs, the risk of death for children is signifi-
cant (Ennis and Harrison, 1950; Agerty, 1952; Lewis et al. , 1955),
regardless of the quality of the medical treatment they receive.
In cases of severe or prolonged nonfatal episodes of lead
encephalopathy, neurological damage occurs that is qualitatively
similar to that often seen following traumatic or infectious
cerebral injury, with permanent and irreversible damage being
more common in children than adults (Mellins and Jenkins, 1955;
Chisolm, 1956, 1968). The most severe effects are cortical
atrophy, hydrocephalus (an abnormal increase in cranial fluid),
convulsive seizures, and severe mental retardation. Permanent
-------
IV-2 9
central nervous system damage almost always occurs in children
who survive acute lead encephalopathy and are re-exposed to
lead (Chisolm and Harrison, 1956) . Even if their blood lead
levels are kept fairly low, 25-50 percent show severe permanent
sequelae including seizures, nervous disorders, blindness, and
hemiparesis (paralysis of half of the body) (Chisolm and Barltrop,
1979) .
Even children without obvious signs of acute lead encephalo-
pathy have exhibited persisting neurological damage. As early
as 1943, Byers and Lord's study of 20 previously lead-poisoned
children indicated that 19 later performed unsatisfactorily in
school, "presumably due to sensorimotor deficits, short attention
span, and behavioral disorders". Effects such as mental retarda-
tion, seizures, cerebral palsy, optic atrophy, sensorimotor
deficits, visual-perceptual problems, and behavior disorders have
been documented extensively in children following overt lead
intoxication or even just known high exposures to lead (e.g.,
Chisolm and Harrison, 1956: Cohen and Ahrens, 1959; Perlstein
and Attala, 1966) .
The extent of the later manifestations seems to relate to
the severity of the earlier observed symptoms. In Perlstein and
Attala, 9 percent of the children studied, none of whom appeared
to have severe symptoms when diagnosed for overt lead poisoning,
were later observed to be minimally mentally retarded and 37
percent showed some lasting neurological sequelae.
At somewhat lower blood-lead levels (i.e., 30-70 ug/dl),
substantial data confirm that a variety of neural dysfunctions
-------
IV-30
occur in apparently asymptomatic children. Several studies
indicate that blood lead levels of 50-70 ug/dl are associated
with IQ decrements of 5 points. Adverse electrophysiological
effects, including markedly abnormal EEC patterns, slow-wave
voltages, etc., are also well documented at levels of 30-70
ug/dl.
De la Burde and Choate (1972, 1975) showed persisting
neurobehavioral deficits in children exposed to moderate-to-high
levels of lead; most of the children appear to have had blood
lead levels above 40 ug/dl. Compared to low-lead control
children — matched for age, sex, race, parents' socioeconomic
status, housing density, mother's 10, number of children in the
family below age 6, presence of father in the home, and mother
working -- the higher lead children averaged about five points
lower in 10 and were seven times more likely to have repeated
grades in school or to have been referred to school psychologists.
Moreover, follow-up studies showed that these effects persisted
for at least three years.
While chelation therapy may mitigate some of these persisting
effects, permanent neurological and cognitive damage seems to
result from very high lead levels, with or without encephalopathy.
In addition, these children also appear more likely to experience
neurological and behavioral impairments later in their childhood.
IV.B.2. Neurotoxicity at Lower Blood-Lead Levels
The adverse effects of lead on neurological functioning,
both on the microscopic (i.e., cellular and enzymatic) level and
the macroscopic (i.e., learning behavior) level, are well docu-
-------
IV-31
merited. On the micro-level, data from experimental animal studies
suggest several possible mechanisms for the induction of neural
effects, including: (1) increased accumulation of ALA in the
brain as a consequence of lead-induced impaired heme synthesis,
(2) altered ionic balances and movement of ions across axonal
membranes and at nerve terminals during the initiation or conduc-
tion of nerve impulses due to lead-induced effects on the meta-
bolism or synaptic utilization of calcium, and (3) lead-induced
effects on the metabolism or synaptic utilization of various
neurotransmitters.
In addition, lead-induced heme synthesis impairment,
resulting in reduced cytochrome C levels in brain cells during
crucial developmental periods, has been clearly associated with
the delayed development of certain neuronal components and
systems in the brains of experimental animals (Holtzman and
Shen Hsu, 1976). (Cytochrome C is a link in the mitochondrial
electron transport chain that produces energy, in the form of
adenosine triphosphate (ATP), for the entire cell.) Given the
high energy demands of neurons, selective damage to the nervous
system seems plausible.
In addition to the effects of lead on the brain and central
nervous system, there is evidence that peripheral nerves are
affected as well. Silbergeld and Adler (1978) have noted lead-
induced blockage of neurotransmitter (acetylcholine) release in
peripheral nerves, a result of lead's disruption of the transport
of calcium across cellular membranes. This disruption of cellular
calcium transport may also contribute to the effects of lead on
-------
IV-3 2
peripheral nerve conduction velocity. Landrigan et al. (1976)
have noted a significant correlation between blood lead and
decreasing nerve conduction velocity in children in a smelter
community. This effect may indicate advancing peripheral
neuropathy.
Paralleling these cellular or biochemical effects are
electrophysiological changes indicating the perturbation of peri-
pheral and central nervous system functioning observed in children
with blood lead levels of approximately 15 ug/dl. These included
slowed nerve conduction velocities (Landrigan et al., 1976),
reaction-time and reaction-behavior deficits (Winneke et al.
1984; Yule, 1984), as well as persistent abnormal EEC patterns
including altered brain stem and auditory evoked potentials
down to 15 ug/dl (Benignus et al. , 1981; Otto et al., 1981, 1982,
1984). The results indicating neurological effects of lead at
such low levels are particularly important because two- and
five-year follow-up studies (Otto et al., 1982, 1984) indicated
some persistent effects.
Aberrant learning behavior has been noted in rats with
blood lead levels below 30 ug/dl. This behavior evidenced both
reduced performance on complex learning problems and signs of
hyperactivity and excessive response to negative reinforcement
(Winneke, 1977, 1982a).
Finally, the cognitive effects of lead in children show
signs of a dose-response relationship. For high level lead
poisoning, adverse cognitive effects in children are indisputable
and mental retardation is a common outcome. For children with
-------
IV-33
somewhat lower blood-lead levels, de la Burde and Choate (1972,
1975) found lesser but still significant cognitive effects,
including lower mean IQs and reduced attention spans. Several
studies discussed in more detail in Section IV.B.2.a. have found
smaller effects at lower blood-lead levels. The precise biolo-
gical mechanisms connected with these effects are not yet clearly
defined, although hypotheses have been put forward.
While some of these effects have only been observed at
higher blood-lead levels, in animals, or in vitro, they show a
consistent dose-dependent interference with normal neurological
functioning. Furthermore, some of these effects have been docu-
mented to occur at low blood-lead levels in children, with no
clear threshold having been demonstrated.
This general pattern of lead's interference in neurological
functioning on the cellular level, including effects below 30
ug/dl, form the background against which we examined the studies
that investigated changes in cognitive processes in children
at low blood-lead levels. Because of the intrinsic difficulties
in performing such studies, and because most investigators have
not employed sample sizes that would permit unambiguous detection
of small effects, it is important to integrate these behavioral
studies with what has been discovered on the molecular and cellu-
lar levels.
IV.B.2.a. Cognitive Effects of Moderate Blood-Lead Levels
The literature on cognitive effects at low to moderate body-
lead levels is extensive. However, most of the studies have
some methodological flaws, and few display indisputable results
-------
IV-3 4
concerning the relationship between IQ effects and changes in
low body-lead levels. The Draft Lead Criteria Document (p. 12-65)
divided the studies into four groups: clinical studies of high
lead children, general population studies, lead smelter area
studies, and studies of children who are mentally or behaviorally
abnormal.
One of the larger and better designed studies of lower-level
cognitive effects was the Needleman et al. study in 1979, which
found a significant inverse correlation between tooth lead levels
and IQ after controlling for age, parent's 10, and socioeconomic
factors. In 1983 an EPA peer-review panel asked for a reanalysis
of Needleman et al.'s data, using different model specifications.
EPA's Office of Policy Analysis also requested a reanalysis us-
ing a continuous lead exposure variable. Needleman submitted a
reanalysis to EPA, and presented the reanalysis to the Clean Air
Science Advisory Committee (CARAC) in 1984. The CASAC stated
that the reanalysis adequately addressed the concerns of the peer
review panel, and recommended that EPA include the study in its
criteria document process. These latter results confirmed the
association between lead and 10 found in the original study,
and showed that the results were quite stable in response to
the inclusion or exclusion of different confounding factors.
The summary table in Chapter 12 of the Criteria Document
(pp. 12-65 to 12-70) indicated that virtually all of the general
population studies showed high lead groups performing more
poorly on a variety of tests used to assess cognitive function.
For more than half of these tests, however, the probability of
-------
IV-3 5
falsely finding an effect due to chance was more than 5 percent;
i.e., less than half of them had a p-value of less than 0.05.
(Significance levels in the studies were reported as probabilities
if they were below 0.05 and as "not significant" otherwise.)
For use in public policy making, rejecting the results of
these studies because so many fail to attain significance at the
5 percent level may be inappropriate for two reasons. First,
policy makers need to be concerned about both type I and type II
errors. Significance tests guard only against the first type
(falsely rejecting the null hypothesis of no effect); they help
ensure that a regulation is not imposed when there is no adverse
effect. Type II errors (failing to reject the null hypothesis
when it is false) also can be costly, however, because they can
result in the underregulation of a real hazard. With small sample
sizes and subtle effects, the probability of a type II error can
be large; in the case of the Smith et al. (1983) study, we
calculated it to be 62 percent if the true decrease in IQ was
2 points.* The probability of a type II error in the other
studies would be even higher, because of their smaller sample
sizes.
* We computed the false positive from the Smith et al. data
as follows. Using Pocock and Ashby (1985), we derived the
standard deviation for the difference of the high and low
lead groups of 1.499. At a 5 percent chance of rejecting the
null hypothesis when it is true, the normal one-tailed test
statistic is 1.65. Therefore, we would reject the null
hypothesis only for differences greater than (1.499) (1.65)
= 2.473. If the difference in the groups was two 10 points,
the probability of the difference being below 2.473 is
given by p (z < [2.47S-2]/ 1.499) = 0.62.
-------
IV-3 6
The second reason for caution in rejecting the results of
these studies is that while several fail to attain statistical
significance individually, they do show a consistent pattern: in
nearly all of them, the children in the higher lead groups showed
lower mean IQs. Figure IV-7 plots the estimated effects, along
with the 90 percent confidence limits.* The higher end of the
90 percent confidence limit corresponds to the critical value for
a one-tailed test at the 0.05 significance level; i.e., studies
in the figure whose upper confidence limits exceed 0 are not
statistically significant at the p = 0.05 level.
The consistent pattern in all of the studies suggested that
combining evidence from all studies would provide a better test
for a significant effect than separate evaluations of the statis-
tical significance of the individual studies. In applying one
of the available joint tests for the existence of a specific
effect, we began with the six general population studies found
in Table 12.1 of the draft Criteria Document. As the result of
personal communication with Harvey we have not included this
study, due to the younger age of the children (2.5) and the
continuous nature of their study design. We did not consider
clinic studies because of their higher lead (typically > 70 ug/dl)
or studies of children exhibiting abnormal behavior. To the five
remining general population studies we added the smelter study
by Winneke et al. (1982b), as the blood lead levels in that study
* Some of the studies did not report p-values or standard errors,
in which cases, we used the data in the study to compute these
values. Where we had insufficient information, we did not
include the study.
-------
FIGURE IV-7.
Mean IQ Difference Between High Lead Groups and Controls,
Adjusted for Socioeconomic Factors (90% Confidence Intervals)
•*•
2 -
5 o
o
wX
-16 -
1 R
f
1
,
.
•
.
.
1
.
•4-
«
' \
1 O 1 1 1 I 1
^Bride Yule Smith Yule and Winneke Needlonan
et al. et al. et al. Lansdown et al. et al.
(1982) (1981) (1983) (1983) (1982) (1979)
-------
IV-3 8
were in the same range as the general population studies. For
three other studies, Winneke (1983), Winneke (1984), and Yule and
Lansdown's study in Leeds, sufficient data were not available to
include their results. These three studies generally found small
or statistically insignificant effects. We also looked only at
Full Scale IQ measures. While not all studies used the same 10
test, the Full Scale IQ measures employed were close enough to
allow us to compare differences between groups and across
studies. Table IV-3 summarizes the relevant information.
In evaluating the reported results of the combined
significance of these studies, it is important to remember that
because each study was performed using different protocols,
study populations, levels of exposure and investigations, the
resulting combined p-value should be interpreted cautiously,
and viewed more for its qualitative implications than for its
precise numerical result. The specific numerical result (joint
p-value) also is sensitive to the studies included. Inclusion
of the studies by Winneke and others discussed above, which were
omitted because of insufficient data, would change the p-value.
The direction of the change is difficult to predict, however,
because while these studies generally found insignicant results,
the inclusion of additional studies will raise the joint p-value
unless the p-values in the individual studies are quite large
(e.g., p > 0.25) .
We used the Fisher aggregation procedure (Fisher, 1970,
p. 99) to estimate the combined significance of the observed
-------
TABLE IV-3. Computation of Joint P-Value from Epidemiological Studies of Cognitive Effects
from Low Level Lead Exposure in Children
Study
McBride et al.
(1982)
/Yule et al.
(1981)
Smith et al.
(1983)
Yule and Lansdown
(1983)
Winneke et al.
(1982a)
Needleman et al.
(1979)
Joint p-value
Internal Lead Levels
Sample Sizes Blood (ug/dl) Teeth(ppm)
Control Exposed Control Exposed Control Exposed
86 86 < 9 19-30
20 21 7-10 17-32
1AQ 1 Rl^ / 1 C N Q
IflD 1DD — — — — — x Z.D > o
80 82 7-12 13-24
26 26 2.4 7
100 58 < 10 > 20
for studies: P(X2? > 31.31) <0.005
10 Difference P-Value -2 In p
1.2a 0.25 2.77
7.6b 0.027 7.22
2.3b 0.067C 5.41
1.8*> 0.13 4.08
5b 0.10 4.82
4.5b 0.03 7.01
31.31
a peabody Picture Vocabulary IQ lest
b Welchsler Intelligence Scale for Children-Revised
c Smith does not report a p-value but Pocock and Ashby report a 95% confidence interval of .4 to -5.5
for the Smith result which implies a standard error of 1.5 and a t-statistic of 1.53.
LO
10
-------
IV-40
effects, and to derive a joint p-value for all of the studies.
To do this, we needed the p-values for all of the individual
studies.
For each study where p-values were not reported, we used
the standard deviation of the IQ measure to compute the p-value
for the difference in the mean IQs across groups. We could not
use this method for the 1983 study by Smith et al. In that
study, the full scale IQ effects were reported as "not significant1
and no standard deviation was given. However, when we computed
the p-value using the standard deviation derived from Pocock and
Ashby, we found that the p-value was 0.067 when comparing high
and low lead groups for the Full Scale IQ.*
The results of our application of the Fisher procedure for
computing a joint probability for the observed results are pre-
sented in Table IV-3. The resulting probability of less than
0.005 indicates that it was extremely unlikely that we could
get the observed pattern of results if there were really no
effect. The overwhelming preponderance of the data (virtually
all studies show high lead groups with lower cognitive ability)
was highly unlikely to have been due to chance.
Only if the studies were consistently biased towards finding
an effect would the robustness of our result be questionable. In
* Pocock and Ashby reported that in the Smith et al. study
the 95% confidence interval for the full scale IQ effect
was -5.5 to +0.4. The t-statistic for 298 degrees of freedom
is 1.9679, yielding a standard deviation of 1.499. Our
calculation used the difference of 2.3 reported by Smith
et al., rather than the difference of 2.55 implied by the
recent work by Pocock and Ashby.
-------
IV-41
at least one case (Smith), a procedure was used that biased against
finding an effect, and biased upward the p-values. These authors
used a two-stage analysis of variance or covariance where the
effects of all covariates (except lead) on IQ were controlled for
in the first stage, and the only residual IQ effects were regressed
on lead in the next step. Many of these covariates (e.g., parental
care, income, and IQ) negatively correlate with lead exposure, and
this procedure attributed all of the joint variation to the non-
lead variable.
These facts, together with the very small p-value calculated
in the joint test, suggest that the combined evidence of cognitive
effects at moderate levels of lead exposure should be treated as
statistically significant. We conclude that the combined results
of available studies of cognitive effect at moderate lead levels
should be taken as evidence of cognitive decrements due to lead.
IV.B.3. The Magnitude of Lead's Impact on IQ
The evidence described above indicates that exposure to lead
can lower children's IQs and reduce their ability to perform well
in school. In Section IV.D, we monetize the cognitive benefits
of reducing these effects using the costs of compensatory educa-
tion. Here, we briefly describe a more direct, but also more
speculative, approach based on the improvements in IQ that might
be expected with reduced exposure to lead in gasoline.
The latest draft of the Criteria Document characterizes the
evidence as suggesting that, on average, blood lead levels of 30
to 50 ug/dl result in a four-point decrement in IQ, and that lead
-------
IV-42
levels of 50 to 70 ug/dl reduce 10 by roughly five points
(de la Burde and Choate, 1972, 1975; Rummo et al., 1979). If we
assume that preventing a blood-lead level over 30 ug/dl avoids,
on average, the loss of four 10 points per child, the gain in
person-IO points from limiting lead in gasoline is substantial.
In 1986, for example, as shown in Table III-7, we estimate that
52,000 fewer children will experience blood lead levels over 30
ug/dl as a result of the 0.10 gplg limit (assuming no misfueling).
If we assume that each child over 30 ug/dl suffers roughly a four-
point 10 loss, that implies that the final rule will yield a gain
of about 200,000 person-IQ points in 1986. Table IV-4 presents
year-by-year estimates for the alternative rules.
This approach suffers from two faults, which cut in opposite
directions. It does not account for the fact that some children
who are prevented by the regulation from going over 30 ug/dl will
do so by a narrow margin (e.g., their blood lead level will be
29 ug/dl when it would have been 31 ug/dl in the absence of the
rule); such children are unlikely to receive the full four point
gain in 10. On the other hand, this approach attributes no bene-
fit to children whose blood lead levels are reduced from very high
levels, but not brought below 30 ug/dl, or to those whose levels
would have been under 30 ug/dl without the rule, but whose levels
decrease further by the reduction in lead in gasoline.
IV.C. Fetal Effects
Because lead passes the placental barrier, a growing
concern in the public health community is that the most sensitive
-------
IV-43
TABLE IV-4. Year-by-Year Estimates of Gain in Person-IQ Points Under Alternative
Rules, Assuming No Misfueling (thousands of person-IQ points)
Rule 1985 1986 1987 1988 1989 1990 1991 1992
Proposed 0 208 188 172 156 144 128 124
Alternative 88 184 180 172 156 144 128 124
Final 88 208 188 172 156 144 128 124
-------
IV-44
population for lead exposure is not children, but fetuses and
newborn infants. This concern is supported by both animal
studies and, recently, human data in the published peer-reviewed
literature.
Crofton et al. (1980) found that the development of
exploratory behavior by rat pups exposed to lead in utero lagged
behind that of control rats. Average blood lead levels on the
21st postnatal day were 14.5 ug/dl for the exposed pups and 4.8
ug/dl for the controls.
Gross-Selbeck and Gross-Selbeck (1981) found alterations in
the operant behavior of adult rats after prenatal exposure to
lead via mothers whose blood lead levels averaged 20.5 ug/dl. At
the time of testing (3 to 4 months, postnatal), the lead-exposed
subjects' blood lead levels averaged 4.55 ug/dl compared to 3.68
ug/dl in the controls. This suggested that changes in central
nervous system function may persist for months after the cessation
of exposure to relatively low blood-lead levels.
Several other papers (McCauley and Bull, 1978; Bull et al.,
1979) have shown that the prenatal exposure of rats to 0.2 percent
lead chloride in the mother's drinking water markedly reduced the
cytochrome C content in the cerebral cortex, thereby possibly
producing an uncoupling of the electron transport chain in the
cortex. This reduction in cytochrome C content occurred at blood
lead levels as low as 36 ug/dl, with delays in the development of
central nervous system energy metabolism being seen as late as 50
days after birth (Bull et al., 1983).
-------
IV-45
Needleman et al. (1984) analyzed data from over 4,000 live
births at Boston Women's Hospital and reported an association
between mild congenital anomalies and umbilical-cord blood-lead
malformation and lead, but only between all minor malformations
and lead. There also were no significant associations between
lead and any major malformations, although given the rate of such
malformations in the general population, a sample this size has
little power to detect such an effect. Holding other covariates
constant, the relative risk of a child demonstrating a minor
malformation at birth increased by 50 percent as lead levels
increased from 0.7 ug/dl to 6.3 ug/dl (the mean cord-lead level).
This risk increased an additional 50 percent at 24 ug/dl.
(Umbilical-cord blood-lead levels are somewhat lower than, but
correspond to, maternal blood-lead levels; Lauwerys et al., 1978.)
A recent analysis by Bellinger and coworkers (1984) also
found an association between increasing cord-lead levels and
deficits in the child's subsequent performance on the Bayley
development scales, after controlling for covariates. Again,
the mean cord-lead levels in this study were very low (under 10
ug/dl) .
Finally, Erickson et al. (1983) found lung- and bone-lead
levels in children who died from Sudden Infant Death Syndrome
were significantly higher (p < 0.05) than in children who died
of other causes, after controlling for age. While this study
suggests a potential relationship between lead and Sudden Infant
Death Syndrome, this issue remains to be more fully evaluated.
-------
IV-46
In addition, lead has been implicated in complications of
pregnancy, including early and still births. Fahim et al. (1976)
found that women who had normal full-term pregnancies had average
blood-lead levels of 14.3 ug/dl, whereas women with early membrane
rupture had average blood-lead levels of 25.6 ug/dl, and women
with premature delivery had average blood-lead levels of 29.1
ug/dl. Wibberly et al. (1977) found that higher lead levels in
placental tissues were associated with various negative pregnancy
outcomes, including prematurity, birth malformation, and neonatal
death.
Bryce-Smith et al. (1977) found bone lead concentrations
in still births of 0.4-24.2 ppm in the rib (average: 5.7) versus
typical infant bone lead levels of 0.2-0.6 ppm.
To assess the effect of EPA's current rulemaking on fetal
exposure, we performed logistic regressions of the probability of
adults (over 15 years) from the NHANES II survey having blood lead
levels above 30 ug/dl. The regression variables were selected
by a stepwise logistic regression procedure that chose all varia-
bles that were significant at the p = 0.05 level. We then used
these regressions to predict the percent of women who would be
above 25 ug/dl in 1986-1992, under the final rule. We multiplied
the change in the percent of women of child-bearing age by the
expected total number of live births each year to estimate the
change in the number of fetuses born to mothers exposed to more
than 25 ug/dl of blood lead; this is shown on Table IV-5. (The
regression coefficients are included in Appendix C.)
-------
IV-4 7
TABLE IV-5. Estimated Decrease in the Number of Fetuses Exposed
in Utero to > 25 ug/dl of Blood Lead
1985 1986 1987 1988 1989 1990 1991 1992
1,275 3,800 3,200 2,800 2,550 2,200 2,000 1,900
IV.D. Monetized Estimates of Children's Health Benefits
The health benefits of reducing children's exposure to lead
are diverse and difficult to estimate quantitatively or to value
in monetary terms. To monetize the benefits, we focused on two
admittedly incomplete measures: savings in expenditures for
medical testing and treatment, and savings in compensatory educa-
tion. These measures of benefit exclude many important factors,
such as reduced pain and suffering, or higher earnings in
later life.
In fact, many children with elevated blood-lead levels are
neither detected nor treated. However, our estimation procedure
assumes that children who go undetected and untreated bear a
burden at least as great as the cost of testing, treating, and
providing compensatory education for those who are detected. So,
all children with high blood-lead levels are assumed to incur
"costs", whether medical expenditure costs or personal costs in
the form of poor health, inadequate learning, etc.
IV.D.I. Reduced Medical Costs
To estimate the benefits of reduced medical care expenses,
we assumed that children with elevated blood lead levels would
receive the treatment recommended by Drs. Piomelli, Rosen, Chisolm,
-------
IV-4 8
and Graef in a recent article in the Journal of Pediatrics (1984).
Those four leading experts in the clinical treatment of lead
toxicity combined their data and clinical experience to develop
optimal diagnosis, treatment, and follow-up protocols. They also
estimated the percentages of children at different blood-lead
levels who would require various types of treatment. Figure IV-7
summarizes the treatment options that we considered, based on
the recommendations of Piomelli et al.
We assumed that administrative expenditures and follow-up
tests would cost $100 for every child found to be over 25 ug/dl
at screening. Of those children over 25 ug/dl blood lead, we
estimated, based on Piomelli et al. (1982) and Mahaffey et al.
(1982), that 70 percent would be over 35 ug/dl FEP. Piomelli
et al. (1984) recommend provocative ethyleneamentetraacetic acid
(EDTA) testing for such children. EDTA testing typically requires
a day in the hospital and a physician's visit; we assumed a cost
of $500 per test. We also assumed that all children receiving
EDTA testing would receive a series of follow-up tests and physi-
cians' visits, at a combined cost of $300.
The purpose of EDTA testing is to see if children have a
dangerously high body-lead burden (a lead excretion ratio over
0.60, per Piomelli et al.). Table IV-6 presents Piomelli et al.'s
estimates of the percentages of children at various blood lead
levels who will require chelation therapy; it ranges from a low of
zero for those under 30 ug/dl to a high of 100 percent for those
over 59 ug/dl.
-------
IV-49
FIGURE IV-8
Flow Diagram for Children with Blood Lead Levels above 25 ug/dl
Blood Lead Levels
>25 ug/dl
elevated
FEP?
no
yes
simple
follow-up
no
hi<
leac
high body
lead burden?*
I
yes
1
long
follow-up
chelation
therapy
I
repeat
chelation?
no
yes
1
long
follow-up
chelation
therapy
NOTES:
*Provocative EDTA or
other test
tChelation therapy, because
of its severe side-effects
and inherent dangers, cannot
be repeated again after this
point
I
repeat
chelation?
no
1
yes
long
follow-up
chelation
therapy
long
follow-upt
-------
IV-50
TABLE IV-6. Percent of Children Requiring Chelation Therapy
Blood Lead Levels Percent
25-30 ug/dl 0
30-39 ug/dl
age three and over 9.6
age under three 11.5
40-49 ug/dl
age three and over 26.0
age under three 37.9
50-59 ug/dl
age three and over 36.0
age under three 49.0
above 59 ug/dl 100.0
-------
IV-51
Based on our analysis of NHANES II, we estimated that, of
those children over 25 ug/dl blood lead, about 20 percent are
between 30 and 40 ug/dl and 10 percent are over 40 ug/dl. Based
on those estimates and the percentages in Table IV-6, we assumed
that 5 percent of the children above 25 ug/dl would require chela-
tion therapy. In addition, we assumed that half of those children
chelated would require a second chelation due to a rebound in
their blood lead level, and that half of those children would
require a third chelation treatment. Thus, we assumed a total of
0.0875 chelations would be required for every child over 25 ug/dl
blood lead at screening. We assumed that chelation would require
five days in the hospital, several physicians' visits, laboratory
work, and a neuropsychological evaluation, for a total cost of
about $2,500 per chelation.
Multiplying each of these costs by its associated probability
and then summing them yields the estimated cost per child found
over 25 ug/dl at screening: 1.0(100) + 0.7(500) + 0.7(300) +
0.0875(2500) = $878.75, which we round to $900. This is lower
than the amount cited in a memo to the docket describing this
new methodology (August 16, 1984). The reason for this difference
is that, in the text of their article, Piomelli et al. recommend
chelation for all children over 50 ug/dl; our earlier memo assumed
that treatment in estimating costs. However, their data actually
indicate that some children in that range may not require chela-
tion, and, in this document, we have made the more conservative
assumption that not all children over 50 ug/dl will receive it.
-------
IV-5 2
Because we have not included welfare losses (such as work
time lost by parents), the adverse health effects of chelation
therapy itself (such as the removal of necessary minerals and
potential severe kidney damage), or such non-quantifiables as the
pain from the treatment, we believe our estimate of the benefits
is conservative. As mentioned previously, these medical costs
are a measure of avoidable damage for all the incremental cases
of lead toxicity, whether detected or not.
IV.D.2. Reduced Costs of Compensatory Education
As discussed earlier, several studies show that moderate-to-
high exposures to lead reduce cognitive ability, as measured by
IQ tests. The studies by de la Burde and Choate (1972, 1975) also
indicate that these cognitive effects, together with lead-induced
behavioral problems, translate into poorer performance in school;
they found that children in their high lead group were seven
times more likely than similar children with lower lead levels
to repeat a grade or be referred for psychological counseling.
Supplementary educational programs may compensate for some of
these effects, though certainly not all of them.
To estimate roughly the cost of such compensatory education,
we relied on data in a study prepared for the Department of
Education's Office of Special Education Program. Kakalik et al.
(1981) estimate that part-time special education for children who
remained in regular classrooms cost $3,064 extra per child per year
in 1978; adjusting for changes in the GNP price deflator yields
an estimate of $4,290 in 1983 dollars. This figure is quite
-------
IV-5 3
close to Provenzano's (1980) estimate of the special education
costs for non-retarded, lead-exposed children.
Based on de la Burde and Choate's finding that cognitive
effects persist for at least three years, we assumed that each
child needing compensatory education would require it for three
years. De la Burde and Choate's high-lead group consisted mostly
of children over 40 ug/dl blood lead, who make up about 10 percent
of all children over 25 ug/dl. A few, however, were in the range
of 30-40 ug/dl, and other studies (as discussed above in Section
IV.B) have found cognitive effects at levels well below 40 ug/dl.
Thus, we assumed that 20 percent of all children over 25 ug/dl
are affected severely enough that compensatory education would
be appropriate. Thus, our estimated average cost per child over
25 ug/dl is (0.20)(3)(4,290) = $2,574, which we round to $2,600.
IV.D.3. Summary of Estimated Benefits
Adding our estimates of compensatory education and medical
costs yields a combined benefit estimate of $3,500 per case avoid-
ed of a child's blood-lead level exceeding 25 ug/dl. Although for
convenience we have computed the average benefit per child over
25 ug/dl, it is important to note that most of the monetized bene-
fits are attributable to reducing lead in children who would be at
much higher levels (multiplied by the fraction of children over
25 ug/dl who are at those higher levels). It is also critical to
reiterate that our estimates are incomplete, omitting many
important categories, and thus are likely to be significant under-
estimates of the benefits of reducing lead in gasoline.
-------
IV-5 4
Table IV-7 presents year-by-year estimates of the monetized
children's health benefits of the alternative rules. They are
simply the estimated reductions in the number of children above
25 ug/dl (from Table III-6) multiplied by $3,500 per case. As
before, they assume that all misfueling is eliminated in each year.
The sensitivities of the results to alternative assumptions about
misfueling are explored in Chapter VIII.
-------
IV-55
TABLE IV-7. Year-by-Year Monetized Benefits of Reducing Children's
Exposure to Lead Under Alternative Rules,
Assuming No Misfueling (millions of 1983 dollars)
Category
Rule 1985 1986 1987 1988 1989 1990 1991 1992
Medical Care
Proposed 0 155 141 130 117 107 95 93
Alternative 65 139 134 130 117 107 95 93
Final 65 155 141 130 117 107 95 93
Compensatory Education
Proposed 0 447 408 374 338 309 276 268
Alternative 187 400 387 374 338 309 276 268
Final 187 447 408 374 338 309 276 268
Total
Proposed 0 602 550 504 455 417 371 361
Alternative 252 539 522 504 455 417 371 361
1987 1988 1989 1990 1991 1992
-------
CHAPTER V
HEALTH BENEFITS OF REDUCING LEAD:
ADULT ILLNESSES RELATED TO BLOOD PRESSURE
Concerns about the health effects of ambient exposure to
lead traditionally have focused on children. Although lead has a
variety of adverse effects on the health of adults, most of them
appear not to be of substantial concern except at very high blood-
lead levels. Recently, however, two new and extensive analyses
of the NHANES II data set have shown a strong and robust relation-
ship between blood lead levels and blood pressure. That finding
has important implications for the benefits of reducing lead in
gasoline, because high blood pressure, in turn, is linked to a
variety of cardiovascular diseases.
This chapter analyzes the health benefits for adults of
reducing lead in gasoline, but is limited in several ways.
First, we evaluated only illnesses related to blood pressure,
although lead has other adverse effects on adults. Second, the
analysis is restricted to males aged 40 to 59, because lead appears
to affect blood pressure only in men, not women, and because the
best data are available for that age range. Finally, most of
the estimates cover only whites, because the existing studies of
disease associated with blood pressure have had insufficiently
large samples of nonwhites. For these reasons, the estimates
contained in this chapter are likely to understate significantly
the adult health benefits of reducing lead in gasoline. The most
important omissions are older males and black males of all ages.
-------
V-2
The estimates presented in this chapter should be treated as
preliminary. They rely heavily on a recent paper by Pirkle et
al. (1985) that has been published in a peer-reviewed journal
(The American Journal of Epidemiology), but has not yet been
widely reviewed. A summary of that paper and the calculations
underlying the estimates in this chapter was placed in the docket
for this rulemaking (Schwartz, "Blood Lead and Blood Pressure",
September 7, 1984). Another recently published paper (Harlan et
al., 1985) also reports a statistically signifcant relationship
between blood pressure in the NHANES II data set. Until the
broader scientific community has had a chance to review these
papers and comment on their findings, EPA will not rely on blood-
pressure-related benefits for this lead-in-gasoline rulemaking.
These health effects will be considered in the Agency's ongoing
deliberations on a ban, however, and are addressed here for
information purposes.
V.A. The Relationship Between Blood Lead and Blood Pressure
This section analyzes the statistical relationship between
blood lead and blood pressure. The first part provides a brief
overview of earlier studies on the subject, while the second part
provides a detailed discussion of a recently completed statistical
analysis of the NHANES II data.
V.A.I. Earlier Studies
Lead has long been associated with effects on blood pressure
and the cardiovascular system, including a paper in the British
-------
V-3
Medical Journal by Lorimer in 1886 that found that higher blood-
lead levels increased the risk of hypertension. Most of the
studies have focused only on hypertension and relatively high
lead-exposure levels, and have not looked for a continuous effect
of lead on blood pressure. Investigators reporting such an
effect include Beevers et al. (1980), Morgan (1976), Richet et
al. (1966), and Dingwall-Fordyce and Lane (1963). Others have
failed to find effects of lead on hypertension that were signifi-
cant at the 95 percent confidence level, although most of them
did find a positive association. These include Ramirez-Cervantes
et al. (1978) and Fouts and Page (1942).
More recently, Batuman et al. (1983) found an association
between chelatable body-lead levels and hypertension in veterans,
and several recent general population studies and lower lead-
exposure studies (Beevers et al., 1976; Kromhout and Coulande,
1984) have found a significant association with blood lead.
Moreau et al. (1982) also found a significant relationship
(p < 0.001) between blood lead levels and a continuous measure of
blood pressure in 431 French policemen, after controlling for
age, body mass index, smoking, and drinking.
An even more recent British study (Pocock et al., in press)
found blood lead significantly related to blood pressure at the
99 percent confidence level, but the authors felt that the small
size of their correlation coefficient suggested no noticeable
effect. However, that conclusion appears to reflect a misunder-
standing of statistics. It is the regression coefficient that
indicates the size of an effect. A correlation coefficient
-------
V-4
confounds that measurement with the variances of the dependent
and independent variables. While their full data set was not
available to us, Pocock et al. presented their grouped data, and
we were able to perform a regression of blood pressure versus
the log of blood lead on their group averages, both before
and after adjustment for confounders. The regressions were
weighted by the inverse of the variance of each group, and con-
firmed their finding that blood lead was a significant predictor
of blood pressure in their data, both before and after adjusting
for covariates. Moreover, the regression coefficient indicated
that the size of the effect was significant, suggesting a change
of 3 mm Hg (millimeters of mercury, the standard measure of blood
pressure) as blood lead goes from 5 to 15 ug/dl.
Weeden (1975) found lead associated with the vascular
renal changes linked to essential hypertension, indicating a
possible causal pathway. Cooper and Gaffey (1974) analyzed
mortality data from 1,267 death certificates for 7,032 lead
workers employed between 1900 and 1969, and found a significant
excess of deaths from hypertension disease and renal disease. A
later analysis of similar data from 1971 to 1975 also found an
increase in cardiovascular and renal disease, but it was no
longer significant at the 95 percent confidence level (Cooper,
1981) .
Animal data also link lead to hypertension. Victery (1982)
found lead associated with a significant elevation of blood pres-
sure in rats with blood lead levels of 41 ug/dl. Importantly,
-------
V-5
this study confirms Beevers et al.'s finding of a sex differen-
tial, with male but not female rats becoming hypertensive. Webb
(1981) examined the vascular responsiveness of tail arteries in
rats exposed to blood lead levels in the 40 ug/dl range that had
suffered increases in systolic blood pressure, and found that the
arteries in exposed rats had increased contraction in response to
stimulation by neurotransmitters.
lannaccone et al. (1981) also reported increased blood
pressure (p < 0.001) in rats at blood lead levels of 38.4 ug/dl,
as well as significant increases in the blood pressure response
to noradrenalin. Perry and Erlanger (1979) found that low level
exposure of rats to lead produced increases of 15-20 mm Hg in
systolic blood pressure. Kopp (1980) repeated those findings
and found electrocardiogram changes, indicating an effect on
the heart itself. Subsequent tissue analysis of the heart showed
reduced levels of ATP in the heart muscle, indicating that heme
synthesis inhibition by lead was affecting energy availability
in the heart itself.
The direct cardiological effects of lead are also indicated
by electrocardiogram changes in lead-poisoned children, which
are reversed by chelation therapy (Freeman, 1965; Silver and
Rodrigues-Torres, 1968). Williams (1977, 1978, 1979) has
shown persistant increased susceptibility to norepinephrine-
induced arrhythmias in rats exposed to lead in the first
three weeks of life.
V.A.2. Analysis of NHANES II Data
In light of these indications of potential effects of lead
-------
V-6
on the cardiovascular system, the relationship between blood
lead levels and blood pressure has recently been explored (Harlan
et al., 1985; Pirkle et al., 1985) using the NHANES II data.
The NHANES II is an excellent data base for this analysis because
of the care given to accurate measurements, the great range of
information on possible confounding factors, and because it is a
representative sample of the U.S. population. As such it avoids
the problems of selection bias, healthy-worker effect, other
occupational exposures, and the choice of controls that confound
many occupational studies. Harlan et al. found blood lead related
to blood pressure for males aged 12 to 74 after controlling for
the traditional variables associated with blood pressure (age,
age-squared, body mass index, race) as well as alcohol consumption,
socio-economic factors, and all nutritional variables suspected
of affecting blood pressure. Moreover, this relationship held
in each year of the NHANES II sample, when analyzed separately,
and this relationship held for both blacks and whites.
Pirkle et al. found that blood lead levels were a statis-
tically significant predictor of blood pressure in adult males.
This relationship held not only when blood lead was evaluated
in a regression with all known factors that have previously been
established as correlated with blood pressure, but also when
that relationship subsequently was tested against 87 additional
variables representing linear and nonlinear functions of every
dietary and serologic variable in the NHANES II survey.
Although finaal judgments about the casual relationship
between blood lead and blood pressure must await further review
-------
V-7
and study, EPA believes that the Pirkle et al. study provides
a reasonable basis for estimating the potential blood-pressure-
related benefits of reducing lead in gasoline. Our analysis
builds on the Pirkle et al. study, and on additional work by
one of its authors (J. Schwartz).
V.A.2.a. Blood Pressure Measurements
Three blood pressure measurements were taken during NHANES
II. A seated measurement was taken as soon as the examinee
entered. Later, a recumbent measurement was taken. A second
seated measurement was taken just before the end of the examina-
tion. It is standard medical practice to prefer the second
seated measurement, because nervousness on just entering a
medical examination center makes the first seated measurement
less stable. All of the results presented are for the second
seated measurement. However, almost all of the regressions and
robustness tests described were performed on all three measure-
ments, and on the average of the first and third seated measure-
ments; all of the conclusions concerning lead's significance
held for all eight regressions (four diastolic, four systolic).
V.A.2.b. Initial Analysis
After replicating the Harlan et al. results for all adult
males, the first goal was to determine if blood lead levels were
related significantly to blood pressure in white males, 40 to 59
years old. This subgroup was chosen because at lower ages both
blood pressure and blood lead vary with age. This collinearity
could artificially mask or enhance the correlation between blood
-------
V-8
lead and blood pressure. Between 40 and 59 years of age, however,
blood pressure is essentially independent of age. Choosing this
subgroup avoids any collinearity problems. We focused on whites
because data relating cardiovascular disease to blood pressure
are less extensive for nonwhites. The established correlates of
blood pressure are age, sex, race, and one of the indices of
relative height-to-weight. Body mass index (BMI = weight/height2)
was used in this analysis. By limiting attention to 40 to 59 year
old white males, there was no need to control for race, sex, or,
to a large degree, age. Although age was only occasionally signi-
ficant in the stepwise analysis, both age and age-squared were
forced into each multiple regression model to be certain any
effect of lead was independent of age.
The natural log of blood lead was more normally distributed,
was more statistically significant, and gave a higher R2 than
untransformed blood lead, blood-lead-squared, blood lead plus
blood-lead-squared, the square root of blood lead, or blood lead
to other fractional powers (0.15, 0.2, 0.3, 0.4). All of the
results reported here are for the natural log of blood lead, but
regressions using lead on the untransformed scale gave very
similar results.
The initial regressions analyzed systolic and diastolic
blood pressures for white males, 40 to 59 years old, with a model
consisting of age, age-squared, BMI, and blood lead. These
regressions were done to determine whether blood lead levels were
significantly associated with systolic and diastolic blood pres-
sures after controlling for age, sex, race, and BMI, which are
-------
V-9
well-documented correlates of blood pressure. Lead was statis-
tially significant (p < 0.01) for both systolic and diastolic
blood pressures in all the regressions (unweighted, weighted, and
weighted with design effects). The regressions also tested
whether this relationship held up when other potentially confound-
ing variables were considered.
V.A.3. Tests of Robustness
The regression models were expanded to incorporate additional
variables, with particular attention directed to the stability
and significance of the lead coefficient in the presence of nutri-
tional factors and blood biochemistries.
A large set of nutritional and biochemical variables from
NHANES II was included in the stepwise regressions. Additional
regression analyses considered potential problems of interaction
terms. To ensure the robustness of the relationships, further
analyses were done to address marginally insignificant variables
and nonnutrition variables. Although our analysis focused on
males aged 40 to 59, additional regressions were also performed
considering all males over age 20.
V.A.S.a. Nutritional and Biochemical Variables
To provide an unusually rigorous test of the independent
significance of blood lead, almost all of the nutritional and bio-
chemical variables in the NHANES II were included in stepwise
regressions. In addition, to account for possible curvilinear
relationships, squared and natural logarithmic transformations
-------
V-10
of almost all of these variables were also included. The vari-
ables are listed in Table V-l. The objective was not to evaluate
the possible association of nutritional or biochemical measure-
ments with blood pressure, but rather to conservatively estimate
the strength and independence of the relationship between blood
pressure and blood lead.
Including these additional 87 variables increases the proba-
bility of variables being found statistically significant due to
chance alone. This complicates the interpretation of nutritional
and biochemical factors, but not the interpretation of the lead
variable; it only makes it more difficult for lead to maintain
its significance.
The general procedure for variable selection was as follows.
First, weighted stepwise multiple linear regression was used to
determine which variables were significantly related (p < 0.05)
to blood pressure (using the Stepwise and MAXR options of the SAS
procedure, STEPWISE). The MAXR procedure was the principle one
used; it determines for any given model size (i.e., number of
variables) the variables that explain the greatest amount of the
variance (i.e., maximize R^). We chose the largest model with
all variables significantly related to blood pressure (p < 0.05).
The Stepwise option, which uses forward selection with backwards
elimination, chose very similar models, and also always chose
blood lead. From the 87 nutritional and biochemical variables,
the weighted stepwise regression selected five additional variables
for diastolic pressure and six additional variables for systolic
pressure using a 5 percent significance test. These were used
-------
V-ll
TABLE V-l. Variables Included in the Stepwise Regression Analyses
age *
age-squared *
body mass index
dietary sodium t
salt shaker sodium
dietary sodium X
salt shaker sodium
dietary potassium t
dietary sodium -
potassium ratio
dietary calcium t
dietary phosphorus t
dietary protein t
dietary fat t
dietary carbohydrate t
dietary cholesterol t
dietary saturated
fatty acids t
dietary oleic acid t
dietary linoleic acid t
dietary iron t
dietary vitamin A t
dietary vitamin C t
dietary thiamine t
dietary riboflavin t
dietary niacin t
serum cholesterol+
serum vitamin C t
serum iron t
serum transferrin saturation
serum zinc t
serum copper t
serum albumin t
hemoglobin t
red blood cell count
ethanol consumption / week t
cigarettes smoked / day
total dietary grams t
total dietary calories t
cigar or pipe smoking
* forced into each regression to remove any possible age
effects on blood pressure.
t the natural log and squared transformation of these
variables were also included in the stepwise regression,
-------
V-12
as the starting model for the SAS procedure SURREGR, which addi-
tionally incorporated the survey design effects. For both
systolic and diastolic blood pressures, one variable from the
weighted stepwise regression failed to maintain significance at
the 5 percent level after the design effects were incorporated.
The final regression results for systolic and diastolic pressures,
after accounting for the weighting and design effects, are given
in Table V-2.
The multiple logistic regressions (of the probability of
hypertension) were also performed using programs from SAS. The
procedure LOGIST was used for the stepwise unweighted multiple
logistic regression, and the procedure NLIN (nonlinear regression)
was used for the weighted logistic regression calculations. The
selection process again chose the largest significant model that
explained the greatest amount of the variance. Calculations of
threshold levels for effects were made using the procedure NLIN
on segmented regression models, which finds the threshold point
that minimizes the sum of the squares of the error terms. The
results of the logistic regression on hypertension are shown in
Table V-3. Note that the logistic regressions included blacks
as well as whites, because these regressions were used only to
predict the effect of lead on the probability of having hyperten-
sion and were not used to estimate the number of cardiovascular
diseases and deaths. As noted earlier, blacks were not included
in the linear regressions because the best available coefficient
for predicting cardiovascular risks included insufficient numbers
of blacks.
-------
V-13
TABLE V-2. Regression of Diastolic and Systolic Blood Pressures
in White Males Aged 40 to 59
Variable
Diastolic
Age
Age2
Body Mass Index
Log (blood lead)
Dietary Potassium
Hemoglobin
Albumin
Log (dietary
vitamin C)
Systolic
Age
Age2
Body Mass Index
Log (blood lead)
Albumin
Log (dietary
Vitamin C)
Log (dietary
riboflavin)
Log (dietary
oleic acid)
Log (serum
vitamin C)
Coefficient
0.2768
-0.0014
1.131
3.954
-0.0018
1.548
3.587
1.838
1.311
-0.0068
1.736
8.436
7.088
2.411
-5.509
3.992
-3.472
t-Statistic
0.17
0.10
8.55
2.85
4.92
3.90
2.50
4.65
0.57
0.30
9.42
3.24
2.50
3.84
3.07
2.49
2.47
Probability
0.8636
0.9321
0.0001
0.0080
0.0001
0.0005
0.0179
0.0001
0.5720
0.7706
0.0001
0.0028
0.0178
0.0005
0.0044
0.0183
0.0184
-------
V-14
TABLE V-3. Weighted Logistic Regression on Probability of Diastolic
Blood Pressure Greater Than or Equal to 90 nun Hg in Men
Aged 40 to 59
Variable
Constant
Log (Blood Lead)
Albumin
Body Mass Index
Hemoglobin
Log(Vitamin C)
Dietary Potassium
Total Carbohydrates
Coefficient
-16.41
0.693
0.0873
1.700
0.0329
0.3585
-0.00058
0.00246
t-statistic
10.13
3.96
3.70
9.34
5.25
5.98
7.47
3.09
p-Value
0.0000
0.0000
0.0001
0.0000
0.0000
0.0000
0.0000
0.0010
-------
V-15
After including the nutritional variables, the blood analytes,
and their curvilinear transformations, lead remained significantly
associated (p < 0.01) with both systolic and diastolic blood
pressures. The magnitude of this relationship, adjusted for the
other significant variables, is shown graphically in Figures
V-l and V-2. Furthermore, segmented regression analyses indicated
there was no threshold blood lead level in the data.
These segmented "hockey stick" regressions fit two regression
lines to the data. One, below the putative blood lead threshold
T, depends on all the variables except lead. The other, for blood
lead levels above T, includes lead. An iterative technique is
used to find the value of T that minimizes the sum of the squares
of the error terms over the full range of both regression lines.
In this case, the error in the regression was minimized at a
threshold of zero; that is, lead was significantly related to
blood pressure at all levels down to zero.
V.A.S.b. Interaction Terms
In multiple regression analysis, another consideration is
the possibility of significant interaction terms. To evaluate
this possibility, an additional weighted stepwise regression
analysis was done for systolic and diastolic blood pressures.
The variables consisted of the linear interaction terms between
the final variables in the model (shown in Table V-2) and the
linear form of all the other variables originally selected for
the initial stepwise regression, including their log and square
transforms (Table V-l). This meant running a stepwise regression
with 162 interaction terms added to the final regression models
-------
V-16
140
x
E
135
CO
CO
111
(C
Q.
o
o
o
o
CO
(0
o
IU
CO
o
<
130
125
120
FIGURE V-l
Ad j lasted Systolic Blood Pressure versus Blood Lead
1 6 10 14 18 22 26 30 34 38
ADJUSTED BLOOD LEAD LEVELS (MICROGRAMS/DECILITER)
-------
o>
X
iu
oc
D
CO
CO
UJ
oc
o.
Q
O
o
CD
O
Zj
O
&
o
ui
h-
CO
92
I 90
88
86
84
82
80
< 78
V-17
FIGURE V-2
Adjusted Diastolic Blood Pressure versus Blood Lead
6
10
14
18
22 26 30 34 38
ADJUSTED BLOOD LEAD LEVELS (MICROGRAMS/DECILITER)
-------
V-18
for systolic and diastolic pressures. Using such a large set
of variables gave a high probability that some variables would
enter at the 5 percent level by chance. However, the purpose
was not to determine if those variables were independently
significant, but, rather, to further test the significance and
independence of the relationship between blood pressure and
blood lead. As expected, several interaction variables entered
the systolic and diastolic regressions, but in each regression
the lead coefficient varied less than 10 percent and remained
significant (p < 0.015).
V.A.3.C. Marginally Insignificant Variables
Three other analyses were done to ensure that this
relationship was robust. First, the original weighted stepwise
regression was extended to include variables significant through
the 15 percent level to see if marginally insignificant variables
influenced the significance of lead. For both systolic and dia-
stolic pressures, lead remained significant and there was little
change in the magnitude of the coefficient.
Second, for diastolic blood pressure, all the variables were
included that were significant between the p = 0.05 and the 0.15
levels, and every possible combination of those variables was
considered. All 255 combinations were added to the variables that
were statistically significant, and a regression was performed
on each one. The coefficient of the log of blood lead varied by
only plus or minus 10 percent from the value we obtained when we
included only significant variables, and the highest p-value for
lead was still less than 0.01.
-------
V-19
The last analysis was the most demanding test of the indepen-
dence of the relationship between blood pressure and blood lead.
Models for diastolic and systolic blood pressures were fit by
weighted stepwise regression to the original model variables
(Table V-l), excluding lead. This gave all of the other variables
and their curvilinear transformations the maximum opportunity to
explain variation that could also be explained by lead. After
obtaining this new final model without lead, a single regression
was run adding the lead variable to the variables of this new
final model. For both systolic and diastolic pressures, lead was
still statistically significant (p < 0.016) and the magnitude of
the lead coefficient changed less than 10 percent from those ob-
tained in the original analysis. The results of all these analyses
indicated that the strength and independence of the relationship
between blood pressure and blood lead were remarkably stable.
Because some people have found small amounts of ethanol
associated with reduced blood pressure, ethanol was also modeled
as a quadratic function of consumption, and with two dummy
variables for light and heavy drinking. The stepwise regression
was repeated, with no change.
V.A.3.d. Nonnutrition Variables
Pirkle et al. then considered nonnutrition variables that
might be associated with blood pressure. In additional runs
completed since then, we have added several other variables.
The complete set is shown in Table V-4. Socio-economic and demo-
graphic factors as well as additional medical history variables
were included.
-------
V-20
TABLE V-4. Nonnutrition Variables Tested in the Stepwise Regression
Demographic Variables Other Personal-History Variables
Family Income Tricep Skinfold
Poverty Index Subscapular Skinfold
Region of the Country Recreational Exercise
Season of the Year Work-Related Exercise
Degree of Urbanization Recent Weight Loss
Residence Inside Central City Family History of Hypertension
Educational Level Kidney Disease
Serum Creatinine
Hypertension Variables
Hypertensive Medication
Low Salt Diet
-------
V-21
Hypertension medication and low salt diet were tested —
not for inclusion in a final model, as they are essentially
indicators of high blood pressure, but rather to see if the
response to lead differed in those groups. The coefficient of
lead did not change appreciably, and lead interaction terms with
the two variables were insignificant.
The other variables in Table V-4 were tested in two ways.
First, the stepwise regression procedure was repeated with them
using all nutritional and serum measurements that were significant
at the p = 0.15 level. The nutritional factors were limited to
those significant at the 0.15 level to give the nonnutritional
factors a greater chance to enter the model. Again, lead was
selected (p < 0.005) and its coefficient changed by less than 10
percent from the original model that included only age, age2,
and body mass index.
The variables in Table V-4 were then added to all of those
on Table V-l (including their nonlinear transforms) and the step-
wise process was repeated -- with the same results. Finally,
the stepwise procedure was rerun using all of the variables in
Tables V-l and V-4 except lead; lead was then inserted into the
model resulting from this procedure. It was still significant
(p < 0.006), with less than a 10 percent change in its coefficient.
Because the presence of two terms to describe the curvilinear
dependence of blood pressure on age might reduce the chances of
variable correlated with age achieving significance, age was modeled
as a single curvilinear function (sine of age), and the stepwise
-------
V-22
regression repeated; the results were the same. In addition,
smoking and drinking were forced into the regression, and lead
was still significant (p < 0.01), with only a 3 percent change
in its coefficient.
Our previous studies have shown that about half of the lead
in people in the NHANES II sample came from gasoline. Tetraethyl
lead has very little cadmium in it, so confounding with cadmium
(which is also suspected of affecting blood pressure) is unlikely.
However, we repeated the regression excluding occupationally
exposed workers, who may also have cadmium exposure. Lead
remained significant (p < 0.01), and its coefficient increased
somewhat. We also regressed gasoline lead directly on blood
pressure, and it was significant.
Although all of these analyses make it clear that
collinearity is not a problem in these regressions, variance
inflation factors were computed; no significant variable had a
variance inflation factor above 1.4. (Variance inflation factors
below 4 are considered acceptable in multiple regression analyses.)
To ensure that the significance of lead in the regression
was not due to the presence of a few influential observations,
influence diagnostic procedures were run. Studentized residuals
were plotted for all the observations, and the largest residuals
were clustered near the middle of the data, where their influence
is slight. Cook's D statistics also were computed for each obser-
vation. The highest Cook's D was 0.029, and the second highest
was 0.023, both of which are very small. Moreover, of the 10
observations with the largest Cook's D statistics, six had posi-
tive residuals and four had negative residuals, indicating that
-------
V-23
the most influential observations split almost evenly on which
way they would influence the lead regression coefficient.
V.A.S.e. Other Age Groups
The 40 to 59 year old age group represents about one-third
of adult males, and is the only one where the confounding of age
and blood lead can be eliminated unambiguously. Additional
regressions were performed, however, to confirm the Harlan et al.
finding of an effect in all adult males. Tables V-l and V-4
contain several variables that Harlan et al. did not consider in
their analysis. Therefore, the stepwise regression analysis was
repeated using all of the variables in both tables, and their
square and natural log transforms as indicated. All males over
the age of 20 were considered. Lead was selected by the regres-
sion, with a p-value less than 0.01. To check whether the rela-
tionship might be substantially different for different age
groups, dummy variables for each 10-year age group (between 20
and 70 years), and interaction terms between lead and those dummy
variables, were inserted in the stepwise regression. Such inter-
action terms check for differences in the lead/blood pressure
relationship without having to subdivide the sample. None of the
interaction terms was significant at even the p = 0.15 level.
V.A.4. Summary of Blood Lead - Blood Pressure Results
The final models for blood pressure, including all
statistically significant variables, are shown in Table V-5.
The final logistic model for the probability of hypertension is
shown in Table V-6.
-------
V-24
TABLE V-5. Regression of Diastolic and Systolic Blood Pressures
in White Males Aged 40 to 59
Variable
Diastolic
Age
Age- squared
Body Mass Index
Blood leadt
Potassium
Hemoglobin
Albumin
Dietary Vitamin Ct
Family history of
hypertension
Recreational
exercise
Coefficient
-0.210
0.003
1.082
4.609
-0.002
0.151
0.354
1.886
2.085
-1.851
F-Statistic
0.02
0.04
67.88
12.19
25.30
16.81
7.42
23.67
4.37
9.48
Probability
0.8960
0.8373
0.0000
0.0014
0.0000
0.0003
0.0104
0.0000
0.0446
0.0042
Systolic
Age
Age-squared
Body Mass Index
Blood leadt
Albumin
Dietary Vitamin Ct
Dietary Riboflavint
Dietary Oleic Acidt
Serum Vitamin Ct
Family history of
hypertension
1.142
-0.005
1.710
8.510
0.695
2.458
-5.101
3.650
3.365
3.683
0.25
0.05
85.90
10.54
6.09
13.78
8.14
5.34
5.81
4.59
0.6226
0.8208
0.0000
0.0027
0.0192
0.0008
0.0075
0.0275
0.0218
0.0399
t log transform
-------
V-25
TABLE V-6. Logistic Regression on Probablity of Blood
Pressure Greater Than or Equal to 90 ntn Hg in
Men Aged 40 to 59
Variable
Constant
Log (Blood Lead)
Albumin
Body Mass Index
Hemoglobin
Log(Vitamin C)
Dietary Potassium
Total Carbohydrates
Recreational Exercise
Coefficient
-15.40
0.793
0.650
0.1571
0.0265
0.3593
-0.00053
0.00286
0.3864
t-Statistic
7.0
3.20
2.06
6.57
3.19
4.22
5.33
2.86
0.128
p-Value
0.0000
0.0014
0.0399
0.0000
0.0015
0.0000
0.0000
0.0080
0.0026
-------
V-26
It is noteworthy that the logarithmic form of the dose-
response relationship suggests a large initial effect, leveling
off at higher blood-lead levels. This may explain why only about
60 percent of the occupational studies (i.e., high lead-exposure
studies) have found an effect that was significant at the 95
percent confidence level, while almost all of the studies of
lower lead levels have found the relationship to be significant.
The other low-exposure studies, the animal data, and the
robustness of these results suggest that the relationship is
causal. Moreover, specific analyses to determine whether there
is a lower threshold below which lead has no effect on blood pres-
sure showed that the data were fit best with a threshold of zero.
V.B. Benefits of Reduced Cardiovascular Disease
Reducing lead in gasoline will reduce blood lead levels,
which in turn will reduce blood pressure and the number of indi-
viduals with hypertension. The reduction in hypertension will
have some direct benefits from reduced medical treatment expendi-
tures. More important, however, will be the indirect benefits
in the form of reduced cardiovascular disease associated with
elevated levels of blood pressure.
This section describes the methods used to estimate the
benefits associated with lowering blood pressure. The first part
deals with estimating the reductions in morbidity and mortality,
while the second discusses the methods used to value those
benefits.
-------
V-27
V.B.I. Reductions in Hypertension and Related Morbidity
and Mortality
Estimating the reduction in hypertension and cardiovascular
disease requires several steps. The first is to estimate the
impact of the reduction in gas lead on levels of lead in adults'
blood. For that step, we used the regression analyses of the
NHANES II data reported in Chapter III. Those regression
coefficients were applied to the NHANES II data base to simulate
the effects of gasoline lead reductions on blood lead levels.
In each case, the blood lead levels in the NHANES II data were
first adjusted to reflect reductions that have occurred since
the time of the survey. The subsequent steps vary with the
condition involved, and are described below.
V.B.1.a. Hypertension
Estimating the change in the number of cases of hypertension
was straightforward; the logistic regression coefficients from
Table V-6 were applied to the individual NHANES II data to predict
the numbers of hypertensives at alternative levels of gasoline
lead. The change due to this regulation was calculated by sub-
tracting the number at the new lead level from the number at the
original lead level (1.10 gplg). Table V-7 reports the year-by-
year estimates for three cases: the Final Rule (0.50 gplg on
7/1/85 and 0.10 gplg on 1/1/86), the Proposed Rule (0.10 gplg on
1/1/86), and the Alternative discussed in the Notice of Proposed
Rulemaking (0.50 gplg on 7/1/85, 0.30 on 1/1/86, 0.20 on 1/1/87,
and 0.10 on 1/1/88). In all three cases, Table V-7 assumes that
the rules eliminate all misfueling; alternative assumptions
-------
V-28
TABLE V-7. Reductions in Cases of Hypertension in Males Aged 40
to 59, Assuming No Misfueling (thousands of cases)
Rule 1985 1986 1987 1988 1989 1990 1991 1992
Proposed 0 1,804 1,727 1,649 1,562 1,489 1,396 1,399
Alternative 639 1,527 1,600 1,649 1,562 1,489 1,396 1,399
Final 639 1,804 1,727 1,649 1,562 1,489 1,396 1,399
-------
V-29
about misfueling are examined in Chapter VIII. These estimates
cover only males aged 40 to 59, but include nonwhites as well as
whites.
V.B.l.b. Myocardial Infarctions, Strokes, and Deaths
Estimating the impact of reduced blood pressure on morbidity
and mortality required several additional steps. Using the NHANES
II data and the regression coefficients in Table V-5, we simulated
the changes in individual blood pressure levels due to reductions
in gasoline lead. Coefficients from two large studies of cardio-
vascular disease were then used to estimate changes in the numbers
of first-time myocardial infarctions, first-time strokes, and
deaths from all causes.
The relationships between blood pressure and cardiovascular
diseases are well established by several large, long-term epidemic-
logical studies. The classic study, which was important in
establishing cholesterol as a major factor in the risk of heart
disease, was the Framingham study (McGee et al., 1976). Extensive
analyses of these data have yielded estimates of cardiovascular
risks associated with several variables, including blood pressure.
Figure V-3 shows the age-adjusted rates of death and heart attacks
as functions of blood pressure from that study.
In the 1970s, the National Institutes of Health funded the
Pooling Project (The Pooling Project Research Group, 1978), which
combined the Framingham data with data from five other long-term
studies to improve the accuracy of the risk coefficients for heart
attacks. The Pooling Project tested the Framingham coefficients
-------
V-30
FIGURE V-3
Adjlasted Rates of Death and Heart Attacks versus Blood Pressure:
Framingham Data
Annual incidence of death by diastolic blood pressure
Males 45-74 (age adjusted rate)
250
I
LI
s,
o
§
u
&
s
§
•O
0
|
225
200
175
150
125
100
<70 70- 75- 80- 85- 90- 95- 100- 105- 110+
74 79 84 89 94 99 104 109
n
&
o
o
o
14
s,
o
.5
Annual incidence of nyocardial infarction by
diastolic blood pressure
Males 45-74 (age adjusted rate)
100 -
90
80
70 -
60
50
40
<70 70- 75- 80- 85- 90- 95- 100- 105- 110+
74 79 84 89 94 99 104 109
-------
V-31
against the other study results and found that their predictive
power was good. It then analyzed the first occurrence of myocard-
ial infarctions (serious heart attacks) in white men who entered
the studies at ages 40 to 59 and who were followed for at least 10
years. Our estimates of the numbers of first-time myocardial
infarctions under alternative standards employ the Pooling
Project's coefficients.
In addition to estimating the risk of heart attacks, the
Framingham study estimated regression equations for the risks of
stroke and death as functions of blood pressure and other vari-
ables. Because the Pooling Project did not include those end-
points, we used the Framingham study coefficients. As with heart
attacks, the estimates for strokes cover only first-time events;
thus, our estimates for strokes and myocardial infarctions are
biased downwards because they exclude second and subsequent heart
attacks and strokes associated with elevated blood pressure. The
regression equation for deaths covers all causes of death; it
includes deaths not just from myocardial infarctions and strokes,
but also from other causes associated with blood pressure (e.g.,
heart diseases other than myocardial infarctions).
Levy et al. (1984) recently tested the Framingham study
regression coefficients to see how well they explained the observed
decrease in cardiovascular mortality in the United States from
1970 to 1980. They found that the coefficients, when coupled
with changes in blood pressure and other cardiovascular risk
factors over that same period, were able to explain about 80
percent of the drop in cardiovascular mortality.
-------
V-32
The Hypertension Detection and Follow-up Program (NEJM, 1983)
found that intervention leading to about a 5mm Hg change in
diastolic blood pressure produced a 20 percent reduction in
overall mortality. The Australian National Trial on mild hyper-
tension also found reductions in morbidity and mortality resulted
from lowered blood pressure (Lancet, 1980). The Multiple Risk
Factor Intervention Trial found that drug therapy to lower blood
pressure reduced cardiovascular disease in persons with normal
resting electrocardiograms (ECGs), but increased it in persons
with abnormal resting ECGs (JAMA, 1982). This suggests an
adverse affect of the drugs used.
To produce estimates for all 40 to 59 year old white males,
the individual risk of each person sampled in the NHANES II was
summed and then averaged. Since the sampled individuals repre-
sent the U.S. population for their specific age-race-sex category,
their average risk represents the average risk for all 40 to 59
year old white men. Because blood lead levels have dropped since
the NHANES II period, we corrected for that change and then
evaluated the effects of the new lead-in-gasoline limits. Again,
only white men were examined because there were too few blacks
in the Framingham study, and their risk might be different from
whites.
The fact that gasoline lead levels would slowly decline even
without new EPA actions created a slight complication. Because
gasoline lead levels fall over time in both our base case and the
low-lead case, the difference in blood lead levels resulting from
the rule will change over time. Therefore, we recalculated the
-------
V-33
risk estimates for each year, assessing the annual change in blood
lead levels due to reductions in gas lead.
The three cardiovascular-risk regression equations all predict
risk over the next 10 years, given current blood pressure, age, and
other characteristics. Presumably, the risk in years 2-10 was
affected by blood pressure in those years, as well as by initial
blood pressure. Because blood pressure levels over time in the
same individual are positively correlated, it is likely that
the regression coefficient in part picked up the effect of future
blood pressure levels. Lacking any data with which to estimate
the pure effect of a one-year change in blood pressure, we divided
the coefficient for 10-year risk by 10. The adjusted coefficient
was then used with the year-by-year predicted changes in blood
pressure to estimate risk reductions. This procedure almost
certainly overcompensates, lending a downward bias to the results,
because current blood pressure is not perfectly correlated with
future blood pressure.
We adjusted the population at risk for the increases in the
U.S. population of white males aged 40 to 59. The regression
from the Framingham study predicting deaths for men aged 40 to
54 was extended to 40 to 59 for data comparability and uniformity.
Because the death rate actually increases with age, this also will
bias the results downward.
Table V-8 reports the resulting year-by-year estimates of
reduced myocardial infarctions, strokes, and deaths for the three
phasedown schedules, as was done in Table V-7 for hypertension.
As before, these estimates assume that misfueling is completely
-------
V-34
TABLE V-8. Reductions in Numbers of Cases of Cardiovascular
Disease and Deaths in White Males Aged 40 to 59,
Assuming No Misfueling
Condition
Rule
1985 1986 1987 1988 1989 1990 1991 1992
Myocardial
infarctions
Proposed
Alternative
Final
Strokes
Proposed
Alternative
Final
Deaths
Proposed
Alternative
Final
0 5,350 5,156 4,956 4,726 4,531 4,274 4,289
1,829 4,467 4,750 4,956 4,726 4,531 4,274 4,289
1,829 5,350 5,156 4,956 4,726 4,531 4,274 4,289
0 1,115 1,074 1,032 984 943 889 892
382 932 990 1,032 984 943 889 892
382 1,115 1,074 1,032 984 943 889 892
0 5,160 4,971 4,778 4,556 4,367 4,119 4,132
1,766 4,310 4,581 4,778 4,556 4,367 4,119 4,132
1,766 5,160 4,971 4,778 4,556 4,367 4,119 4,132
-------
V-35
eliminated. They indicate that the rule being promulgated will
have a large impact on the incidence of cardiovascular disease.
We estimate that the reduction of lead in gasoline in 1986 alone
will result in 5,350 fewer myocardial infarctions; 1,115 fewer
strokes; and 5,160 fewer deaths from all causes among white
males aged 40 to 59. Extending the analysis to men of other
ages and to nonwhites would substantially increase these estimates.
V.B.2. Monetized Benefit Estimates
Valuing reductions in morbidity and mortality is a difficult
and, to say the least, controversial task. For morbidity, we have
restricted our estimates to avoided medical costs and foregone
earnings associated with diseases. These estimates clearly
are too low, for they fail to account for other important losses
associated with disease, including pain and suffering (e.g.,
the paralysis that often follows a stroke). For valuing the
reduction in mortality risk, we have chosen a fairly conservative
estimate ($1 million) from the large range obtained from studies
of occupational risk premiums.
V.B.2.a. Hypertension
Whether or not it results in coronary or cerebrovascular
disease, high blood pressure is a significant chronic illness.
It also generates economic costs, in the form of drugs, physicians'
visits, hospitalization, and work loss. We used data from the
NHANES II and from the National Institutes of Health to estimate
the value of avoiding a case of high blood pressure.
-------
V-36
The NHANES II ascertained how many times per year a person
saw a physician because of high blood pressure. The weighted
average, for males 40 to 59 years old with diastolic blood pres-
sure over 90mm, was 3.27 visits per year. We assumed a cost of
$35 per visit, for an annual total of $114.
The same population was forced to remain in bed an average
of 0.41 days per year because of high blood pressure. At the
average daily wage ($80), that translates to $33 per year.
NHANES II also found that 29 percent were on medication for
hypertension; assuming a drug cost of $200 per year for those
on medication yields an annual cost of $58.
The National Hospital Discharge Survey (1979) found that,
excluding those with heart disease or cerebrovascular disease,
people with high blood pressure used 3.5 million of the occupied
hospital bed-days that year; dividing by the 60 million people
the NHANES II identified as having high blood pressure gives a
rate of 0.058 hospital bed-days per person per year. We have
assumed that these results apply to the 40 to 59 year old age
group, as well. Using a daily hospital cost of $400 yields an
annual cost per hypertensive of $23.
Summing these estimates yields a total of $228 per
hypertensive per year. It should be noted that only 29 percent
of the people with blood pressure above 90mm in the NHANES II
were on medication, in part because some of them had not pre-
viously been detected as having high blood pressure. Therefore,
the average cost for a detected case will be higher. For example,
Weinstein and Stason (1977) used an average cost of $200 in 1975
-------
V-37
dollars, or about $450 in 1983 dollars, for treatment of patients
undergoing medical care for hypertension. Nevertheless, we have
conservatively used $220 as the value of avoiding one case of
high blood pressure for one year.
V.B.2.b. Myocardial Infarctions
Our estimate of the benefits of reducing the incidence of
myocardial infarctions relies heavily on Hartunian et al. (1981),
who estimated the medical expenses and lost wages associated with
a variety of diseases. Under the category of myocardial infarc-
tions (MI), Hartunian et al. examined three types of cases:
sudden death, fatal MI, and nonfatal MI. ("Sudden death" was
classified as a myocardial infarction in the Pooling Project
regression coefficients we used.)
For each category and each age group, Hartunian et al.
obtained data on the type of medical services needed (e.g., ambu-
lance and coronary intensive care unit), the fraction of cases
using each service, and the costs in 1975 dollars. They also
determined the annualized recurrence and follow-up costs, by age,
for each condition. These were then discounted (using a 6 percent
real discount rate) to the time of initial occurrence to estimate
the cost, in current dollars, of each new case. The resulting
estimates were $96 for sudden death and $7,075 for both fatal and
nonfatal Mis.
We have adjusted these 1975 estimates in three ways to
reflect 1983 conditions. First, we inflated them to 1983 dollars.
Because most of the costs were hospital-related, with the rest
-------
V-38
principally being physicians' fees, we inflated the Hartunian et
al. cost estimates by a weighted average of 80 percent of the
change in the Consumer Price Index (CPI) for hospital rooms and
20 percent of the change in the CPI for physicians' charges.
Approximately 90 percent of the Hartunian et al. MI costs were
hospital-related, not physicians' fees, and hospital costs rose
faster than physicians' fees, so this approach is conservative.
The second adjustment in the 1975 estimates involves changing
cost indices. Because cost indices only account for increased
costs of the same procedure, in this case principally the initial
hospitalization for a heart attack, they do not reflect the cost
of new or different procedures. Since 1975, the fraction of people
suffering coronary heart disease who subsequently undergo coronary
bypass operations has increased substantially. The number of by-
pass operations tripled in seven years, from 57,000 in 1975 to
170,000 in 1982, while the number of cases of coronary heart
disease has remained relatively constant (National Centers for
Health Statistics, Hospital Discharge Survey, and unpublished
data). Based on the Hartunian et al. data, 7.1 percent of MI
cases in 1975 had subsequent bypass operations. Assuming that
they shared proportionately in the tripling of the bypass-opera-
tion rate, we estimated that an additional 14 percent of Mis now
result in a bypass operation. Hartunian et al. estimated the
cost of bypass operations at $6,700 in 1975 dollars, or $16,800
in 1983 dollars. Adding 14 percent of this cost to the other
direct costs yields an estimate of the total direct costs in
1983 dollars of $20,100 for an MI and $240 for sudden death.
-------
V-39
Our third adjustment involved discount rates. Hartunian
et al. used a 6 percent real discount rate to present value the
future year costs, whereas this analysis employs a 10 percent
discount rate. Fortunately, Hartunian et al. performed sensi-
tivity calculations for other discount rates, including 10 per-
cent. Making all of these adjustments, the costs per case are
$18,100 for an MI and $216 for sudden death.
Hartunian et al. also obtained data indicating the proba-
bility distribution of cases among the different categories. Of
the total number of cases in these three categories, about 22.5
percent were sudden deaths and the remaining 77.5 percent were
fatal or nonfatal Mis. Applying those percentages to the
medical-cost estimates derived above yields a weighted average
of $14,076 per myocardial infarction.
Hartunian et al. calculated the present value of fore-
gone earnings based on reduced labor force participation using
data on each type of heart disease, broken down by sex and 10-year
age categories. We have used those results, with several modi-
fications. First, we excluded foregone earnings for fatal heart
attacks, because we valued the reduction in mortality risks
separately (see Section V.B.2.d., below). Second, we adjusted
for the increase in average non-farm compensation from 1975 to
1983, using information from Data Resources, Incorporated.
Finally, we again used a discount rate of 10 percent, rather than
the 6 percent used by Hartunian et al. in their base case
analysis.
-------
V-40
The resulting estimates of foregone earnings are $90,000 for
heart attack victims under 45; $47,000 for those between 45 and
54; and $22,000 for those over 55. Based on data from the
Pooling Project and NHANES II, 16.1 percent of nonfatal heart
attacks in men between 40 and 59 occur in those under 45, 50.9
percent occur in those between 45 and 54, and 33 percent in those
55 and older. Using those percentages yields a weighted average
for lost earnings of $45,670 per attack. Combining that earnings
estimate with the earlier one for medical costs yields a total
benefit per myocardial infarction avoided of about $60,000.
V.B.2.C. Strokes
Our estimates of the benefits of avoiding strokes also rely
on Hartunian et al., with similar adjustments. (Unlike myocardial
infarctions, we have not adjusted their medical cost estimates for
strokes to reflect any changes in medical treatment since 1975.)
Table V-9 presents the estimates for three types of stroke —
hemorrhagic, infarctive, and transient ischemic attacks (TIA) —
by age. The averages are based on the distribution of types of
strokes and incidence of strokes by age. The overall average is
$44,000 per stroke avoided.
We have been unable to estimate a value for avoiding the
loss in quality of life that occurs in stroke victims. This
is a significant omission. For example, of the people in the
NHANES II who reported having had a stroke in the past, 45 percent
suffered paralysis in the face and 13 percent still had at least
partial facial paralysis, 54 percent suffered paralysis in at
-------
V-41
TABLE V-9. Benefits of Reducing Strokes (dollars per case)
Type of Stroke
Age
Hemorrhagic
35-44
45-54
55-64
Inf arctive
35-44
45-54
55-64
Transient ischemic attacks
35-44
45-54
55-64
Medical
Expenses
12,600
13,300
17,200
17,600
18,100
23,600
3,184
3,184
3,184
Foregone
Earnings
41,000
26,000
11,000
71,000
43,000
14,000
1,114
3,076
8,280
Total
53,600
39,300
28,200
88,600
61,100
37,600
4,298
6,260
11,464
Weighted average
44,000
-------
V-42
least one arm and 21 percent remained paralyzed, 59 percent had
numbness in arms or legs and 28 percent had remaining numbness,
30 percent had vision impairment and 13 percent remained visually
impaired, and 50 percent had speech impairment with 22 percent
continuing to suffer from speech impairment. While we have no
estimates of people's willingness to pay to avoid the risk of
these profound injuries, common sense suggests that it is high.
V.B.2.d. Mortality
Valuing reductions in mortality is highly controversial.
Over the past decade or so, a substantial literature has developed
on the subject. Economists are in general agreement that the
best conceptual approach to use is the willingness-to-pay (WTP)
of the individuals involved. The appropriate value is not
the amount that an individual would pay to avoid certain death,
but rather the total sum that a large group of individuals would
pay to reduce small risks that sum to one; for example, the
amount that 10,000 people would pay to reduce a risk to each of
them of one in ten thousand.
Several studies have estimated WTP based on implicit tradeoffs
between risk and dollars revealed in market transactions. Most
of these studies (e.g., Thaler and Rosen, 1976; Smith, 1974 and
1976; Viscusi, 1978) have studied labor markets, based on the
premise that, all else being equal, workers must receive higher
wages to accept a higher risk of being injured or killed on the
job. Such studies typically regress wages on risk and a variety
of other explanatory variables (e.g., levels of education required,
-------
V-43
worker experience, whether or not the industry is unionized,
location, and non-risk working conditions). In such regressions,
risk might be measured as the number of fatalities per 1,000
workers per year. The coefficient for that variable is then
interpreted as the amount of extra wages needed to compensate
for a 0.001 risk of death. Dividing the coefficient by the unit
of risk yields the estimate of WTP to avoid a statistical death.
For example, if the coefficient is $500, the estimated WTP is
$500,000 (= $500/0.001).
A few studies have estimated WTP in nonoccupational settings,
Blomquist (1977), for example, estimated the implicit cost-risk
tradeoffs that individuals make in deciding whether or not to
take the time to put on seat belts.
None of these studies yields definitive answers. All suffer
from data limitations (e.g., incomplete information on possible
confounding variables and on the extent to which individuals
perceive the risks they face). Not surprisingly, given these
problems, the studies also yield a wide range of estimates. A
recent survey of the literature prepared for EPA found a range
of $400,000 to $7 million per statistical life saved (Violette
and Chestnut, 1983). Based on that survey, EPA1s RIA guidelines
do not attempt to set any specific value, but rather recommend
that range. To simplify the presentation of the results, this
RIA uses a single value from the lower end of that range, $1
million per statistical life saved. Although we do not present
any formal sensitivity analyses on this value, the results in
Chapter VIII show that the net benefits are so large that they
-------
V-44
would remain positive whatever part of that broad range were
used; even at $400,000 per statistical life saved, the estimated
benefits would be many times higher than the costs.
V.B.3. Summary of Blood Pressure Benefits
Table V-10 summarizes the benefits of reducing the numbers
of cases of hypertension, myocardial infarctions, strokes, and
deaths due to high blood pressure. As in earlier tables, these
estimates assume that misfueling is eliminated. These are con-
servative estimates for several reasons:
(1) The hypertension estimate covers only males aged 40
to 59.
(2) The other estimates cover only white males aged 40
to 59.
(3) We have not assigned any value to reduced pain and
suffering associated with hypertension, myocardial
infarctions, and strokes.
(4) We have not estimated any health benefits for adults
other than those related to blood pressure.
In addition, of course, some readers may quarrel with the value
assigned to reduced risk of mortality; we have chosen a single
value for convenience, not because we believe any particular
value can be defended strongly. Despite these limitations, the
estimated benefits of the final phasedown rule are large, total-
ing $5.9 billion in 1986.
As discussed at the beginning of this chapter, these esti-
mates should be treated as tentative. Although the two key
studies (Pirkle et al., 1985; Harlan et al., 1985) recently have
been published in peer-reviewed journals, they have not yet been
-------
V-45
TABLE V-10. Year-by-Year Estimates of Blood Pressure Benefits,
Assuming No Misfueling (millions of 1983 dollars)
Rule 1985 1986 1987 1988 1989 1990 1991 1992
Proposed 0 5,927 5,707 5,484 5,227 5,008 4,722 4,736
Alternative 2,033 4,955 5,262 5,484 5,227 5,008 4,722 4,736
Final 2,033 5,927 5,707 5,484 5,227 5,008 4,722 4,736
-------
V-46
widely available for public review. As a result, EPA has not
relied on blood-pressure-related health effects in reaching a
decision on the final phasedown rule. These potentially serious
health effects will be considered by EPA, however, in connection
with a possible ban on lead in gasoline, and extensive
review and comments will be sought on them.
-------
CHAPTER VI
BENEFITS OF REDUCING POLLUTANTS OTHER THAN LEAD
Decreasing the amount of lead in gasoline will reduce
emissions of several pollutants in addition to lead. Most of
these reductions will result from decreased "misfueling," the
misuse of leaded fuel in vehicles equipped with pollution-control
catalysts. In such vehicles, leaded gasoline poisons the cata-
lysts, greatly reducing their effectiveness in controlling emis-
sions of hydrocarbons (HC), nitrogen oxides (NOX)> and carbon
monoxide (CO). Reducing lead in gasoline should affect emissions
from misfueling in two ways. First, it will be more expensive to
produce 89 octane leaded gasoline at 0.10 gplg than to produce 87
octane unleaded gasoline. This change in relative manufacturing
costs should alter retail price differentials (although, as
discussed in Chapter 2, it may not make unleaded cheaper than
leaded at the pump), thus reducing the incentive to misfuel. In
addition, even for those vehicles that continue to be misfueled,
it will take considerably longer to destroy the effectiveness of
catalysts with 0.10 gplg leaded gasoline than it does now with
1.10 gplg.
All three of these pollutants have been associated with
damages to health and welfare, and contribute to ambient air
pollution problems covered by National Ambient Air Quality
Standards (NAAQS). CO is itself a "criteria pollutant," covered
by a NAAQS. NOX is the composite formula tor nitrogen oxide (NO)
and nitrogen dioxide (N02); N02 is covered by a NAAQS. Although
most NOX is emitted as NO, some of it is chemically transformed
-------
VI-2
in the atmosphere to NO2. NOX and HC both contribute to the
formation of ozone (03), another criteria pollutant. In addition,
certain hydrocarbons (in particular, benzene) have been linked to
cancer.
Independent of its effect on misfueling, reducing lead in
gasoline also will reduce emissions of ethylene dibromide (EDB),
which is added to leaded gasoline as a "scavenger" to reduce the
build-up of lead deposits in engines. Because EDB is added in
proportion to the amount of lead, tightening the lead standard
will reduce the amounts added to gasoline. EDB has been linked
to increased risk of cancer.
This chapter examines the impacts of reducing these
pollutants, focusing on the three associated with misfueling:
HC, NOX, and CO. Section A estimates the emissions caused by
misfueling. Section B addresses the ozone-related health and
welfare effects, which account for the vast majority of the
benefits that we were able to quantify. Section C discusses the
health and welfare gains associated with pollutants other than
ozone. We have tried to estimate as many of the effects of
these pollutants at ambient concentrations as possible, but
our quantitative estimates are subject to considerable uncer-
tainty and provide incomplete coverage of the potential effects
of emission reductions.
In both Sections B and C, the estimates of health effects
are presented in physical rather than monetary units, but non-
health effects (such as crop losses) are estimated in dollars.
Finally, in Section D, we estimate the monetized economic benefits
-------
VI-3
of eliminating misfueling using two methods. The first values
directly the health and welfare effects estimated in Sections B
and C. That method is conceptually the more appropriate one,
but it omits some important categories because of incomplete
quantification. These emissions are likely to lend a downward
bias to the direct estimates. This downward bias is most obvious
in the case of CO, for which we have not monetized any benefits,
but also is of major concern for NOX, for which we have identified
but have been unable to quantify, several potentially significant
benefit categories, in particular health effects and damages from
acid deposition. Even in the case of HC, for which we have
quantified significant ozone-related benefits, the estimates may
be biased downward significantly because of our inability to
quantify ozone's impacts on chronic health conditions and forests,
nor have we estimated direct health effects of any HCs other than
benzene. Further, there is considerable uncertainty in those
categories we have included.
The second method monetizes the emission reductions using
the values implied by the cost of the pollution control equipment
needed to meet the emission limits set by Congress. Our final
monetized benefit estimates, used in later chapters to compute
the total and net benefits of alternative rules, are the averages
of these two methods.
The basic methodology used in this chapter is the same as
that employed in Schwartz et al. (1984) and in the preliminary
RIA. The analysis has been refined in several areas, however.
-------
VI-4
First, to estimate emissions associated with misfueling, we have
used the results of EPA's 1983 tampering and misfueling survey,
the results of which were not available for use in the earlier
documents. Second, the estimates of the health effects associated
with ozone exposure rely on an updated statistical model. Third,
in this document we do not rely on a peoperty-value-based estimate
of NOX estimates, because closer examination of the underlying
study suggested that it had not adequately accounted for con-
founding factors, in particular other pollutants whose concentra-
tions may covary with NOX. Finally, and most importantly,
additional and more detailed analysis of the impact of HC and NOX
emissions on ozone has caused us to revise downwards significantly
our estimates of the rule's impact on ozone concentrations.
Throughout this chapter, we estimate the effects and monetized
benefits of eliminating misfueling altogether. For lead levels
other than a complete ban on all leaded gasoline, this is probably
an overly optimistic assumption. Chapter VIII presents estimates
based on a broader range of alternative assumptions about the
impacts of different rules on misfueling.
VI.A. Emissions Associated with Misfueling
"Misfueling" or "fuel switching" refers to the use of leaded
gasoline in a vehicle originally designed and certified to use
unleaded gasoline. Because leaded regular gasoline is cheaper
and higher in octane than regular unleaded, some drivers deliber-
ately misfuel their vehicles in an attempt to reduce expenses or
to improve vehicle performance.
-------
VI-5
Misfueling can occur by removing or damaging the nozzle
restrictors installed in the fuel inlets of vehicles with cata-
lytic converters, by using an improper size fuel nozzle, or by
funneling leaded fuel into the tank. Sometimes gasoline retailers
sell gasoline that is mislabeled or contaminated (U.S. EPA,
1983a), but this accounts for less than 1 percent of misfueling.
It is illegal for service stations or commercial fleet
owners to misfuel or to allow the misfueling of vehicles origin-
ally equipped with catalytic converters. Federal law does not
apply to individuals who misfuel their own vehicles, however.
Using leaded gasoline in vehicles with catalytic converters
damages this pollution control equipment, and can increase
emissions of HC, CO, and NOX by as much as a factor of eight.
Table VI-1 shows the emissions increases caused by misfueling on
a per-mile basis. The estimates distinguish between pre- and
post-1981 vehicles because emission standards changed in that year,
leading to changes in the design of catalysts and other emission
control devices. In vehicles manufactured before 1981, misfueling
has no effect on NOX emissions, but does cause relatively
large increases in HC and CO emissions.
Misfueling is a significant problem. Several recent surveys
by EPA have shown that a substantial number of vehicles are mis-
fueled with leaded gasoline. According to the 1983 survey
(U.S. EPA, 1984e), about 15.5 percent of light-duty vehicles
designed to use unleaded gasoline are misfueled with leaded. The
1982 survey (U.S. EPA, 1983a), showed a lower rate, about 13.5
-------
VI-6
TABLE VI-1. Increase in Emissions Due to Misfueling (grams/mile)
Light-Duty Vehicle Model Years HC CO NOX
1975 to 1980 2.67 17.85 0.0
1981 and later 1.57 11.07 0.71
Source: U.S. EPA, Office of Mobile Sources, "Anti-Tampering and
Anti-Misfueling Programs to Reduce In-Use Emissions from
Motor Vehicles," May 23, 1983.
-------
VI-7
percent. Misfueling rates apparently vary by the age of the
vehicle, by whether it is in an area with an Inspection and
Maintenance (I/M) mobile source enforcement program, by whether
it is part of a commercial fleet, and other factors. Table VI-2
provides 1983 misfueling rates by model year of vehicle and by
T/M status. We assumed for our analysis that the rates of
misfueling by age of vehicle would stay constant in the absence
of new regulations.
The EPA surveys probably underestimate real misfueling rates
by a significant margin, primarily because vehicle inspections for
misfueling are voluntary, which would bias the results downward
(assuming that misfuelers are less likely to agree to have their
vehicles tested). In some areas, the rates of drivers refusing
inspections were very high. In the 1982 survey, the refusal rates
ranged from 1 to 8 percent in I/M areas, and from 3 to 44 percent
in non-I/M areas.
To estimate the reduction in emissions that would be achieved
by eliminating misfueling, we combined the data in Tables VI-1 and
VI-2 with our fleet model (described in the Appendix), which pro-
jected the number of vehicles of each model year and their annual
mileage. For each year of our projection, we first estimated the
number of vehicles that would have misfueled for the first time
in that year. (We assumed that no emission reductions would
result from stopping the misfueling of vehicles that already had
their catalysts destroyed by misfueling in earlier years.) We
then projected over the remaining lifetime of the vehicle the
expected excess pollutants it would have emitted due to misfuel-
-------
VI-8
TABLE VI-2. Misfueling Rates in 1983 (percent)
Model
Year
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
Weighted
Overall
Misfueling Rates
1.6
5.1
8.0
10.0
9.0
19.6
19.0
22.6
25.5
25.9
Average:* 15.5
I/M Areas
8.7
4.8
3.2
7.9
6.2
17.9
7.3
23.2
13.9
16.0
Non-I/M Areas
0.0
5.2
9.0
10.5
9.7
20.0
21.6
22.4
28.0
28.0
This weighted average does not account for the number of miles
driven by each model year.
-------
VI-9
ing. The projections account for the facts that older vehicles
drive fewer miles per year, that survival rates decline with age,
and that the efficiency of emission control devices deteriorates
with age. These projected emission streams were then discounted
(at a 10 percent real rate) back to the year in question. Table
VI-3 presents the resulting year-by-year estimates of the dis-
counted emissions avoided by eliminating misfueling. Note that
the estimate for each year is not of actual emission reductions
achieved in that year, but rather the discounted value of emis-
sion reductions due to stopping the first-time misfueling of
vehicles in that year. The numbers of tons of HC and CO controlled
remain fairly constant from 1986 through 1992, at over 300,000
tons of HC and more than 2.5 million tons of CO. The estimates
of reduced NOX emissions increase, from 94,000 tons in 1986 to
150,000 tons in 1992, because of the increase in the proportion
of post-1981 vehicles. (Recall that misfueling does not increase
NOX emissions in vehicles manufactured before 1981.)
VI.B. Health and Welfare Effects Associated with Ozone
Hydrocarbons and nitrogen oxides react photochemically to
form ozone ("smog"). Ozone in turn affects health, materials
damage, and vegetation. To estimate the ozone-related effects of
reducing emissions of HC and NOX, we employed a two-step process.
First, as described below in Section VI.B.I, we estimated the
health and welfare effects of a 1 percent change in ozone. Second,
as discussed in Section VLB.2, we estimated the relationship
between reducing HC and NOX and the subsequent decrease in ozone.
-------
VI-10
TABLE VI-3. Year-by-Year Estimates of Reductions in Emissions,
Assuming No Misfueling (thousands of metric tons)
Pollutant 1985 1986 1987 1988 1989 1990 1991 1992
Hydrocarbons 155 305 303 303 303 308 320 331
Nitrogen Oxides 40 94 107 119 130 139 145 150
Carbon Monoxide 1,067 2,116 2,114 2,122 2,131 2,174 2,255 2,333
Total 1,262 2,515 2,524 2,544 2,564 2,621 2,720 2,814
-------
VI-11
The material in these sections closely parallels McGartland and
Ostro (1985). Several of the studies relied upon are EPA con-
tractor reports in progress or in draft; as such they have not
undergone full peer review and should be considered preliminary.
VI.B.1. Effects of a 1 Percent Reduction in Ozone
In calculating the effects of a 1 percent change in ozone, we
relied primarily on dose-response estimates. That is, we applied
change in ambient levels. Occasionally, as a validity check of
the benefit estimates, we interpolated from existing aggregate
damage estimates to project the impacts of a single pollutant or
of a given change in ambient levels. Regardless of the approach,
the benefit estimates are uncertain and should be interpreted
with caution. Unless noted otherwise, we assumed a constant
benefit per ton of pollution control over the relevant range.
The effects of ozone on human health, vegetation, materials,
and ecosystems were summarized in the EPA Air Quality Criteria
for Ozone and Other Photochemical Oxidants (U.S. EPA, 1978). In
addition, we have relied on the considerable amount of research
that has become available since that document was finished. As
part of EPA's periodic review of the ozone NAAQS, the Office of
Research and Development currently is updating the Criteria
Document. Nothing in this report is intended to prejudge or
supercede the outcome of that process.
VI.B.I.a. Health Effects of Reducing Ozone
Studies of the effects of ozone on human health have
investigated the relationships between changes in ozone concen-
-------
VI-12
trations and changes in lung function; decrements in physical
performance; exacerbation of asthma; incidence of headaches;
respiratory symptoms, such as coughing and chest discomfort;
eye, nose, and throat irritation; and changes in blood parameters
(U.S. EPA, 1978; Goldstein, 1982; Ferris, 1978).
Uncertainty remains about whether a threshold level exists
for ozone and, if so, at what level. For example, McDonnell et al.
(1983) found a nonlinear relationship between health and ozone
exposure that "flattened" at ozone levels below 0.18 parts per
million (ppm) — a level above the ambient concentrations in most
metropolitan areas. If such a threshold exists, the health bene-
fits of reducing ozone from its current levels would be minimal.
Population studies by Zagraniski et al. (1979) and Lebowitz et al.
(1984), however, suggest effects may be occurring at ambient
levels as low as 0.08 ppm. Moreover, other studies (Portney and
Mullahy, 1983; Hasselblad and Svendsgaard, 1975) do not support
the existence of any threshold for health effects.
Hammer et al. (1974) found associations between increased
oxidants and respiratory symptoms (such as cough and chest discom-
fort) and other symptoms (such as eye irritation and headache) in
young, healthy adults. They obtained the symptom rates from daily
diaries and adjusted them by excluding days on which subjects
reported fevers. Makino and Mizoguchi (1975) found a correlation
between oxidant levels and eye irritation and sore throats in
Japanese school children. Lippmann et al. (1983) and Lebowitz
et al. (1982, 1983, 1984) found evidence of decreased athletic
performance, increased prevalence of acute symptoms, and
dysfunction of pulmonary systems resulting from ozone exposure.
-------
VI-13
In addition to these studies of the general population,
Whittemore and Korn (1980), Linn et al. (1981), Bates and Sizto
(1983), and others have shown that asthmatics and people with
other chronic respiratory diseases may he particularly sensitive
to ozone or other oxidants. Even low levels of exposure to photo-
chemical oxidants have been shown to provoke respiratory symptoms
in individuals with predisposing factors, such as smoking or
respiratory illness (Zagraniski et al., 1979).
There is also evidence linking reduced respiratory function
-- measured as Forced Expiratory Volume (FEV) and Forced Venti-
lating Capacity (FVC) -- to ozone exposure. For example, McDonnell
et al. (1983) reported an association for normal subjects while
exercising. Folinsbee et al. (1984), Horvath et al. (1979), and
Adams and Schelegle (1983) found an association between decrements
in FEV and ozone exposure.
Unfortunately, it is difficult to estimate from these studies
the potential economic benefits of reducing health effects related
to ozone exposure because they did not estimate dose-response
functions. Most were designed to investigate potential thresholds,
or simply to determine if any relationship existed between ozone
and particular effects. In addition, studies using lung function
changes as the health endpoint fail to provide a measure that can
be valued in economic terms.
Recent work by Portney and Mullahy (1983, 1985) at Resources
for the Future (RFF) is an exception. They considered the effect
of alternative levels of ozone on various health measures, combining
individual health data from the Health Interview Survey (HIS)
-------
VI-14
with data on pollution concentrations during the same period
covered by the survey. In the HIS, interviewees provided infor-
mation on their health status during the two weeks preceding the
survey. As their health measure, Portney and Mullahy focused on
the number of days of restricted activity due to a respiratory
condition (RADRESP). The RADRESP measure included days when the
symptoms were relatively minor, as well as those when they were
serious enough to confine individuals to bed or to make them
miss work.
Portney and Mullahy regressed RADRESP on a dozen or more
independent variables, including socioeconomic and demographic
factors, chronic health status, urban variables, and other pol-
lutants, as well as ozone. As their ozone measure, they used
the daily maximum one-hour concentration (measured in parts per
million) averaged over the two-week period covered for the indi-
vidual. They considered several different specifications and
functional forms. In their ordinary least squares (OLS) regres-
sions, they tried various forms of the ozone measure, including
the square and the square root as well as the untransformed vari-
able. The first part of Table VI-4 summarizes the OLS results.
In all three specifications, the ozone coefficient was positive,
but not significant at the 95 percent confidence level. For the
specifications using the linear ozone term, the coefficient on
ozone represents the average change in RADRESPs per person per
two weeks for a one ppm change in ozone. Thus, for example,
that coefficient predicts that reducing the average daily maximum
-------
VI-15
TABLE VI-4. Regression Results for Portney and Mullahy Study
on Respiratory Symptoms Related to Ozone
Ozone Specification
OLS
Linear
Square root
Squared
Ozone
Coefficient
1.2185
0.8076
0.4667
t-statistic
1.13
1.66
0.07
F-statistic
2.743
2.867
2.636
Poisson 6.8827 1.97 N.A.
(log-likelihood ratio = -1395.4)
-------
VI-16
ozone concentration by 0.01 ppm for one year for a population of
1 million adults would decrease the number of RADRESPs by 316,800
(= 0.01 x 1.22 x 1,000,000 x 52/2).
In subsequent analysis, Portney and Mullahy (1985) estimated
the relationship using a Poisson model, which can be written as:
E(RADRESP) = exp(XB),
where E(RADRESP) is the expected number of RADRESPs and XB is the
vector of independent variables and their coefficients. The
second part of Table VI-4 summarizes the results of the Poisson
model. Because the Poisson model is nonlinear, the ozone coeffi-
cient is slightly harder to use for extrapolation. The Poisson
model, however, appears to fit the data better; RADRESPs have a
Poisson-like distribution.
In separate models estimating RADRESPs for children, Portney
and Mullahy (1983) did not find any consistently significant
effects. As a lower bound, therefore, we assumed no affect on
RADRESP for children in the general population. However, incom-
plete data for children and the reliance on parents to report
child-related health effects may explain this result; a restric-
tion in activity probably was less likely to be reported for a
child. Public health scientists continue to debate whether
children are as susceptible to ozone as adults. Older children,
for example, may not be as susceptible because they typically
have large excess lung capacity. On the other hand, damages to
the lungs of a child may result in a chronic respiratory condition
-------
VI-17
in adulthood. Therefore, to place a plausible upper bound on our
estimates, we applied the adult coefficients to children as well.
To estimate the change in RADRESPs due to a 1 percent change
in ozone, we simulated the change using the data on individual
exposures and characteristics constructed by Portney and Mullahy.
Their study matched the 1979 HIS with air quality data, weather
stations, and other area-specific data. Using the estimated
regressions, we rolled back the exposure of each person in the
data base by 1 percent. Using data from the Census Bureau, we
assumed a population of 230 million in 1984, with 70 percent of
the total above age 17.
The changes in total annual RADRESPs for adults predicted by
the Poisson and linear models were quite similar, 2.1 and 2.4
million, respectively. Because the Poisson model provided a
better fit of the data, we used it as the basis for our estimates.
For our high and low estimates of adult effects, we used plus or
minus one standard deviation of the ozone coefficient. For child-
ren, our low estimate was zero and our high estimate was 0.90
million cases (based on the adult coefficient). For our medium
or point estimate, we used the midpoint of those extremes, 0.45
million cases. These results should be interpreted cautiously,
because cross-sectional studies of this type can be extremely
sensitive to model specification, functional form, omitted
and confounding variables, and the ambient air monitors used.
Portney and Mullahy (1985) also used a multinomial logit
model to estimate the marginal impact of ozone on the two types
of RADRESPs — the more serious ones resulting in a day of bed
-------
VI-18
rest or of lost work, and the less serious ones that resulted in
a more minor restriction of normal activity. That analysis
suggested that a marginal change in ozone was three times as
likely to cause a day of minor restricted activity than a day of
bed rest or lost work.
To provide an alternative estimate of respiratory conditions
and a separate estimate of nonrespiratory irritations, we used the
results of a statistical reanalysis of the Hammer et al. (1974)
study discussed earlier. In an unpublished paper, Hasselblad
and Svendsgaard (1975) fit simple logistic regressions to estimate
the relationship between ozone concentration (measured as a
daily maximum hourly concentration) and eye irritation, headache,
coughing, and chest discomfort. The probability of a response
at an ozone level, X, measured in parts per hundred million
(pphm), was given as:
p(X) = C + (1 - C)/[l + exp (-A - BX)]
Table VI-5 presents the estimates of the parameters for the four
outcome measures. These coefficients must be interpreted cau-
tiously, because Hasselblad and Svendsgaard did not control for
some possible confounding variables, in particular temperature
and humidity. In a later, published paper, Hasselblad (1981)
fit multiple logistic regression models to these same data, but
that study paper does not report the regression coefficients we
needed to make our estimates.
-------
VI-19
TABLE VI-5. Regression Coefficients Relating Respiratory
and Non-respiratory Symptoms to Ozone
Condition
Respiratory Effects
Cough
Chest discomfort
Non-respiratory Effects
Eye irritation
Headache
A B
-2.98 0.0092
-3.53 0.0023
-4.96 0.0907
-4.88 0.0470
C
0.0450
0.0166
0.0407
0.0976
-------
VI-20
We used these estimated dose-response functions with the
individual information in the Portney and Mullahy data set to
simulate the effects of a 1 percent reduction in ozone. For
coughs and chest discomfort, the Hasselblad and Svendsgaard coeffi-
cients yielded a total of 1.54 million adult cases per year for
a 1 percent reduction in ozone, compared to the estimate of 2.1
million based on the Portney and Mullahy Poisson model. These
estimates are remarkably consistent, as the RADRESP measure used
by Portney and Mullahy included other symptoms besides cough and
chest discomfort. In addition, the Hammer et al. sample used by
Hasselblad and Svendsgaard consisted of student nurses, who were
young and generally healthy, while the Portney and Mullahy sample
was more representative of the general population. For those
reasons, our estimates for respiratory conditions rely solely on
Portney and Mullahy's results.
The results of Hasselblad and Svendsgaard also can be used to
estimate the number of nonrespiratory conditions, such as eye
irritation and headache, possibly related to exposure to ozone
and other photochemical oxidants. There is evidence suggesting
that these symptoms are not related to ozone per se, but rather
to other oxidants, such as peroxyacetyl nitrate (PAN), whose
production may be proportional to that of ozone.
To account for this possibility, we used the estimates based
on Hasselblad and Svendsgaard as point estimates for nonrespiratory
irritations (headache and eye irritation). Unfortunately, these
researchers did not report the standard errors, so a confidence
interval could not be determined.
-------
\ / VI-21
These ozone health benefits reflect the likely acute effects
generated by intense, short-term exposure to ozone. Long-terra
exposure to ozone also may affect the health of some people, but
the epidemiological evidence on chronic ozone effects is sparse.
One of the available studies, Detels et al. (1979), compared the
effects of prolonged exposure to different levels of photochemical
oxidants on the pulmonary functions of both healthy individuals and
individuals with chronic obstructive pulmonary disease. Persons
exposed to an annual mean of 0.11 ppm of oxidant, compared to a
control group exposed to 0.03 ppm of oxidant, showed statistically
significantly increased chest illness, impairments of respiratory
functions, and lower pulmonary function.*
While the epidemiological evidence of the effects of long-term
exposure to ozone is sparse, several animal experiments have demon-
strated effects on lung elasticity, blood chemistry, the central
nervous system, the body's ability to defend against infection, and
the rate at which drugs are metabolized (U.S. EPA, 1983f). Unfor-
tunately, it is not possible to extrapolate those results to humans.
Therefore, we could not quantify the chronic health effects attri-
butable to ozone, but we believe that some of these effects may be
present at current ambient levels.
* At workshops related to the development of the Criteria Document
for ozone, some shortcomings in this analysis were noted. For
example, the study group was also exposed to higher levels of
NO2 and 804, and there were some questions about the adequacy of
the measurement of ozone exposure, about the subject selection,
and about the test measures. Although it is both reasonable and
likely that long-term exposures affect health, the failure to
correct for the effects of other pollutants raises uncertainties
about the specific findings.
-------
VI-22
Table VI-6 sununarizes our estimates of the acute health
effects of a 1 percent change in ozone. The estimates of respir-
atory effects are based on the Portney and Mullahy results, while
the nonrespiratory effects are derived from Hasselblad and
Svendsgaard's analysis. In the latter case, high and low estimates
are not reported because Hasselblad and Svendsgaard did not calcu-
late standard errors for their estimates. As discussed above,
these estimates omit any quantification of chronic health effects.
VI.B.l.b. Ozone Agricultural Effects
Ozone, alone or in combination with sulfur dioxide and
nitrogen dioxide, is responsible for most of the U.S. crop damage
attributed to air pollution (Heck et al., 1983). Ozone affects
the foliage of plants by biochemical and cellular alteration,
thus inhibiting photosynthesis and reducing plant growth, yield,
and quality.
Early studies of ozone-related damages used generalized
relationships between ozone concentrations, yield, and economic
loss. Insufficient information precluded the construction of an
economic model with credible dose-yield data. Thus, for example,
Freeman (1982), in a general survey of the literature, could only
conclude that the total agricultural damages from ozone ranged
from $1.0 to $4.0 billion in 1978 dollars.
Recent work by the National Crop Loss Assessment Network
(NCLAN) suggests that prior studies have underestimated ozone-
related damages. NCLAN'S estimated dose-yield functions for
soybeans, wheat, corn, peanuts, cotton, barley, and sorghum have
-------
VI-23
TABLE VI-6. Estimated Health Effects of a 1 Percent Reduction in
Ozone (millions of days per year)
Condition
Low
Estimate
High
Estimate
Medium
Estimate
Respiratory Effects
Bed Rest/Work Loss
Adults
Children
Subtotal
Minor Restrictions
Adults
Children
Subtotal
Nonrespiratory Effects
Headaches
Eye Irritation
Subtotal
N.A.
N.A.
N.A.
N.A.
-------
VI-24
provided more accurate information on ozone's effects on crops.
Kopp (1983, 1984) and Adams et al. (1984) incorporated these
functions into models of agricultural production and demand to
estimate the benefits of ozone reduction strategies.
Kopp constructed a detailed microeconomic model of farm
behavior for over 200 producing regions in the United States.
The NCLAN dose-yield functions are directly incorporated in Kopp's
model of the supply side of each crop for each region. Because
estimates of the demand and supply elasticities for these crops
are used in the analysis, it gives a good indication of the actual
change in economic welfare. Kopp's simulations suggest that a 1
percent reduction in ozone would produce total benefits of roughly
$110 million (1983 dollars) per year for the seven major crops
covered by NCLAN, as shown in Table VI-7. These seven crops
accounted for only about 80 percent of the total value of U.S.
crop production (USDA, 1982). If we increase Kopp's estimate by
assuming that ozone damages to all other crops occur in the same
proportion as their relative value, we conclude that the benefits
of a 1 percent change in ozone are roughly $137 million annually.
Adams et al. (1984) used a different approach. By
incorporating the NCLAN dose-yield functions into an existing
quadratic programming model, they calculated ozone benefits for
six of the seven crops covered by Kopp (they did not include
peanuts). They estimated that a 10 percent reduction in rural
ozone would result in annual benefits of roughly $674 million
(1983 dollars). Assuming linearity and increasing the estimate
to account for omitted crops, we estimated $90 million in benefits
-------
VI-25
TABLE VI-7
Crop
Annual Agricultural Benefits of a 1 Percent Ozone
Reduction (millions of 1983 dollars)
Estimate
Soybeans
Corn
Wheat
Cotton
Peanuts
Sorghum
Barley
Total assessed
50.8
7.6
21.1
20.4
5.0
4.6
0.2
109.8
-------
VI-26
for a 1 percent change in ozone levels. Unfortunately, Adams
et al. did not calculate the benefits on a crop-by-crop basis,
so it is not possible to make a detailed comparison with Kopp's
estimates.
For our best estimates, we used Kopp (1983, 1984) and Adams
et al. (1984), because they used the superior NCLAN data and
based their estimates on economic measures of welfare loss. The
two estimates are fairly close. We concluded that a 1 percent
reduction in rural ozone would produce $90 to $140 million in
annual agricultural benefits per year, with a point estimate of
$114 million. We qualify these numbers by noting that the esti-
mates do not reflect unreported small "truck farm" sales or any
averting activities that farmers may undertake, such as planting
pollutant resistant crops; we believe that these categories are
likely to be small. On the other hand, the welfare measures do
not reflect either the effects of crop subsidy programs at the
state and federal levels, or the effects of drought, both of which
are likely to reduce the marginal welfare impacts of ozone. The
effects of subsidies are likely to be particularly significant.
VI.B.I.e. Ozone Effects on Nonagricultural Vegetation
Forests and ornamental plants also may suffer substantial
damages from exposure to ozone. The preliminary draft of the
Ozone Criteria Document discusses the issue:
The influence of 03 on patterns of succession and
competition and on individual tree health is causing
significant forest change in portions of the temperate
zone.... Long-term continual stress tends to decrease
the total foliar cover of vegetation, decrease species
richness and increase the concentrations of species
-------
VI-27
dominance by favoring oxidant-tolerant species. These
changes are occurring in forest regions with ozone
levels (1-hour maximum) ranging from 0.05 ppm (111
ug/m3) to 0.40 ppm (785 ug/m3) (U.S. EPA, 1983).
Additional evidence of significant damages from ozone associated
with nonagricultural vegetation is provided by McLaughlin et al.
(1984).
Unfortunately, no careful quantitative studies of the type
done by NCLAN have been performed for nonagricultural vegetation.
Heintz et al. (1976) have estimated losses to ornamental plants of
$100 million per year in 1973 dollars. Inflating to 1983 dollars
using the Farm Products Index, and assuming linearity, yields
estimated annual benefits of $1.4 million for the reduction in
damages to ornamentals from a 1 percent reduction in ozone.
Damage to forests is potentially a much larger concern in
terms of reduced production and decreases in recreation and aethes-
tic values. In a very small contingent valuation study, Crocker
and Vaux (1983) found that the shift of an acre of the current mix
of severely, moderately, and unharmed timberland in the San
Bernardino National Forest into the unharmed category would gener-
ate additional annual recreational benefits of between $21 and $68
per acre per year. These findings are difficult to generalize for
the rest of the nation because ambient ozone levels are unusually
high in the San Bernardino area (and the authors provide no dose-
response function for extrapolating to areas with lower concentra-
tions), and because other site attributes and visitors' socioeco-
nomic characteristics have very large and significant effects on
users' willingness to pay to reduce damages to forests.
-------
VI-28
We also lack data on the impact of ozone on commercial
forests. Most commercial forests are located in areas with
relatively low ozone concentrations, so damages may be small.
Weyerhauser Corporation staff recently reported "negligible or
nonexistent" damages from ozone (personal communication from Jack
Larson of Weyerhauser). In light of the NCLAN findings with
respect to crops, however, the possibility of significant ozone
damage to forests cannot be dismissed.
VI.B.l.d. Ozone Materials Damage
Ozone directly damages many types of organic materials,
including elastomers, paint, textile dyes, and fibers. It can
increase the rigidity of rubber and synthetic polymers, causing
brittleness, cracking, and reduced elasticity. Ozone exposure
also can generate other effects, such as avoidance costs (pur-
chase of specially resistant materials) and aesthetic losses.
Only the direct costs are incorporated in this analysis, however.
In his survey of the literature, Freeman (1982) suggested
that annual materials damages from oxidants and NOX amount to
approximately $1.1 billion (1978 dollars). We updated that esti-
mate using the U.S. Government price indices for rubber and tex-
tile products and the index of personal consumption expenditures
for durable goods; this yielded an estimate of $2.25 billion for
1983. Assuming linearity, a 1 percent ozone reduction generates
a benefit of roughly $22.5 million annually.
-------
VI-29
We obtained an alternative estimate of the benefits of
reduced material damage by using dose-response information
incorporated in the 1978 Criteria Document for ozone. The text
contains per capita economic damage functions for elastomers,
textiles, industrial maintenance, and vinyl paint. We used a
population-weighted mean value of ozone of 0.03 ppm (60 ug/m3)
from the draft of the new Criteria Document (U.S. EPA, 1984), a
population estimate of 230 million, and the indices cited above.
This method yielded annual benefits of $15 million (in 1983
dollars) for a 1 percent reduction in ozone. Averaging these
two estimates (with slightly more weight given to the lower esti-
mate which does not include any NOx-related damages) yields a
point estimate of $18 million annually, with a range of $15 to
$22.5 million.
VI.B.I.e. Summary of Benefits of a 1 Percent Change in Ozone
Table VI-8 summarizes our estimates of the effects of a 1
percent reduction in ozone. The estimated health effects include
roughly 2.5 million days of respiratory symptoms resulting in some
restriction in activity, and 4.4 million days of non-respiratory
irritations (headaches and eye irritation). On the nonhealth side
of the ledger, increases in agricultural crop production dominate,
with a total of $114 million. Materials damage is next, with $18
million, while ornamental plants contribute $1.4 million. In all
cases, the estimates are subject to considerable uncertainty; the
"high" and "low" estimates provide only a partial indication of
that uncertainty, as they reflect only the ranges in available
-------
VI-30
TABLE VI-8. Summary of Estimated Effects of a 1 Percent
Reduction in Ozone
Effect
Low High Medium
Estimate Estimate Estimate
Health (millions of cases)
Respiratory Symptoms
Bed rest/Work loss
Minor restrictions
0.36
1.07
1.23
3.60
0.64
1.91
Nonrespiratory Symptoms
N.A.
N.A.
4.37
Other (millions of dollars)
Agricultural crops
Ornamental plants
Materials damage
Unquantified Benefit Categories
Chronic health effects
Forest damage
90.0 140.0 114.0
N.A. N.A. 1.4
15.0 22.5 18.0
-------
VI-31
estimates or statistical uncertainty in the parameter estimates
from particular studies. A more complete accounting for
uncertainty — which would include different functional forms
for the dose-response functions, omitted categories, etc. —
would yield substantially broader ranges. The omitted categories,
in particular chronic health effects and damage to forests,
however, suggest that these estimates are too low, perhaps by a
substantial margin.
VLB.2. Linking NOX and HC Reductions to Ozone Effects
The previous section estimated the effects of reducing ozone.
Linking those effects to emissions of HC and NOX requires estimat-
ing the relationships between those pollutants and ozone concen-
trations. Section VI.B.2.a describes the general process of ozone
formation, while Section VI.B.2.b presents rough quantitative esti-
mates of the impact of HC and NOX emissions on ozone concentrations
in both urban and rural areas.
VLB.2.a. The Process of Ozone Formation
Ozone changes are influenced by the amount of solar
radiation and the ratio of, and changes in, the concentrations of
NOX and HC. The general process by which ozone (03) is formed is
illustrated by the N02-NO-03 cycle. These reaction equations can
be written as:
(VI-1) NC-2 + sunlight —> NO + O
(VI-2) 0 + 02 '—> 03
(VI-3) 03 + NO > N02 + 02
-------
VI- 3 2
Ozone is not emitted in any measurable quantity. The first
two reactions represent its only significant source. However,
these reactions are fully reversed by the third one. Because NO
comprises roughly 90 percent of man-made NOX emissions, reaction
(VI-3) suggests that additional NOX emissions would reduce ozone
concentrations by scavenging the ozone for one of the oxygen
atoms, thereby producing ©2 and N02 . Therefore, increases in NOX
cannot increase ozone, unless some other species oxidizes NO to
NO2 without scavenging ozone for the necessary oxygen atom. This
additional reaction is:
(VI-4) Organic radicals (from HC) + NO — > N©2 + miscellaneous
products
With reactions (VI-1), (VI-2), and (VI-4), NO2 is recycled,
allowing for the generation of excess 03. That is, HC permits
ozone formation while bypassing the ozone destructive step of NO
+ 03 — > NO2 +02. In short, with a plentiful supply of HC , NOX
will not have to scavenge ozone to form NO2 . Instead, NOX com-
bines with the HC to form additional NO2 , which in turn makes
more ozone. Therefore, in areas with "excess" HC, NOX control
will reduce ozone, but controlling HC will be relatively ineffec-
tive. However, when there is not enough HC to react with avail-
able NOX, NOX will scavenge ozone to form NO2 • In this situation,
NOX control will result in less ozone scavenging (and more ozone).
HC control, on the other hand, will be effective in reducing
ozone because the greater the reduction in HC, the more NOV is
X
forced to scavenge ozone.
-------
VI-33
Atmospheric scientists have shown that the HC/NOX ratio is
relatively small in metropolitan areas. Thus, in such areas, a
reduction in HC will reduce ozone, but depending on the specific
HC/NOX ratio, controlling NOX may increase ozone. In rural areas,
HC/NOX ratios are larger. Therefore, if the ozone in such areas
results primarily from local emissions, controlling local NOX
emissions may be a more effective way of reducing ozone than local
HC controls.
VI.B.2.b. Quantitative Estimates of Impacts of HC and NOX on Ozone
Several models have been developed to simulate the effect of
changes in HC and NOX on ozone (U.S. EPA, 1983f, 1984b; Systems
Applications, Inc. [SAI], 1984). EPA's Office of Air Quality
Planning and Standards (OAQPS) also has developed three sets of
regional Empirical Kinetic Modeling Approach (EKMA) isopleths for
use in the ozone NAAQS review. Finally, the data base obtained
from state implementation plans (SIPs) for ozone summarizes all of
the modeling results from the SIPs. A review of these models and
their general results is provided in McGartland and Ostro (1985).
From these models, we generated estimates of the "average" impact
of changes in HC and NOX on urban and rural peak ozone concentra-
tions (relevant for health effects and materials damage) and on
rural average ozone concentrations (more relevant for agricultural
effects), measured in terms of elasticities (i.e., the percent
change in the ozone variable with respect to a 1 percent change
in emissions of the pollutant). The results are summarized in
Table VI-9; the discussion of their derivation follows.
-------
VI-34
TABLE VI-9. Estimated Ozone Reductions from 1 Percent Reduction in
Rural and Metropolitan HC and NOy (percent reduction)
Reduction
in Peak
Urban Ozone
Reduction
in Peak
Rural Ozone
Reduction
in Average
Rural Ozone
HC
1 Percent Reduction
in Urban HC
0.60
0.05
0.13
1 Percent Reduction
in Rural HC
0
0.36d
0.36d,e
1 Percent Reduction
in Urban
1 Percent Reduction
in Rural N0«
-0.3 to -0.05
0
-0.1 to +0.1
(0.04)b
0.05 to 0.2d
(0.15)c
-0.09
0.05 to 0.2d,e
(0.15)c
a Estimates varied by a fairly wide range.
estimate of -0.1.
We used the point
Sparse evidence does not permit the estimation of a single
point estimate with a reasonable degree of certainty. In many
areas (e.g., the Northeast Corridor), wind trajectories would
probably transport metropolitan ozone to other urban centers
or over the ocean. In these cases, rural ozone would not
change. In other areas, however, the increased ozone from
urban NOx reductions would travel to rural areas, but lower
rural N©2 levels would reduce ozone. Therefore, the overall
effect could be positive or negative. As a conservative estimate,
we assumed that a 1 percent uniform reduction in metropolitan
reduces rural ozone by 0.04 percent.
We used a conservative high point estimate
c Estimates varied.
of 0.15.
d Results from ongoing analysis at SAI will soon narrow the
range of estimates.
e The lack of evidence made it necessary to assume that average
rural ozone changes by the same percent as peak rural ozone.
-------
VI-35
The ozone-SIP data base presents estimates of how urban HC
affects peak urban ozone for every city not attaining the ozone
standard. Analysis of these data by OAQPS (U.S. EPA, 1984)
suggests that a 1 percent reduction in hydrocarbon emissions will
reduce metropolitan ozone on average by 0.6 percent. This estimate
is based on many EKMA urban simulations throughout the U.S.; as
such, it represents the best available estimate available of the
percent reduction in urban ozone from a 1 percent reduction in
HC. However, there is considerable variance from area to area.
Estimates of how urban HC affects rural average and peak
ozone levels are presented by SAI (1984). SAI simulated regional
air quality under four basic scenarios — mornings and afternoons,
winter and summer. The two summer scenarios (relevant for esti-
mating agricultural benefits) indicate that average ozone changes
by 0.13 percent for every 1 percent change in urban HC during the
agricultural growing season. Peak rural ozone changed by much
less: roughly 0.05 percent for every 1 percent change in urban
HC. This is mainly because the HC and ozone plumes are widely
dispersed by the time they reach rural areas.
Estimates of how rural HC affects rural ozone generally were
not available. Therefore, we used the three sets of regional EKMA
isopleths developed by OAQPS. By estimating data points at HC/NOX
ratios typical of rural areas, we estimated that, on average, a
1 percent reduction in rural HC will reduce peak rural ozone by
0.36 percent. EKMA models are not, however, suitable for estimat-
ing changes in average ozone. As our best estimate, we assumed
that average ozone changes by the same percentage as the peak.
-------
VI-36
For rural areas this may be a reasonable approximation, as rural
ozone levels do not fluctuate as much as urban ozone levels and
thhe plumes would still be highly concentrated in the same area.
To estimate the effects of NOX reductions on urban and rural
ozone, we used SAI (1984) and the EKMA isopleths developed by
OAQPS. By taking an average of data points from the isopleths, we
calculated that, on average, a 1 percent reduction in metropolitan
NOX will increase ozone by 0.1 percent.
SAI estimated how rural ozone is affected by changes in NOX
emissions. A 1 percent reduction in urban NOX increases average
rural ozone by 0.09 percent. Peak rural ozone would probably not
change as much because the urban plume is widely dispersed when
it reaches the rural areas. The SAI analysis showed average rural
ozone increasing by a small amount in some situations, but, on
average, there probably is a small decrease in rural ozone given a
reduction in urban NOX emissions. As a conservative estimate, we
assumed that a 1 percent reduction in urban NOX reduces peak rural
ozone by 0.04 percent.
We again relied on the EKMA isopleths generated by OAQPS to
estimate how rural NOX emissions affect peak rural ozone. On
average, the isopleths indicate that a 1 percent reduction in
rural NOX will reduce peak rural ozone by 0.1 to 0.2 percent. We
used an estimate of a 0.15 percent decrease in peak ozone for
every 1 percent decrease in rural NOX. We also used that estimate
to quantify how average ozone reacts to changes in rural NOX.
(The rural plumes would still be concentrated in a small area.)
-------
VI-37
Four caveats should be considered in connection with the
results summarized in Table VI-9:
1. Many of the results are based on averages of
different simulation results.
2. Our estimates represent how ozone is affected on
average across the U.S. They should not be used
for estimating ozone changes for a specific area.
3. Most of the literature reports estimates of changes
in peak ozone. Peak ozone is the appropriate measure
for most of our benefit categories, but for one
important benefit category (agriculture), changes in
daily maximum 7-hour average ozone levels are needed.
For rural emission changes, we assumed that the mea-
sures of average ozone move in direct proportion to
changes in peak ozone.
4. Some of our estimates of how rural emissions affect
ozone were derived from models better suited to
represent urban areas (e.g., EKMA).
Because of these qualifications, the estimates in Table VI-9
should be regarded as highly uncertain. The estimates for NOX
are particularly uncertain, because much less modeling has been
done of the NOx-ozone relationship than for HC.
We next calculated the population-weighted effects on ozone
of 1 percent changes in national emissions of HC and NOX. (We
used population as the weighting factor because the nonagricul-
tural benefits are proportional to it.) We calculated weighted
average elasticities using the estimates in Table VI-9 and the
fact that about two-thirds of the non-agricultural benefits are
attributed to metropolitan area and one-third to non-metropolitan
areas (based on relative populations). For hydrocarbons, the
estimates in Table VI-9 suggest that a 1 percent reduction in
emissions would reduce metropolitan ozone by 0.60 percent
(= 0.60 + 0); similarly, non-metropolitan ozone would fall by
-------
VI-38
0.41 percent (= 0.05 + 0.36). Weighting these two changes by
the split in non-agricultural benefits between metropolitan and
non-metropolitan areas yields an overall benefit-weighted elasticity
of:
(VI-5) 0.67 (0.60) + 0.33 (0.41) = 0.537
Multiplying this weighted elasticity times the national effects
of a 1 percent change in ozone yields the same result as
computing separately, and then summing, the metropolitan and
non-metropolitan effects of 1 percent changes in hydrocarbons.
Similarly, for NOX, the estimated benefit-weighted average
elasticity for non-agricultural benefits was:
(VI-6) 0.67 (-0.1 + 0) + 0.33 (0.04 + 0.15) = -0.004
Note that reductions in NOX are predicted, on average, to
cause a slight increase in peak ozone levels. Using the full
range of elasticity estimates in Table VI-9, however, this
weighted elasticity ranges from -0.22 to +0.066.
For agricultural crops, we assumed that the relevant measure
would be the change in average rural ozone levels; we gave no
weight to changes in urban ozone. For HC, the estimated elasticity
of average rural ozone with respect to national emissions was:
(VI-7) 0.13 + 0.36 = 0.49
Thus, we estimate that 1 percent reductions in metropolitan and
non-metropolitan HC emissions reduce average rural ozone by 0.49
percent. For the NOX point estimates, the elasticity was:
-------
VI-39
(VI-8) -0.09 + 0.15 = 0.06
Note that this estimate suggests that nationwide NOX reductions
will slightly decrease rural ozone levels, despite the fact that
urban NOX control, on average, is predicted to increase ozone;
the increase in urban ozone that is transported to rural areas
does not offset the reduction in ozone produced in rural areas.
Again, however, the range of uncertainty is large. Using the
ranges in Table V-9 generates elasticity estimates for the impact
of national NOX control on average rural ozone ranging from -0.04
to +0.11.
VI.B.2.C. Ozone-Related Effects Per Ton of HC and NOX Controlled
We used the estimated weighted average elasticities, together
with the estimates of reductions in HC and NOX emissions to predict
changes in ozone. We needed to predict changes in two ozone
measures: average rural ozone (for the agricultural benefits)
and changes in peak ozone levels, averaged over the nation (for
the other ozone benefit categories).
To apply the elasticities, we needed to convert the emission
reduction to percentages; because our two largest ozone-related
benefit categories (agriculture and restricted activity days) were
based on 1978 and 1979 air-quality data, we computed the changes
as percentages of 1979 emissions, which were about 21.9 million
tons for HC and 21.3 million tons for NOX. Thus, the predicted
emissions due to misfueling in 1986 (from Table VI-3) translate
to 1.4 percent (= 305,000/21.9 million tons x 100%) and 0.44
percent (= 94,000/21.3 million x 100%) for NOX. Using these
-------
VI-40
percentages together with the estimated elasticities for peak
ozone (equations VI-5 and VI-6) yields the following estimated
percentage change in peak ozone due to emission reductions in
1986:
(VI-9) (0.537)(1.4%) + (-0.004)(0.44%) = 0.75 percent.
Similarly, for average rural ozone, the estimated decline in
ozone due to emission reductions in 1986 is:
(VI-10) (0.49)(1.4%) + (0.60X0.44%) = 0.71 percent.
Table VI-10 presents the year-by-year estimates of changes in
ozone due to eliminating misfueling.
Finally, the predicted changes in the two ozone measures
were combined with the estimates in Table VI-8 to predict the
year-by-year ozone-related effects of the rule. The changes
in average rural ozone were used for agricultural crops, while
the changes in peak ozone were used for the other benefit
categories. Table VI-11 presents the results for 1986. It is
important to remember that these estimated effects are subject
to all of the uncertainties and omissions discussed earlier
in connection with our estimates of the benefits of a 1 percent
reduction in ozone. In addition, however, they also reflect
the uncertainties in the estimates relating HC and NOX emissions
to ozone; these uncertainties are particularly great in the
case of NOX.
-------
VI-41
TABLE VI-10. Estimated Ozone Reductions due to Eliminating
Misfueling (percent of 1979 level)
Ozone
Measure 1985 1986 1987 1988 1989 1990 1991 1992
National
peak 0.38 0.75 0.74 0.74 0.74 0.75 0.78 0.81
Rural
average 0.36 0.71 0.71 0.71 0.71 0.73 0.76 0.78
-------
VI-42
TABLE VI-11. Quantified Ozone-Related Effects Due to Elimination
of Misfueling in 1986
Effect
Health (millions of cases)
Respiratory
Bed rest/Lost work
Minor restrictions
Non respiratory
Other (millions of dollars)
Agricultural crops
Ornamental plants
Materials damage
LOW
Estimate
0.27
0.80
N.A.
63.80
N.A.
11.19
High
Estimate
0.92
2.69
N.A.
99.25
N.A.
16.79
Medium
Estimate
0.48
1.43
3.26
80.81
1.04
13.43
Unquantified Benefit Categories
Chronic health effects
Forest damage
-------
VI-43
VI.C. Health and Welfare Effects Not Related to Ozone
Thus far, our analysis has focused only on HC and NOX, and
only on their ozone-related effects. In this section, we discuss
other health and welfare effects associated with those pollutants
and with CO and EDB. In most cases, the discussion is only quali-
tative because we have been unable to make quantitative estimates.
VI.C.I. Hydrocarbons
In addition to the benefits associated with reduced ozone,
reductions in hydrocarbon emissions may affect sulfate concentra-
tions and will reduce exposure to benzene. Although quantitative
estimates are not possible for the former, we can make some rough
estimates of the health benefits of reduced benzene exposure.
VI.C.I.a. Impact on Sulfates
Hydrocarbons are a factor in the formation of sulfates.
In particular, S02 oxidizes faster when the amount of hydroxide
radicals in the atmosphere increases (which is, in turn, a function
of the amount of HC in the atmosphere). However, the ability to
quantify these complex relationships has just been developed, and
experts at SAI and EPA's Office of Research and Development believe
that the total change in sulfates is highly dependent upon many
factors (e.g., cloud cover, current HC and NOX concentrations, and
oxidant and sulfur dioxide levels) for which we have only limited
data.
A recent modeling analysis by SAI (1984) indicated that a
10 percent reduction in HC could reduce sulfates in urban areas
during certain times of the year by about 2 percent. However,
-------
VI-44
because of the uncertainty surrounding this estimate, and the
uncertainty in interpolating this to much smaller changes in HC,
we did not try to quantify the reduction in sulfates in this
analysis.
Vl.C.l.b. Impact on Benzene and Other Aromatics
Reducing lead in gasoline will affect benzene emissions in
two ways. First, because benzene is a hydrocarbon, reducing mis-
fueling will reduce benzene tailpipe emissions along with other
hydrocarbons. Second, to increase octane with less lead, refin-
eries will increase the severity of their catalytic reforming,
which in turn increases the amount of benzene in reformate, one
of the components blended into gasoline. Benzene emissions are
of particular concern because benzene is believed to be a leukem-
ogen, and has been listed by EPA as a hazardous air pollutant
under Section 112 of the Clean Air Act.
Benzene comprises about 4 percent of tailpipe hydrocarbon
emissions (U.S. EPA, 1983c). Assuming that catalysts are as
effective in eliminating benzene as other hydrocarbons, that
would imply that benzene would be roughly 4 percent of the reduced
HC emissions from eliminating misfueling. Thus, for example,
in 1986, we estimate that eliminating misfueling would eliminate
0.04 (305,000) = 12,200 tons of benzene.
Reducing lead is likely to increase the overall benzene
content of gasoline because additional reforming creates more
benzene in the reformate. Benzene (along with all other aroma-
tics) has a a poor octane response to lead, however, so as the
-------
VI-45
aromatic content of reformats increases it will be more economical
for refiners to increasingly direct such stocks to unleaded gaso-
line, and use less reformate and more of other higher lead-response
components (e.g., alkylate) in leaded gasoline. Based on the DOE
refining model, which finds the blending pattern that gives the
least cost, we estimated that the benzene content of gasoline will
rise from about 1.5 percent to 1.6 percent in unleaded, but will
fall from 0.62 percent to 0.38 percent in leaded gasoline (at
0.10 gplg).
Raising the overall benzene content of the gasoline pool will
increase evaporative emissions of benzene from gasoline marketing
facilities (service stations, etc.), but based on EPA analyses of
the impacts of regulating gasoline marketing (U.S. EPA, 1984a), we
estimated the increase to be less than 400 tons per year. Moreover,
because of the shift of benzene from leaded to unleaded gasoline,
evaporative emissions from vehicles should fall by about the
same amount, because unleaded gasoline is used in vehicles with
evaporative control devices. In addition, switching benzene
from leaded to unleaded should reduce tailpipe emissions of
benzene because unleaded gasoline is used in vehicles with cata-
lytic converters, which oxidize about 90 percent of the unburned
hydrocarbons in the exhaust. In 1986, this should reduce benzene
tailpipe emissions by almost 7,000 tons.
Table VI-12 summarizes our estimates of reductions in
emissions of benzene for 1986, assuming that misfueling is elimi-
nated. They total about 18,000 tons, of which about two-thirds
are due to the elimination of misfueling and one-third are due
-------
VI-46
TABLE VI-12. Reductions in Benzene Emissions in 1986, Assuming
No Misfueling (tons)
Sources of Emissions Amount
Reduced Misfueling
Tailpipe emissions 12,200
Changes in fraction of benzene
in leaded and unleaded gasoline
Marketing evaporative -366
Vehicle evaporative 377
Tailpipe 6,230
TOTAL 18,441
-------
VI-47
to reduced tailpipe emissions resulting from the change in the
relative benzene contents of leaded and unleaded gasoline.
Evaporative emissions account, on net, for less than 0.1 percent
of the change.
EPA1s Carcinogen Assessment Group (CAG), using a "plausible
upper bound" unit-risk estimate, estimated that in 1976, automobile
emissions of benzene resulted in up to 50.9 deaths from leukemia
(U.S. EPA, 1979a). Since then, however, the CAG has revised its
unit-risk estimate for benzene downwards by slightly more than 8
percent, yielding an estimate of 47.3 leukemia cases for 1976.
The CAG estimate was based on an emissions estimate of 202,000 tons
of benzene. Scaling down the CAG risk estimate of 47.3 cases by
emissions yields 4.4 (= 47.3 x 18,441/202,000) cases of leukemia
eliminated in 198fi.
We also have performed similar calculations for emissions of
total aromatics, with similar results. Reducing lead in gasoline
should reduce overall emissions of aromatics, both because of
reduced misfueling and because of shifts in the relative fractions
of aromatics in leaded and unleaded gasolines. On net, emissions
of aromatics should fall by several tens of thousands of tons per
year (Schwartz, 1984, in Docket EN-84-05) as a result of reducing
lead in gasoline to 0.10 gplg. We have not included any benefits
for these reductions, however, except as they relate to benzene
specifically.
VI.C.2. Nitrogen Oxides
Besides the ozone-related effects discussed above, nitrogen
oxides also may affect health and welfare directly. NOX emissions
-------
VI-48
are believed to affect health and materials, to contribute to
reductions in visibility, and are associated with acid deposition.
In addition, damage to vegetation has been demonstrated experiment-
ally. Unfortunately, specific dose-response information related
to NOX is sparse. As a consequence, there is great uncertainty in
the benefit estimates and many effects cannot be quantified. How-
ever, a few of the benefit calculations are presented to provide
a partial measure of the effects of NOX emissions on economic
welfare. For health effects, only qualitative evidence is provided.
While there may be acid rain benefits as well, we have not included
them because of uncertainties about the role of NOX in acid
deposition.
As discussed earlier, NOX represents the composite formula for
NO (nitric oxide) and NC>2 (nitrogen dioxide). NO is the dominant
oxide released initially, but atmospheric interactions result in
the conversion of NO to NO2- Based on the results of smog chamber
tests and modeling experiments, Trijonis (1978, 1979) concluded
that maximum and average NO2 concentrations tend to be proportional
to initial NOX contributions.
VI.C.2.a. Visibility Benefits from Reduced NOV
NO2 is a reddish-brown gas that reduces visibility by
absorbing and discoloring light. In contrast, particulate
matter scatters light to reduce visual range. While particulate
matter accounts for almost all of the damage to visibility in
the East, for some western regions NO2 may play a significant
role in determining visual range.
-------
VI-49
To bound the value of improved visibility per ton of NOX
reduced, we used the results of Brookshire et al. (1976). Their
contingent valuation study showed that recreators were willing to
pay $1.2 million per year (annualized) to avoid visibility re-
ductions that would result from a planned Kaiparowits power plant
in southern Utah. Similar studies, by Randall et al. (1974) and
by Blank et al. (1977) produced comparable results for the
value of visibility benefits in the Four Corners region of the
Southwest. However, these estimates apply only to this particular
region, known for its scenic vistas.
In the Kaiparowits study, climatic conditions, emission
controls, and other factors allowed the investigators to assume
that the major visibility-related impact would be the coloration
of the sky by NO2- Given the projected power plant emissions of
80,000 tons of NOX per year, the estimate of $1.2 million of
potential damage translates to about $2.1 million in 1983 dollars
(based on Consumer Price index). For 1986, we project a 94,000
ton reduction in NOX emissions if misfueling is eliminated. Using
the Brookshire et al. estimate and assuming linearity yields a
benefit estimte of about $2.4 million. We stress that this
figure is probably an upper bound, even for sensitive regions.
Although the Kaiparowits area is not densely populated, a large
number of recreators use the site and they typically place a
high value on protecting clean areas. Further, it is doubtful
that NC>2 has a noticeable impact on eastern visibility, where
-------
VI-50
range is limited by buildings and largely influenced by particulate
matter (based on a conversation with Shep Burton of Systems
Applications, Inc.).
VI.C.2.b. Health Benefits of Reducing NOy
Evidence of health effects related to NOX is provided in the
Air Quality Criteria for Oxides of Nitrogen (EPA, 1982b), the OAQPS
staff paper on the NAAQS for nitrogen oxides, the Clean Air
Science Advisory Committee's (CASAC's) cover letter (EPA, 1982c),
the NOX Regulatory Impact Analysis (EPA, 1982e), and the published
literature. These studies can be divided into four classes:
(1) Animal toxicology studies. Animals are exposed to
controlled levels of NC>2. Researchers have the option
of using invasive techniques to investigate the effects
of N02.
(2) Controlled human-exposure studies. These are clinical
studies in which humans are exposed to NO2 in enclosed
chambers. They are limited typically to examining the
effects of a single, short-term (acute) exposure.
(3) Outdoor epidemiological studies. Health indicators
of cross-sectional groups are statistically related
to real-world outdoor ambient concentrations. This
class of studies is generally most appropriate to
assess the benefits of controlling outdoor air pollu-
tion, since health effects are related directly to
the control variable of interest.
(4) Indoor epidemiological studies. Health measures of
cross-sectional groups are statistically related to
indicators of indoor pollutant concentrations. For
example, the "gas stove" studies investigate the
effect of indoor air pollution on individuals living
in homes with gas stoves (a significant source of NC«2)f
compared with people living in homes with electric
stoves.
Most animal toxicology studies involving NC>2 emphasize peak
exposure, with concentrations (2 ppm to 20 ppm) roughly 40 to 400
times the annual average ambient N02 standard of 0.05 ppm. An
-------
VI-51
additional limitation of these studies in assessing benefits is
that there is no generally accepted method for extrapolating expo-
sure response results from animal studies to humans. Thus, these
studies may lend support to the human studies, but are of little
help in quantifying health effects.
The majority of the controlled human experiments have
examined the effects of NC>2 on healthy adults by exposing them
to single, short-term concentrations in enclosed chambers. A
very limited number of clinical studies also have examined some
potentially sensitive populations (asthmatics and chronic bron-
chitics), although others (children) have yet to be tested.
In general, these studies, as summarized in the OAQPS staff
paper, indicate that healthy adults are not affected by concen-
trations of 1 ppm or less. When considering the effect on more
sensitive individuals, small reductions in pulmonary function may
appear in the range of 0.5 to 1.5 ppm. Also, some studies have
shown increased sensitivity to agents inducing bronchoconstriction
at 0.1 to 0.2 ppm levels of N02- Unfortunately, dose-response
functions relating NOX to economically-significant health endpoints
do not exist.
Successful outdoor epidemiological studies are scarce because
investigators must separate many confounding effects and health
hazards. Moreover, some of the studies used in setting the exist-
ing annual standard of 0.053 ppm (100 ug/m3) have been criticized
because of NC>2 measurement problems. Therefore, this class of
studies does not provide any adequate basis for estimating a
dose-response function.
-------
VI-52
Indoor epidemiological studies have focused on comparisons
of homes with gas stoves (which emit NOX) to those with electric
stoves. Some of these studies reported an association between
nitrogen dioxide and either lung function changes or respiratory
effects (EPA, 1982), but the most recent studies, including Ware
et al. (1984), no longer find a statistically significant rela-
tionship between children living in homes with gas stoves and
the incidence of respiratory illness. Ware et al., however,
continue to find small statistically-significant decreases in
in pulmonary function associated with NOX, although the estimates
may be confounded by the influence of parental education.
Because of the mixed results concerning respiratory symptoms
and the lack of a dose-response estimate for pulmonary function
changes, no health benefits are quantified for the anticipated
reductions in NOX emissions.
VI.C.2.C. NOy Effects on Vegetation
Data concerning the effects of NOX on plant growth and yield
also are limited. Nevertheless, it it reasonable to assume that
NOx-induced reductions in the assimilative capacity of plants
through altered metabolism, leaf injury, or abscission affect
plant growth. MacLean (1975), however, concluded that average
NC>2 concentrations are well below the threshold curve for damage
to growth. In fact, the maximum NC>2 concentrations recorded in
Los Angeles for 1966 would just begin to damage growth. Even in
Los Angeles, however (the only major city in the U.S. that exceeds
the NO2-NAAQS), average NO2 concentrations are below the likely
threshold when averaged over longer periods.
-------
VI-53
Although NC>2 by itself is unlikely to damage plants at
existing outdoor levels, several studies have demonstrated syner-
gistic and antagonistic effects. The Criteria Document (EPA,
1982) concluded that "concentrations of N02 between 0.1 ppm and
0.25 ppm can cause direct effects on vegetation in combination
with certain other pollutants" (pp. 12-44). But these data
indicate that rural areas are still not at risk, because, in
general, concentrations will not rise above 0.1 ppm for suffi-
cient time for damages to occur. Therefore, we limit our estimate
of benefits to account for vegetation damages in urban areas.
The plants most in danger would be ornamental vegetation,
but even these damages are likely to be small. Leighton et al.
(1983) concluded from their review that "ozone appears to account
for more than 90 percent of total vegetation damages" (p. 60;
see also Heck et al., 1982, and Page et al., 1982). If NOX
accounted for the other 10 percent of the total ornamental
vegetation damage, as an upper-bound we could use the high estimate
of ozone ornamental vegetation damages of $140 million to impute
total NOX damages of roughly $16 million. For the 0.44 percent
reduction in NOX predicted for 1986, that would imply benefits
of only $62,000.
VI.C.2.d. NOy Effects on Materials
Field studies and laboratory research have demonstrated
that nitrogen oxides can significantly fade textile dyes.
Barrett and Waddell (1973) estimated damages at $280 million
(inflated to 1983 dollars using the Textile Products and
-------
VI-54
Apparel Index) for NOX damage in this category. The basis for
the estimates included not only the reduced wear life of textiles
of moderate fastness due to NOX, but also the costs of research
and quality control. The major share of the costs is the extra
expense of using dyes of higher NOX resistance and of using
inhibitors. Additional costs are also incurred for dye application
and increased labor expenditures. The factors relating higher
costs in the textile industry to NOX are discussed in Chapter 8
of a report of the National Academy of Sciences (1976).
For 1986, assuming linearity, the Barret and Waddell
estimate implies that materials benefits related to NOX control
would be about $1.2 million (0.44% x $280 million).
VI.C.2.e. Acid Deposition Benefits
Acid deposition occurs when NOX and S02 emissions are
chemically altered into acids in the atmosphere and transported
over long distances, or when the precursor emissions are acidified
after being deposited in dry form on plant, soil, or building sur-
faces. Both wet and dry forms of acidic deposition are harmful to
aquatic, terrestrial, and material resources. While these damages
are potentially important in some regions, basic scientific under-
standing of the effects on resources and transport processes is
quite limited.
The computation of damages requires several crucial pieces
of information. Unfortunately, many of these data are so uncer-
tain that confidence intervals of damage estimates are exception-
ally wide. Dollar estimates of damage require concentration-
response functions and a tabulation of the resources at risk.
-------
VI-55
Reliable concentration-response data are lacking in nearly all
resource areas, especially forestry and materials, and inventory
data are unavailable for most material resources. Finally, the
benefit estimate requires information on the relative contribution
of nitrates and sulfates to total damages, as well as the rela-
tionship between the tonnage of NOX emissions and nitrate depo-
sition. All of these are the subjects of intensive ongoing
research. Until more of this research is completed, quantitative
damage estimates will not be accurate enough to be useful in a
policy context.
Damage from acid deposition may be significant, although
there is much uncertainty. Nitrates account for roughly 30
percent of total acidic loadings, but the damage information
available indicates that the proportion of nitrate to total
damages may be less than that implied by the relative emittant
loadings.
VI. C.3. Carbon Monoxide
At current ambient concentrations, exposure to CO may cause
health effects in some individuals. Persons with cardiovascular
disease appear to be at highest risk, but those with chronic
respiratory disease, pregnant women, and the elderly also are
believed to be sensitive to CO exposure. Unfortunately, clinical
dose-response functions relating low-level CO exposure to parti-
cular health effects, when estimated, have not been conclusive.
Therefore, it is not possible to estimate the impact of reduced
CO emissions on health endpoints. We have, however, been able
-------
VI-56
to estimate roughly the impact of reducing CO emissions on the
numbers of people with carboxyhemoglobin (COHb) levels that may
pose some risk.
VI.C.3.a. Health Effects of CO
Probably the greatest concern about CO exposure is its effect
on the cardiovascular system. At moderate levels of exposure, CO
reduces exercise time before the onset of angina pectoris. This
clinical phenomenon is believed to result from insufficient oxygen
supply to the heart muscle, and is characterized by spasmodic chest
pain, usually precipitated by increased activity or stress, and is
relieved by rest. Typically, atherosclerosis, which causes a
narrowing of the arteries in the heart (coronary heart disease),
predisposes a person to attacks of angina.
Angina pectoris, by definition, is not associated with
permanent anatomical damage to the heart. Nonetheless, the dis-
comfort and pain of angina can be severe, and each episode of
angina may carry some risk of a myocardial infarction. However,
epidemiological studies have not provided conclusive results on
the association between CO exposure and the incidence of myocard-
ial infarction.
The health effects from exposure to CO are associated with
the percentage of total blood hemoglobin that is bound with CO,
producing COHb, which reduces the oxygen-carrying capacity of the
blood. The median concentrations of COHb in blood are about 0.7
percent for nonsmokers and about 4 percent for smokers. At least
one clinical study (Anderson et al., 1973) associated reduced
-------
VI-57
exercise time until the onset of pain with COHb levels of 2.9
percent in patients with angina pectoris. At 4.5 percent COHb,
this same study reported an increased duration of angina attacks.
The potential health improvements from reduced CO may be
significant, for two reasons. First, there are many people in
the population believed to be sensitive. EPA has estimated that
5 percent of the U.S. adult population — roughly 11.5 million
people — has definite or suspected coronary heart disease. Of
this group, as many as 80 percent have suspected or definite
angina pectoris (U.S. EPA, 1980). Additional large subgroups of
the population may be particularly sensitive to exposure to CO,
including individuals with pre-existing conditions that
compromise oxygen delivery to various tissues, enhance oxygen
need, or elevate the sensitivity of tissues to any oxygen
imbalance. Other sensitive groups may include:
0 people with peripheral vascular diseases, such
as atherosclerosis and intermittent claudication
(0.7 million people);
0 people with chronic obstructive pulmonary
diseases (17 million people);
" people with anemia or abnormal hemoglobin types
that affect the oxygen-carrying capacity of the
blood (0.1245 million people);
0 people drinking alcohol or taking certain
medications (e.g., vasoconstrictors);
0 the elderly;
0 residents of and visitors to high altitude areas;
and
-------
VI-58
0 fetuses and infants (3.7 million total live births
per year).*
The second reason for potentially significant impacts of
reduced CO exposure is that the blood of many people shows
concentrations of COHb above 2.9 percent, the lowest level of
COHb where adverse effects are believed to occur. Data from the
second National Health and Nutrition Examination Survey (NHANES
II) indicated that for the U.S. population over twelve years of
age, 2 percent of those who have never smoked, 3 percent of
former smokers, and 66 percent of current smokers exceeded 2.9
percent COHb at the time of the survey (U.S. DHHS, 1982).
Other health effects have been reported at comparable or
higher COHb levels. For example, several investigators have
found statistically significant decreases in work time until
exhaustion in healthy young men with COHb levels at 2.3 to 4.3
percent (Horvath et al., 1975; Drinkwater et al., 1974; Raven
et al., 1974). At higher COHb levels (5.0 to 7.6 percent and
above), investigators have reported impairment in visual percep-
tion, manual dexterity, ability to learn, and performance of
complex sensorimotor tasks in healthy subjects. Finally, Klein
et al. (1980) showed that, at 5.0 to 5.5 percent COHb, healthy
young men had decreased maximal oxygen consumption and decreased
time at strenuous exercise before exhaustion.
* Animal studies showed that pregnant females exposed to CO
had lower birth weights and increased newborn mortality, and
their newborns had lower behavioral levels, even when no
effects on the mother were detected. In addition, a possible
association has been reported between elevated CO levels and
Sudden Infant Death Syndrome (Hoppenbrouwers et al., 1981.)
-------
VI-59
Vl.C.S.b. Change in Numbers of People Above 2.9 Percent COHb
For three specific subgroups of the U.S. population —
current smokers, ex-smokers, and never-smokers — we calculated
the number of people who would move below 2.9 percent COHb (the
lowest level frequently associated with specific health effects)
because of the reduction in CO emissions. We also calculated the
number of "sensitive" individuals who would shift below 2.9
percent COHb. This 2.9 percent level should not be construed as
a "threshold" level, but rather as a level at which some increased
health risk has been detected.
We considered several complex relationships to make these
rough estimates. Because of resource and data limitations, we
assumed simple one-to-one relationships in some of the linkages
between changes in CO emissions and the ultimate changes in COHb
levels. Specifically, we assumed that a given percent change in
CO emissions would generate a similar percent change in ambient
CO. We also assumed a linear relationship between changes in
ambient CO and the mean percent COHb level in blood. For small
changes, these assumptions may be fairly accurate. For larger
changes in CO emissions, there is greater uncertainty concerning
the impact on COHb levels. For example, it is well known that
the binding affinity of hemoglobin may be nonlinearly related to
the level and change in CO. Mage et al. (1984), however, found
a very high correlation (R^ = 0.967) between maximum COHb levels
and five-hour averages of CO concentrations, which suggests
that our assumption of linearity is a good approximation.
-------
VI-60
In estimating the number of people whose COHb levels would
exceed 2.9 percent, we assumed that the geometric standard devia-
tion of the distribution of COHb levels was roughly constant for
small changes in mean COHb levels; this implies that changes in
CO emissions will shift the whole distribution proportionately.
Subject to these simplifying assumptions, the results are indi-
cative of the potential magnitude of the effect of reduced CO
emissions on COHb levels.
To estimate the reduction in COHb levels, we had to (1)
calculate the average annual CO emissions from auto and resident-
ial sources* for 1976-1980; (2) project 1988 emissions from these
two sources; (3) determine the shape (mean and standard deviation)
of the frequency distribution of COHb levels for 1976-1980 from
the NHANES II for current smokers, former smokers, and never-
smokers; (4) adjust the mean of this distribution for 1988, based
on the changes in CO emissions; (5) calculate the change in mean
COHb levels with and without misfueling; and (6) multiply the
change in the probability distribution by the appropriate popula-
tion group to predict the number of people now expected to be
below 2.9 percent COHb.
The average annual CO emissions from automobiles and
residences during 1976 to 1980 were 73.3 and 3.9 million tons,
respectively, for a total of 77.2 million tons (U.S. EPA, 1982f).
* We focused on auto emissions and residential sources (gas
ovens, heating, etc.) because they have been identified as
the most serious sources of human exposure. Cigarette
smoking patterns and rates are assumed to be unaffected by
any misfueling policy.
-------
VI-61
The estimate of auto emissions for 1988 is 36.5 million tons.*
To estimate residential CO emissions, we determined the relation-
ship of residential CO emissions over time (U.S. EPA, 1982f),
and projected that relationship into the future. Based on this
analysis, residential CO emissions would be 8.3 million tons in
1988. Thus, CO emissions from these two sources were estimated
to drop to 44.8 million tons in 1988, a 42 percent decrease.
Data on blood carboxyhemoglobin levels were collected as part
of the NHANES II study. Because of the seasonal and geographic
pattern of the survey during the four-year period, it may not
reflect the full range of potential exposure. Specifically, the
Northeast was not sampled during the winter, and no high-altitude
cities were sampled. Therefore, the survey results may somewhat
understate total CO exposure.
For each of the subgroups — current smokers, ex-smokers, and
never-smokers — we fit the observed COHb levels with a lognormal
distribution. We took the log of each observation in the NHANES
and calculated the mean and the standard deviation of log(COHb)
The baseline projections for CO for 1988 were calculated as
follows:
For CO we started with EPA emission factors generated in the
draft model of MOBILE III for on-road vehicles in 1988 of
approximately 26.65 g/mi. The emission factor was reduced by
0.75 (to 20 g/mi) to adjust for I/M areas and the state of
California, which has its own, more stringent, emission con-
trols. We assumed 159.6 million on-road vehicles traveling an
average 11,436 miles (see Appendix B). Multiplying:
20 g/mi x 11,436 mi/vehicle x 159.6 x 106 vehicles
1 x 106 g/metric ton
36.5 x 106 metric tons
-------
VI-62
for each of the three groups. The mean COHb level (M) is then:
M = exp (u + s2/2),
where u is the mean of the lognormal distribution and s2 is the
variance. To estimate the 1988 COHb baseline, we reduced the
mean COHb level by the predicted change in COHb (P) over time.
We calculated the difference, log(M) - log(M-P), for each sub-
group. This difference was then subtracted from the log(COHb)
for each person in the subgroup to obtain a new distribution.
This procedure shifts the distribution down while keeping the
geometric standard deviation constant. Thus, it assumes
constant proportional changes, rather than equal absolute
reductions, in COHb levels across subgroups.
To calculate P, we needed to determine the impact of the
reduction in CO emissions from 1978 to 1988 on mean COHb levels.
We assumed a linear relationship between ambient CO and percent
COHb, with a slope coefficient of 0.16 (adapted from Ferris,
1978; U.S. EPA, 1979a; and Joumard et al., 1981). We used an
average ambient CO level of 3.127 ppm for 1978 (Council on
Environmental Quality, 1980). Thus, the projected 42 percent
change in ambient CO between 1978 and 1988 would reduce mean
percent COHb levels, on average, by 0.210 percentage points
(0.42 x 3.117 x 0.16 = 0.210).
Given the 1988 baseline distribution of COHb levels, we
determined the effect of eliminating all misfueling and reduced
-------
VI-63
the mean by 5 percent, the estimated CO reduction.* We used the
same procedure of converting to a log change, reducing the COHb
level of each population subgroup by that amount, and re-averaging.
From this new distribution, we could then calculate the change
in the number of each subgroup under 2.9 percent COHb due to
this reduction in the mean. Since the NHANES II data are weighted
to represent the population of the United States, the weighted
average gives us the change in the total number of people in
each group who would have COHb levels above 2.9 percent. The
results indicated that over 400,000 people would shift below
the 2.9 percent COHb level if we could eliminate misfueling:
112,000 current smokers; 62,000 exsmokers; and 227,000 never-
smokers.
Another indication of the impact of the reduction in CO is
generated by considering the reduction in COHb levels for some of
the sensitive subgroups identified above. This includes those
with suspected and definite angine (9.2 million), peripheral
vascular disease (0.7 million), chronic obstructive pulmonary
disease (17 million), anemia (0.1245 million), and fetuses (3.7
million). Totalling our projections for 1988, we estimated that
there will be 31 million sensitive individuals. In addition, we
included a subset of the elderly, who may have an increased risk
"* The actual change in CO was calculated using the estimates of
avoided emissions in Table VI-3 and projections of total emis-
sions in 1988. Assuming motor vehicles emit 86 percent of all
CO from transportation and residential fuel combustion, the
reduction of 2.2 million metric tons is 5 percent (2.2 x 0.86/
36.5) of the total.
-------
VI-64
but who may not have been counted within these other groups. As
an approximation, we added 25 percent of those age 65 and above,
to obtain a total population at risk of 39 million people.
Since we did not have data on the COHb distribution for these
sensitive subgroups (including the elderly and those with coronary
heart disease), we assumed their COHb distribution was similar
to that of never-smokers. This assumption was made because some
in this sensitive group probably take measures to reduce the
impact of air pollution and other irritants on their health. How-
ever, it probably generates a low estimate because many in this
group continue to smoke, and others are former smokers. Neverthe-
less, we applied the previously determined change in the portion
of never-smokers who would shift below 2.9 percent COHb because
of the reduction in misfueling (0.000865) to the group considered
sensitive. It suggested that 33,700 sensitive people will shift
below the 2.9 percent COHb level in 1988.
VI.C.4. Ethylene Dibromide Emissions
Most of the ethylene dibromide (EDB) manufactured in the
United States is added to leaded gasoline as a scavenger for the
lead. Reducing the lead concentration of gasoline would result
in an equal reduction of EDB use, which is of concern to EPA
because it is a potential human carcinogen. EDB from leaded gas-
oline enters the air through three routes: tailpipe emissions,
evaporative emissions from cars, and evaporative emissions from
the retail and distribution chain of gasoline.
-------
VI-65
Sigsby et al. (1982) estimated EDB emissions from tailpipes
and evaporative emissions from cars. They concluded that EDB
emissions at the tailpipe were approximately 0.37 percent of the
amount in gasoline. However, all of the emissions tests were
done on retuned and adjusted cars, and under somewhat artificial
test procedures; the 0.37 percent survival was the average of
the tests. The EPA federal test procedure showed an average EDB
survival of 0.69 percent. Since EPA test of on-the-road vehicles
have generally shown substantially higher actual emissions than
in retuned cars undergoing the federal test procedure, we used
the 0.69 percent survival factor in this analysis. Multiplying
this by the projected leaded gasoline demand in 1986 yields an
estimated reduction in EDB tailpipe emissions of 143 metric tons.
Shed tests of the evaporative emissions of EDB generally
indicate that the EDB evaporative emissions were 1/20,000 of the
hydrocarbon evaporative emissions (Sigsby et al., 1982). In the
absence of better data, we assumed the same ratio of refueling
evaporative emissions. This results in an estimated 34 additional
metric tons of EDB emissions, based on our fleet model.
Finally, EPA estimated (EPA, 1984a) that hydrocarbon emissions
in the retail and distribution chain of gasoline were 407,000 metric
tons in 1982. Based on our estimate of 37.5 percent leaded
gasoline demand in 1986, we estimated that leaded gasoline would
produce 153,000 metric tons of hydrocarbon emissions. Using the
ratio of EDB to hydrocarbon emissions, this suggests 7.6 metric
tons of EDB emissions from the retail and distribution stages
that would be reduced if all misfueling was eliminated. In
-------
VI-66
total, we estimate 185 metric tons of EDB emissions from all
three sources (tailpipe, evaporative, and marketing emissions)
would be avoided by reducing lead in gasoline in 1986.
In addition, EDB from gasoline storage tanks has been leaking
into underground aquifers. Data do not exist, however, to esti-
mate the magnitude of the benefit of reducing the contamination of
ground water by EDB.
VI.D. Monetized Benefit Estimates
To estimate the monetized benefits of reducing emissions of
pollutants other than lead, we used two different methods. The
first used the direct estimates of health and welfare effects,
combining the nonhealth estimates with monetized estimates of the
health effects. As discussed earlier, that method is conceptully
correct, but suffers from substantial uncertainties and from
our inability to quantify some potential significant benefit
categories. Our second method uses the cost of pollution
control equipment destroyed by misfueling as a proxy for the
benefits of eliminating misfueling.
VI.D.I. Value of Quantified Health and Welfare Benefits
Our direct approach yields monetary estimates for only a
limited number of benefit categories, most of them related to
ozone. The ozone-related benefits fall into two major categories:
health-related and other. The second category already has been
estimated in monetary terms (Table VI-11) for 1986. The ozone-
related health effects, however, thus far have been stated only
in physical terms.
-------
VI-67
For the more serious category of respiratory conditions
(days of bed rest or work loss), we used the average daily
wage ($80) as a lower bound. This is the value of the lost
output to society for an employed individual. Because some
of these conditions also involve additional medical expense,
we used an upper bound of $120, giving a mean value of $100
per day.
To value a day of more minor restrictions in activity due
to respiratory conditions, we relied upon Loehman et al. (1979),
whose survey results suggested a willingness to pay of $2.31 to
prevent a day of minor coughing, $4.90 to prevent minor shortness
of breath, and $8.17 to prevent minor head congestion. We con-
verted these estimates from 1978 dollars to 1983 dollars to yield
values of $3.50 to $12.50 for avoiding a minor restricted day,
with a point estimate of $8.
We were unable to find estimates in the literature for the
value of avoiding headaches or eye irritation. These conditions,
however, seemed less serious than the respiratory effects dis-
«
cussed above, so we used a value of $3 per case, just below the
lower end of the range from the Loehman et al. study.
In addition to the ozone-related benefits, we were able to
place rough monetary estimates on three additional benefit
categories: (1) reduction in leukemia cases associated with
benzene exposure; (2) NOx-related visibility benefits; and (3)
N0x-related materials damages. The monetary estimates for the
last two categories already have been discussed. For the
-------
VI-68
leukemia cases, we note that the types of leukemia associated
with benzene exposure are almost invariably fatal. As discussed
in the previous chapter, valuing reductions in risks to life is
controversial, with a wide range of values found in the litera-
ture. For our high and low estimates, we used the range in the
EPA guidelines, $400,000 to $7 million per statistical life
saved. For our point estimate, we used the same value employed
in the previous chapter for blood-pressure-related fatalities,
$1 milliom per case.
Table VI-13 summarizes the monetized benefit estimates for
1986. The total covers a broad range, from $113 million to $305
million, with a "medium" estimate of $171 million. For the
reasons discussed earlier, plus uncertainties about the valuation
of health effects, even this broad range does not capture the
full extent of the uncertainty with respect to the categories
estimated. Moreover, it has not been possible to monetize many
of the possible benefits; the bottom of Table VI-13 presents a
partial list of those omitted categories.
VI.D.2. Implicit Value Based on Cost of Control Equipment
Our second method of valuing reduced emissions of HC,
NOX, and CO is based on the cost of pollution control equipment
destroyed by misfueling. This method assumes, implicitly, that
the benefits of controlling mobile-source emissions are at least
equal to the cost of the equipment needed to meet emission
-------
VI-69
TABLE VI-13. Monetized Benefit Estimates Due to Elimination
of Misfueling in 1986 (millions of dollars)
Low High Medium
Effect Estimate Estimate Estimate
Quantified Benefitsf Ozone-Related
Health
Respiratory
Bed rest/Lost work 21.5 110.1 47.8
Minor restrictions 2.8 33.6 11.4
Nonrespiratory 9.8 9.8 9.8
Other
Agricultural crops 63.8 99.2 80.8
Ornamental plants 1.0 1.0 1.0
Materials damage 11.2 16.8 13.4
Quantified Benefitsf Not Ozone-Related
Leukemia (benzene) 1.8 30.8 4.4
Visibility (NOX) 0.0 2.4 1.2
Materials damage (NOX) 1»2 1.2 1.2
TOTAL MONETIZED BENEFITS 113.1 305.0 171.0
Unquantified Benefit Categories
Chronic health effects due to ozone
Forest damage due to ozone
Direct health effects of HCs other than benzene
Sulfate-related damages due to HC
Acid-precipitation damages due to NOX
Health effects due to NOX
Vegetation damages due to NOX
Health effects due to CO
Health effects due to EDB
-------
VI-70
standards. Emissions control equipment costs about $283 per
vehicle (U.S. EPA, 1981).*
One complication for this method of valuing emissions is
that many catalysts are not destroyed by misfueling until they
have been in use for several years. Thus, it was necessary to
prorate the cost of the pollution control devices. To do so, we
first estimated the tons of emissions that would be controlled
over the life of the pollution control device. These estimates
accounted for declines in the efficiency of the devices over
time, declines in annual miles driven as vehicles age, and scrap-
page rates. We then discounted the emissions controlled back to
the first year (to make them comparable to the cost estimate), and
divided them into the S283 cost of pollution control equipment.
That yielded a cost of $153 per ton controlled.** Note that this
*The costs are "retail price equivalents," which are 30 percent
to 50 percent of the manufacturers' suggested retail price of
the components (catalysts, oxygen sensors, etc.) as replacement
parts. There may be a small upward bias in this estimate, but
we used the lower estimate in the cited report ($250, compared
to a $425 upper bound). Converting to 1983 dollars gave a
cost of $283. (About half of the cost of oxygen sensors and
other equipment was allocated to fuel efficiency, not pollution
control.)
**Using 1981 emissions standards of 0.41 grams per mile for
HC, 3.4 g/mi for CO, and 1.0 g/mi for NOX gave us 4.81 g/mi
for all pollutants in each future year. We multiplied 4.81
g/mi by E/(1-E) (where E equals the catalytic converter
efficiency in that year) and by 10,000 mi/yr, and divided by
1000 g/Kg to get kilograms controlled in each year by one
catalytic converter. We then discounted the estimate for
each year back to the first year of the catalytic converter's
life. Summing these present values gave us an estimate of
1.848 tons controlled by each car's catalytic converter over
a 10-year life. A cost of control equipment of $283, divided
by 1.848 tons, gave us $153 per ton.
-------
VI-71
estimate covers the sum of HC, NOX, and CO emissions; because
the devices control all three pollutants simultaneously, it was
not possible to generate pollutant-by-pollutant estimates using
this method.
We then calculated year-by-year benefits by multiplying
that estimate of $153 per ton of pollutant controlled times the
projected reduction in total emissions that would be achieved in
each year if misfueling were totally eliminated. For example,
as shown in Table VI-3, we estimate that eliminating misfueling
in 1986 would reduce the present value of emissions of the three
pollutants by 2.515 million tons. Thus, our estimate by this
method of the benefits for 1986 is $385 million ($153 x 2.515
million tons).
VI.D.3. Summary of Benefits of Controlling Pollutants
other than Lead
In this chapter, we have developed two approaches to valuing
the reductions in emissions that would be achieved by eliminating
misfueling. The first method, based on direct estimates of
health and welfare benefits, yielded an estimate of $171 million
for 1986. To generate estimates for other years, we scaled the
1986 estimates in Table V-13, with the scaling factor depending
on the benefit category. For the ozone-related agricultural
benefits, we scaled the estimates in proportion to the estimated
change in average rural ozone levels (from Table VI-10). For
the other ozone-related benefits, we used the predicted change
in peak ozone levels (also from VI-10). For the benzene-related
-------
VI-72
leukemia cases, we scaled by hydrocarbon emissions (from Table
VI-3), because most of the reductions in benzene result from
reduced emissions of hydrocarbons from misfueled vehicles.
Finally, the two NOx-related categories, visibility and materials
damage, were scaled by NOX emission reductions (also from Table
VI-3). This scaling procedure assumed that benefits are propor-
tional to emission reductions over the relevant ranges. The
results are shown in the first line of Table VI-14.
The caveats discussed earlier with respect to the direct
estimates for 1986 apply to these estimates as well. They
include not only the significant uncertainties associated with
the estimated categories, but also the omission of several
benefit categories due to limited data. For ozone, the
omissions include chronic health effects and forest damage.
For HC we have not estimated any direct health effects for
hydrocarbons other than benzene, nor have we quantified the
link between hydrocarbons and sulfate formation. For NOX, we
have not quantified benefits related to acid precipitation or
to vegetation, nor have we quantified any healt effects. We
also have been unable to generate monetized benefit estimates
for CO, although we have made made some rough estimates of the
numbers of people whose COHb concentrations may fall below a
potential harmful exposure to EDB. As a result of all of these
omissions, it is likely that the direct estimates are too low,
perhaps by a substantial margin.
-------
VI-73
The second method, based on the implicit cost per ton of
pollutants controlled by catalysts, yielded a value of $153 per
ton for the sum of HC, NOX, and CO emissions. The second line of
Table VI-14 shows the estimates based on that method. The last
line shows the averages of two methods; the summary estimates in
Chapter VIII are based on those averages. In that chapter, we
also estimate the benefits if misfueling is only partially
eliminated by the rule.
-------
VI-74
TABLE VI-14. Year-by-Year Monetized Estimates of Benefits of
Reduced Emissions of Conventional Pollutants,
Assuming No Misfueling (millions of 1983 dollars)
Method
Direct Estimate
Control Device
Average
1985
86
193
140
1986
170
385
278
1987
169
386
278
1988
170
389
280
1989
170
392
282
1990
173
401
288
1991
180
416
299
1992
186
431
310
-------
CHAPTER VII
VEHICLE MAINTENANCE, FUEL ECONOMY, AND
ENGINE DURABILITY BENEFITS
Switching from leaded to unleaded gasoline, or using fuel
with a lower lead content, provides benefits to vehicle owners.
The principal benefits are lower maintenance costs from lead-
induced corrosion of exhaust systems and engines. Reducing lead
in gasoline is also likely to increase fuel economy. Eliminating
lead altogether, however, may cause premature valve-seat recession
in a few engines designed to rely on lead as a valve lubricant.
Recognizing the problem of excessive deposits of lead in
engines, refiners add scavengers to leaded gasoline to prevent
such deposits. These scavengers — primarily ethylene dibromide
(EDB) and ethylene dichloride (EDO — form compounds (e.g., halo-
gen acids and lead salts) that accelerate corrosion of exhaust
systems and engine components. Section A of this chapter discusses
the maintenance benefits associated with reducing lead (and its
scavengers) in gasoline, and presents monetary estimates for
three categories: exhaust systems, spark plugs, and oil changes.
Reducing or eliminating lead in gasoline increases fuel
economy. The refining processes used to produce octane without
lead yield gasoline that is "denser" (i.e., has a higher energy
content per gallon). Lead fouls oxygen sensors in newer cars
that are misfueled by their owners; this also reduces fuel effi-
ciency. Section B presents the methods used to estimate these
fuel economy benefits.
-------
VII-2
Section C addresses the issue of engine durability. Most
modern engines have hardened valve seats or other features
designed to minimize valve-seat wear. However, many older
(pre-1971) cars, designed to operate on leaded gasoline, do
not have hardened valve seats. In such engines, lead can play a
positive role, forming a protective veneer that "lubricates" the
exhaust-valve seat, thus reducing abrasive and adhesive wear
that can erode the seat, requiring major engine repairs.
Concern about potential valve-seat damage in some engines
was the primary reason EPA proposed reducing lead to 0.10 gplg,
and did not propose banning it altogether until 1995; tests
indicate that 0.10 gplg provides a margin of safety to protect
against premature valve-seat wear. As discussed in Section C,
however, it appears that valve-seat recession may be less of a
problem than the Agency believed at the time of the August pro-
posal .
Large studies of vehicles in use have not detected signifi-
cant valve-wear problems when older engines are switched to
unleaded gasoline. Moreover, studies also indicate that lead
can cause other potentially serious problems that reduce the use-
ful lives of engines, such as accelerated ring and bearing wear
and increased rates of valve burnout. In part because of reduced
concern about potential valve-seat wear, EPA is now considering a
ban on lead in gasoline to the take effect as early as 1988.
-------
VII-3
VII.A. Maintenance Benefits
Reducing lead in gasoline can result in less frequent
replacement of exhaust systems and spark plugs and less frequent
oil changes. Our estimates of maintenance savings are based
primarily on tests of vehicles in use, either commercial fleets
(e.g., taxis) or vehicles owned by individuals for personal use.
In such tests, most of which were performed in the late 1960s or
the early 1970s, the maintenance records of vehicles operated on
unleaded gasoline were compared to those of vehicles using leaded
fuel. In addition, we have supplemented such data with the
results of laboratory tests and, in a few cases, theoretical
calculations reported in the literature.
Estimating the maintenance benefits was complicated by the
fact that in most of the studies, the leaded gasoline averaged 2.3
gplg or more (the levels that were typical when the tests were
performed), but current leaded gasoline averages only 1.10 gplg.
In addition, the tests usually examined the benefits of switching
to unleaded gasoline, not to a very low-lead gasoline of the type
permitted by the rule being promulgated. As a result, we were
forced to interpolate from limited data, first to estimate how
many of the benefits already have been reaped in reducing leaded
gasoline to 1.10 gplgi and second to predict what additional
benefits could be reaped by further reducing, but not eliminating,
lead in gasoline.
VII.A.I. Exhaust Systems
Vehicles experience fewer exhaust system failures using
unleaded gasoline than leaded because of the difference in
-------
VII-4
acidity in exhaust gas condensates. In cars using leaded fuel,
these condensates have a PH ranging from 2.2 to 2.6, while for
unleaded cars the range is 3.5 to 4.2 (Weaver, 1984b). This higher
acidity accelerates corrosion in mufflers and tailpipes.
Table VII-1 summarizes the results of four studies,
involving nine different fleets of vehicles, that examined the
effects of leaded gasoline on exhaust-system replacement rates.
(The effects on spark plug replacements were also studied; this
is discussed in Section VII.A.2.) All of the studies found
demonstrable increases in expected lifetimes (measured in
miles) of exhaust systems in unleaded vehicles when compared
with comparable leaded vehicles. The minimum increase in average
life of the exhaust system for unleaded vehicles was 86.5 percent
(the Wintringham et al., 1972, Detroit fleet).
The average exhaust-system replacement rates for leaded
cars varied greatly among the different studies, ranging from 1
per 20,500 miles (Gray and Azhari, 1972, model year 1967 vehicles)
to 1 per 58,800 miles (Pahnke and Bettoney, 1969). (The rate
per mile for the Gray and Azhari, 1972, study of consumers'
personal use cannot be computed because of inadequate data.)
Four of the fleets showed virtually no replacements of exhaust
systems in the vehicles using unleaded gasoline. Averaging the
results of all these studies, we found about one exhaust system
replacement every 56,000 miles for cars using leaded fuel, and
essentially none for vehicles using unleaded fuel during the
test periods.
-------
VII-5
TABLE VII-1 Summary of On-Road Studies of Spark Plugs and Exhaust Systems
STUDY
REPLACEMENT RATES
PER 11,000 MILES (OR PER 1 YEAR)
Pahnke & Bettoney
(DuPont, 1969)
Gray & Azhari (1972)
( Amoco )
MY 1967:
MY 1968:
Gray & Azhari (1972)
( Amoco )
Wintringham et.al.,
(Ethyl, 1972)
Detroit:
Baton Rouge:
Hick ling Partners
(Environment Canada)
(1981)
Municipal Fleets
SPARK I
UNLEADED
.534
.373
.307
.247
weight*
.440
.347
2.9 times
w/leaded
>LUGS
LEADED
.726
.840
1.085
.295
id avg .
.677
.519
as many
rehicles
EXHAUST !
UNLEADED
.0033
.149
0
.004
weightec
.155
.004
2.4 times
many for
vehicles
exclude T<
fleet)
SYSTEMS
LEADED
.187
.535
.217
.071
1 avg.
.289
.358
as
Leaded
(they
Dronto
AVG. MPV/YR
11,400
7,500
7,500
Not reported
14,575
16,850
(Unknown)
ACCUMULATED
AVG. MVP
65,000
24,000
17,000
1 to 6 yrs.
72,883
84,260
23,810 leaded
24,990 unleaded
TYPE OF SERVICE
Personal Use
Commuting and
business use
Personal Use
(Consumer
Panel)
Employee Fleet
(Business and
Personal Use)
Municipal Service
t OF VEHICLES
LENGTH OF TEST
59 matched pairs/
4.7 years
12 matched pairs/
2 and 3 years
151 matched
pairs/1-5 years
31 matched pairs
33 matched pairs/
5 years
835/5 years
LOCATION
South New Jersey
and Wilmington,
Delaware
Chicago and
suburbs
Eastern states
concentrated
in Mid-Atlantic
Detroit
Baton Rouge
Montreal
Edmonton
Toronto
-------
VII-6
It is useful to look most closely at the Ethyl Corporation
(Wintringham et al., 1972) findings, as vehicles in that study
had the greatest mileage and there is a clear geographic distinc-
tion between the fleets, which highlights the effects of climate,
The Baton Rouge fleet, after an average of over 84,000 miles of
travel per car (compared to a projected lifetime of 100,000
miles), had virtually no exhaust system repairs for unleaded
vehicles, but a rate of about 1 per 31,000 miles for leaded
vehicles. By comparison, the Ethyl Detroit fleet, after about
73,000 miles of travel per vehicle, had a rate for unleaded
vehicles of one exhaust system replacement per 46,000 miles, but
a rate of 1 per 24,000 miles for leaded vehicles. The main
reason for the different experiences in Baton Rouge and Detroit,
the authors concluded, was the greater external corrosion due to
road salts in the colder climate.
The Detroit results are consistent with the Environment
Canada findings (Hickling Partners, 1981) for two municipal
fleecs, which had 42 percent fewer exhaust system replacements
(at equivalent mileage) for cars using unleaded fuel in a cold
climate. On the other hand, the DuPont (Pahnke and Bettoney,
1971) and Amoco (Gray and Azhari, 1972) findings, conducted in
the mid-Atlantic region, in Chicago, and in the eastern U.S.,
were closer to Ethyl's in Baton Rouge; there were virtually no
exhaust repairs for vehicles using unleaded fuel. However, the
evidence suggests that exhaust system corrosion rates do vary
with climate and we have incorporated that variation into our
estimates.
-------
VII-7
Weighting Ethyl's findings for Detroit and Baton Rouge
according to the portion of registered cars in Sunbelt versus
Snowbelt states in 1982 (43 percent and 57 percent, respectively,
according to the Motor Vehicle Manufacturers Association [MVMA],
1983), mufflers nationally would last an average of three times
longer with unleaded fuel than with leaded. Unfortunately,
these studies were conducted on fleets of vehicles over several
years, but for less than the lifetimes of the vehicles.
It is possible, therefore, that the studies ended shortly before
many of the unleaded vehicles required exhaust system replacements
Perhaps the replacement rates for unleaded vehicles would have
increased significantly had the fleets traveled another 10,000
to 20,000 miles. The reported findings, thus, may have overesti-
mated the differences between unleaded and leaded vehicles.
Because of this concern, we have assumed that mufflers on
vehicles using unleaded fuel would last only twice as long (in
miles) as those on vehicles using leaded fuel.
We assumed mufflers on vehicles using leaded gasoline would
last about 50,000 miles; in the studies we reviewed, the leaded
fleets averaged about 20,000 to 60,000 miles between exhaust
system replacements.* Applying our factor of two yielded an
* Passing references in the literature and several commenters
have suggested that the metallurgy of exhaust systems was
upgraded during the 1970s, e.g., changing from cold-rolled
milled steel to chromium stainless steel. Since the more
durable metal would corrode less easily, the commenters
suggested, this design improvement might affect performance
and our estimates of benefits might be substantially over-
stated. However, on the improved exhaust systems, only the
parts from the exhaust manifold to the catalytic converter are
stainless steel. The remaining components of the exhaust
system (exhaust pipe, muffler, and tailpipe) are generally
still made of rolled steel. These are the parts that we esti-
mated would corrode from leaded gasoline. Thus, this tech-
nology change should have no effect on our estimates of savings.
-------
VII-8
expected lifetime of 100,000 miles for vehicles using unleaded
gasoline. Based on a cost of $120 per exhaust system replacement,
the savings per mile are ($120)(1/50,000 - 1/100,000) = $0.0012/
mile, or $12.00 per year for a vehicle driven 10,000 miles
annually.
This estimate must be applied with care, because it is based
on comparisons of vehicles operated on leaded gasoline with a
lead content over 2 gplg to vehicles operated on unleaded gasoline.
We were uncertain as to how much, if any, of the benefits of
reduced exhaust system corrosion might already have been reaped
as a result of reducing lead to its current level of 1.1 gplg.
Fortunately, Gray and Azhari (1971) examined exhaust system cor-
rosion rates using leaded gasoline at both 2.3 gplg and 0.5 gplg;
they found no difference between the two types of leaded gasoline.
(Both showed corrosion rates 10 to 20 times higher than those
with unleaded.) This finding suggests that no reductions in
exhaust system corrosion are reaped until the lead content falls
below 0.5 gplg. Thus, we estimated that vehicle owners switching
from leaded gasoline at 1.1 gplg to unleaded will experience
savings of $0.0012/mile. This estimate applies to misfuelers
who are deterred by the rule, and in the case of a complete ban.
Estimating the exhaust system benefits for vehicle owners who
use low-lead (0.10 to 0.50 gplg) gasoline was more problematic.
For lack of better information, we assumed linearity between zero
and 0.5 gplg; e.g., at 0.10 gplg, owners would get 80 percent of
the benefits, or 0.8(0.0012) = $0.00096/mile. This translates
to $9.60 per year for a vehicle owner driving 10,000 miles yearly.
-------
VII-9
VII.A.2. Reduced Fouling and Corrosion of Spark Plugs
The corrosive effects of lead and its scavengers also reduce
the useful life of spark plugs. As shown in Table VII-1, all of
the fleet studies showed longer intervals between spark plug
changes for vehicles operated on unleaded than on leaded. The
increases ranged from 19 percent (Gray and Azhari, 1972, consumer
use study) to 350 percent (Gray and Azhari, 1972, for model year
1968 vehicles); on average, the gain was about 80 percent.
To estimate benefits, we assumed that the average interval
between spark plug changes with leaded gasoline at 2.3 gplg would
be 10,000 miles.* That is roughly consistent with manufacturers'
recommendations in the early 1970s, before new cars used unleaded.
Applying the 80 percent improvement estimated above for users of
unleaded would allow an interval of 18,000 miles between changes.
As with the exhaust system data, these tests used leaded
gasoline at about 2.3 gplg and unleaded, so we had to make
adjustments to account for savings due to the change from 2.3 to
1.1 gplg. In 1971, Toyota reported (Champion, 1971) that fouling
* In practice, consumers appear to have changed spark plugs less
frequently; the average for leaded vehicles in the fleet tests
was about 15,000 miles. Owners who delay spark plug changes,
however, suffer losses due to decreased fuel economy; which
usually exceed the cost of replacing spark plugs at the appro-
priate interval. For consumers who change spark plugs less
frequently than optimal, the benefits of unleaded or reduced
lead gasoline will come through added fuel economy (since the
spark plugs will degrade less on the unleaded gasoline) rather
than reduced spark plug changes. For example, Graver et al.
found that spark plugs with the wrong gap and orientation led
to decrements of up to 7 percent in fuel economy. In general
these benefits will be higher, so our use of the replacement
costs is conservative.
-------
VII-10
of spark plugs occurred at the same rates with leaded gasoline
at 0.2 gplg as with unleaded gasoline. In 1972, Union Oil reported
(Champion, 1972) that spark plug performance was similar for
unleaded and leaded gasoline at 0.5 gplg. With either type of
gasoline (unleaded or 0.5 gplg), Union reported spark plugs
lasted four times longer than with leaded gasoline containing
3.0 gplg. These findings suggest that there is a threshold
below which further reductions in lead yield no additional gains
in spark plug life.
For lack of better information, we assumed that the threshold
occurred at 0.5 gplg, and that the relationship between lead and
spark plug life from that level to 2.3 gplg was linear. Thus, we
assumed that the reduction in lead from 2.3 gplg to 1.1 gplg
increases the interval by (2.3-1.1)/(2.3-0.5)(80 percent) = 53
percent, or from 10,000 miles to 15,333 miles. At $18 per spark
plug change, we estimated that reducing lead from 1.1 gplg to 0.5
gplg or below will provide benefits of (1/15,333 - 1/18,000)($18)
= $0.000174/mile. The annual benefit for a car owner driving
10,000 miles per year would be $1.74.* Note that this estimate
applies both to those who switch from leaded to unleaded and to
those who use low lead gasoline.
* By contrast, if car owners replace their spark plugs less
frequently than they should, and the fuel economy penalty of
increased spark plug degeneration from leaded fuel is only
0.5 percent, the benefits would be about double this estimate.
-------
VII-11
VII.A.3. Extended Oil Change Intervals
The combustion products that deposit on engine surfaces
cause corrosion and rusting. Engine oil accumulates much of the
debris from this corrosion, as well as some portion of the gaso-
line lead. According to at least one estimate, up to 10 percent
of the lead in gasoline ends up in the used oil, comprising up
to 50 percent of the weight of engine oil sludge (Gallopoulos,
1971).
The particles and corrosive products that accumulate in the
oil cause substantial wear in the engine, and the internal engine
rust may cause hydraulic valve lifter sticking (Cordera et al.,
1964). Besides the long-term engine wear that reduces the
durability of the engine, the vehicle driver may also experience
excessive valve noise and other performance degradation due to
this premature contamination of oil. Although rusting can occur
even in the absence of the halogen acids derived from lead salts,
these compounds are the major cause of internal rusting under
normal driving conditions (Weaver, 1984b).
The fleet studies summarized in Table VII-1 generally did
not consider oil changes or, if they did, found little difference
between the behavior of drivers using leaded and those using
unleaded. This result should not be surprising, as it is unlikely
that the vehicle owners in the studies were aware of the impact of
eliminating lead on engine oil. Presumably most drivers today are
similarly unaware, and follow the recommendations in their owners'
manuals (or habit) in changing their oil. If unleaded gasoline
increases the useful life of engine oil, however, switching to
-------
VII-12
unleaded will yield benefits in the form of improved engine dur-
ability, even if oil change intervals do not change. Thus, in
these cases, our estimates may be viewed as a proxy for improved
engine durability.
Gallopoulos (1971), of the General Motors Corporation, was
one of the first people to investigate the potential impacts of
unleaded gasoline on oil-change intervals. He concluded that
with unleaded gasoline it might be possible to decrease the
frequency of oil changes from 2 or 3 per year to only 1 per
year, but added that further investigation was needed.
Pless (1974), also of General Motors, reported more
conclusive results based on experiments with taxicabs under con-
ditions that took an unusually severe toll on oil quality. In
a group of 20 taxis (1970 model year), Pless found less piston
varnish, ring wear, and used-oil insolubles for the unleaded
vehicles after 16,000 miles of stop-and-go service.
On a fleet of 1972 taxis, Pless (1974) compared unleaded
vehicles after 16,000 miles without an oil change with leaded
vehicles (2.7 grams of lead per gallon) after 8,000 miles. The
results indicated less sludge, oil ring deposits, compression
ring wear, cam and lifter wear, and oil degradation for the
unleaded vehicles with extended oil change intervals, compared to
the leaded taxis with "normal" oil changes. While the unleaded
vehicles had somewhat greater plugging of oil filters, Pless
concluded that this was not a significant finding. Finally,
another fleet traveling predominantly short trips (closer to
typical consumer driving patterns) led Pless to conclude:
-------
VII-13
A combination of unleaded gasoline and doubled
oil change interval allowed significantly less
ring wear, and directionally less sludge,
varnish, and cam and lifter wear than did the
combination of leaded gasoline and "standard"
oil-change interval.
Subsequent to these findings, both General Motors and Chrysler
recommended lengthened periods between oil changes.
Gergel and Sheahan (1976), of the Lubrizol Corporation,
reported results similar to those of Pless, but did not find any
significant plugging of oil filters. They concluded that engine
wear was the limiting factor in extending oil change intervals,
suggesting a maximum of 12,000 miles between changes for unleaded
gasoline engines.
The evidence indicates that there is a relationship between
lead additives and oil change intervals shown through reduction
in engine and engine-parts wear (from reduced abrasive lead
particles or reduced rust), oil degradation, and general engine
and engine-part cleanliness (e.g., lack of deposits and sludge).
One indication of this relationship is the fact that manufacturers'
recommended intervals between oil changes have more than doubled
since the introduction of unleaded gasoline, and in a recent
statement to EPA the MVMA stated that using unleaded gasoline
allows the doubling of oil change intervals. (Some of the increase,
however, has reflected improved oil quality.)
To estimate the benefits of increased oil change intervals,
we assumed an interval of 5,000 miles with leaded gasoline (at
2.3 gplg) and, following Pless1 results, a doubling of the
interval to 10,000 miles with unleaded. Based on a cost per oil
-------
VII-14
change of $10.50,* the estimated benefit is then (1/5,000 -
1/10,000) ($10.50) = $0.00105/mile. The annual benefit to a car
owner would be about $10.50, based on 10,000 miles per year.
As before, this estimate is based on changing from gasoline
containing about 2.3 gplg of lead to gasoline with no lead. To
estimate what the benefits would be in going from 1.1 gplg to
lower levels, we relied on Cordera et al. (1964), who examined
the effects on engine rust of varying concentrations of lead (and
its scavengers, EDB or EDC) in gasoline. They evaluated valve-
lifter rusting at 0, 0.53, and 3.2 gplg; rusting decreased
nonlinearly with reductions in lead (and its scavengers), with
the sharpest declines occurring at low lead levels. Fitting a
smoothed curve to their data suggests that about 12.7 percent of
the total reduction in rusting would occur in going from 2.3 gplg
to 1.1 gplg, leaving 87.3 percent, or 0.873($0.00105/mile) =
$0.00092/mile in benefits for switching from leaded at 1.1 gplg
to unleaded. Based on that same curve, we estimated that going
from 1.1 to 0.10 gplg yielded 58.3 percent of the total benefit,
or 0.583(0.00105) = $0.00061/mile.
VII.A.4. Summary of Maintenance Benefits
Table VII-2 summarizes our maintenance estimates on a per
mile basis. They total $0.00038/mile for changing from 1.1 gplg
* This assumes four quarts of oil at $1.50 each, plus half an oil
filter (assuming the filter would be replaced every other oil
change) at $4 each, plus 15 minutes of labor at $10 per hour.
-------
VII-15
TABLE VII-2. Estimated Maintenance Benefits Per Mile
(cents/mile)
Standard (gplg)
Category
Exhaust systems
Spark plugs
Oil changes
Total
0.50
0.000
0.017
0.021
0.038
0.10
0.096
0.017
0.061
0.174
0
0.120
0.017
0.092
0.229
-------
VII-16
to 0.5 gplg, $0.00174/mile for tightening to 0.10 gplg, and
$0.00229/mile for eliminating lead altogether.
To calculate the benefits in each year, we combined those
estimates with estimates from our fleet model of the numbers of
miles driven by light-duty vehicles of different types. For
1986, for example, we estimate that legal leaded users of light-
duty vehicles will travel 307 million miles and that misfuelers
will travel 174 million miles. For a standard of 0.1 gplg,
assuming that it eliminates all misfueling, the estimated benefit
is then ($0 .00174/mile)(307 million miles) + ($0.00229/mile)(174
million miles) = $933 million. Table VII-3 presents year-by-
year estimates for the three alternative schedules presented in
earlier chapters. As before, the estimates assume that all
misfueling is eliminated; alternative assumptions are explored
in Chapter VIII.
Note that these monetized estimates of maintenance savings
apply only to light-duty vehicles (cars and light-duty trucks),
because we did not have data on such savings for other classes of
vehicles, such as heavy-duty trucks and busses. It is likely,
however, that such vehicles would also reap maintenance savings.
In 1986, we estimate that they will account for about one-quarter
of the demand for leaded gasoline; their share will grow in later
years. Consequently, these estimates understate the benefits in
these categories.
-------
VII-17
TABLE VII-3. Year-by-Year Estimates of Maintenance Benefits,
Assuming No Misfueling (millions of 1983 dollars)
Category
Rule
Spark Plugs
Proposed
Alternative
Final
Exhaust Systems
Proposed
Alternative
Final
Oil Changes
Proposed
Alternative
Final
Total
Proposed
Alternative
Final
1985
0
46
46
0
95
95
0
112
112
0
252
252
1986
84
84
84
503
356
503
347
267
347
933
706
933
1987
77
77
77
473
411
473
330
287
330
880
775
880
1988
73
73
73
450
450
450
318
318
318
840
840
840
1989
69
69
69
433
433
433
309
309
309
811
811
811
1990
67
67
67
422
422
422
303
303
303
792
792
792
1991
65
65
65
415
415
415
301
301
301
780
780
780
1992
64
64
64
412
412
412
301
301
301
776
776
776
-------
VII-18
VII.B . Improved Fuel Economy
Reducing the lead content of gasoline should improve fuel
economy in three ways: by increasing the energy content of
gasoline through more intense processing, by reducing the fouling
of oxygen sensors in misfueled late-model vehicles, and by reducing
the fouling of spark plugs. Energy content and oxygen sensor
benefits are discussed and monetized below. The third source of
benefits was covered, at least in part, by our estimate of increased
intervals between spark plug changes, and hence is not monetized
in this section.
VII.B.1. Energy Content
Increased reforming and isomerization of gasoline to replace
the octane lost through lead reductions increases the density
(energy content) of gasoline. Unleaded gasoline also generates
more deposits in engine combustion chambers, which increases
compression and engine efficiency slightly. Exxon (1978) has
estimated that these effects could cause a 1 to 1.5 percent
improvement in fuel economy-
To estimate the benefits of increased fuel economy from
denser gasoline, we computed the changes in density predicted by
the DOE refinery model at different lead levels. Because the
predicted change represents a relatively small difference between
two estimated large numbers and depends on the precise methods
used by refiners to raise octane, these estimates are subject to
substantial uncertainty. We used a formula developed by the
Society of Automotive Engineers (1979) to estimate the change in
fuel efficiency as a function of density. Finally, we multiplied
-------
VII-19
the estimated savings by the retail price of a gallon of gasoline,
(For this calculation, we used the retail price, $1.10 per gallon
in 1983 dollars, rather than the refinery gate price because a
reduction in gasoline consumption — which would result from
greater fuel economy -- yields savings in distribution and
retailing costs, as well as refining costs.)
VII.B.2. Reduced Fouling of Oxygen Sensors
For vehicles with oxygen sensors and closed-loop catalyst
systems, reducing lead in gasoline offers additional gains in
fuel efficiency to the extent that it reduces misfueling. In
such vehicles, lead fouls the oxygen sensor, thus reducing its
ability to optimize engine performance for maximum fuel economy.
In a recent paper, Armstrong (1984) presented data showing that
replacing the oxygen sensor as well as the catalyst in a misfuel-
ed vehicle reduced hydrocarbon emissions, indicating that leaded
gasoline causes the oxygen sensor to require a fuel mixture that
is too rich. She found that replacing the oxygen sensor reduced
tailpipe emissions by an average of 0.13 grams per mile. Because
the catalyst oxidizes most of the extra hydrocarbons that the
engine wasted with a lead-fouled sensor, it is necessary to
divide that number by (1 - catalyst efficiency) to estimate the
reduction in wasted hydrocarbons. In Armstrong's sample, the
average efficiency of the catalysts was 83.4 percent, so the
excess consumption of hydrocarbons in the misfueled vehicles was
0.13/(1 - 0.834) = 0.783 grams per mile. If the sensor is func-
tioning properly, these hydrocarbons are burned in the engine,
increasing fuel economy.
-------
VII-20
To estimate the benefits associated with reduced hydrocarbon
consumption, we estimated the number of post-1981, sensor-equipped
vehicles that would be misfueled for the first time in each year
from 1985 through 1992. We then computed the discounted (at a
real rate of 10 percent) number of miles that such vehicles
would travel, on average, over their remaining lifetimes, multi-
plied by 0.783 grams per mile and converted the resulting grams
of hydrocarbon to equivalent gallons of gasoline. Finally, we
multiplied total gallons of gasoline for each year by the price
of gasoline ($1.10).
VII.B.3. Summary of Fuel Economy Benefits
Table VII-4 presents the year-by-year estimates of fuel
economy benefits, assuming, as in past chapters, no misfueling.
The estimates are dominated by the savings due to higher fuel
density. These savings fall over time because as demand for
leaded fuel declines, the baseline fuel density rises. The savings
due to reduced oxygen sensor fouling increase over time because
the number of misfueled post-1981 vehicles would grow in the
absence of the new rule.
VII.C. Engine Durability
Lead in gasoline can have both positive and negative effects
on the durability of engines. The primary concern with unleaded
gasoline has been premature valve-seat wear in engines designed
to use leaded gasoline. Such effects have been demonstrated in
laboratory and track tests, although tests of vehicles in normal
use have failed to find any significant acceleration of valve-seat
recession. Section VII.C.I focuses on an evaluation of valve
-------
VII-21
TABLE VII-4. Year-by-Year Estimates of Fuel Economy Benefits,
Assuming No Misfueling (millions of 1983 dollars)
Category
Rule 1985 1986 1987 1988 1989 1990 1991 1992
Fuel Density
Proposed 0 168 150 97 114 113 148 140
Alternative 57 106 106 97 114 113 148 140
Final 60 168 150 97 114 113 148 140
Oxygen Sensors
Proposed 0 22 25 27 30 32 34 35
Alternative 11 22 25 27 30 32 34 35
Final 11 22 25 27 30 32 34 35
Total
Proposed 0 190 175 124 144 145 182 175
Alternative 68 128 132 124 144 145 182 175
Final 68 190 175 124 144 145 182 175
-------
VII-22
seat recession.
Tests indicate that lead and its scavengers can increase
the wear of other major engine components, and consequently,
shorten the useful lives of engines using leaded gasoline.
These negative effects of lead are discussed in section VII.C.2.
VII.C.I. Valve-Seat Recession
We reviewed two types of research in evaluating the potential
for valve-seat recession. The first type of study was engine
tests on dynamometers, done using either unusually high engine
loads to test valve durability, or cycles that simulated typical
driving patterns, or a combination of the two. The second type
of study involved on-road vehicle tests, ranging from high-load
studies to surveys of consumers' experiences. The advantage of
engine tests is their greater measurement precision and control
over test conditions. The on-road studies, on the other hand,
are more likely to reflect "real world" effects.
VII.C.I.a. Laboratory and Track Studies of Valve-Seat Recession
Laboratory studies suggest that exhaust-valve recession
results from abrasion and adhesion on the valve seat when engines
operate continuously under high temperatures, loads, or speeds.
(For detailed discussions of the mechanisms of valve wear, see
Godfrey and Courtney, 1971; Giles, 1971; or Kent and Finnigan,
1971.)
Several researchers have examined rates of valve recession
as a function of engine operating variables and the amount of
lead in the fuel. Giles (1971) and Godfrey and Courtney (1971)
-------
VII-23
were consistent in finding that recession rates were a function
mostly of engine speeds. The shape of this function apparently
varied significantly by vehicle models and years.
Table VII-5 summarizes the available laboratory and track
studies of valve recession as a function of lead concentrations.
Note that most engine studies of valve recession were conducted
at speeds and loads much greater than normal driving patterns.
For example, Giles and Updike (1971), of TRWs Valve Division,
conducted six dynamometer tests simulating vehicle speeds from
50 to 100 mph. These tests, combined with the other evidence,
led them to conclude that:
exhaust valve recession in engine I accelerates rapidly
above 70 mph.... The data shown here also indicate that
the average driver, who seldom exceeds 70 mph, should not
experience significant engine deterioration while using
lead-free gasoline. The salesman, however, who drives
15,000 turnpike miles per year at 80 mph, may well expect
valve train problems. (p. 2369)
Their data showed the rate of valve-lash loss actually decreased
slightly between 50 and 70 mph (wide open throttle at 2000 and
2800 rpm, respectively). Felt and Kerley (1971), of Ethyl
Corporation, also found that valve recession (using unleaded
gasoline) was about two-thirds lower for vehicles traveling at 60
mph than for those traveling at 70 mph, despite going 22 percent
to 280 percent more miles.
These studies were designed either to investigate the
mechanisms causing exhaust-valve-seat recession, or to show
the importance of leaded fuel combustion products in reducing
valve wear. They did not usually test for the likelihood of
-------
VII-24
TABLE VII-5. Summary of Findings of Track and Dynamometer Studies of Lead
Levels and Valve Recession
Study
Findings and Conditions
Doelling, 1971
Felt and Kerley, 1971
Fuchs, 1971
Giles et al., 1971
Giles and Updike, 1971
Godfrey and Courtney, 1971
Kent and Finnegan, 1971
Pahnke and Bettoney, 1971
U.S. Army, 1971
Engine tests at about 65 mph showed that between
0.04 and 0.07 gplg was sufficient protection.
Lead levels of 0.14, 0.07, 0.04, and 0.0 gplg.
Excess wear in continued high speed operation
(70-95 mph). Much lower wear rates in intermit-
tent operation using oil with metal additives.
Engine tests showed 0.5 gplg virtually eliminated
valve recession. Lead levels of 0.5 and 0.0.
Rapid wear on engines with unhardened valve seats
at engine speeds typical of 80 mph or greater.
Little or no excess wear on unleaded with hardened
valve seats at maximum engine speed. Limited
testing of heavy-duty truck engines with inserts
showed no increased wear.
No excess wear at less than 3000 rpm, excess wear
on unhardened valve seats above that. No excess
valve wear at 3500 rpm with hardened valve seats.
High load and speed are cause of valve recession
on unleaded.
High load engine tests showed 0.20 gplg was
sufficient protection. Lead levels of 3.0, 0.5,
0.2, and 0.0 gplg.
High load engine tests showed severe valve
recession at 0.0 gplg, none at 0.5 gplg.
Dynamometer test of three vehicles and three
stationary generators showed no excess valve-seat
wear. Generators at maximum rpm and load; had
hard valve-seat inserts. Vehicle engines at
maximum torque, wide open throttle, and 3200 rpm;
one each with unhardened, hardened, and hardened
inserts
-------
VII-25
valve recession under normal driving conditions. In particular,
intermittent operation at high speeds may well result in substan-
tially different test patterns than continuous operations. Felt
and Kerley, for example, found no significant protective effect
of metal additives in engine oil in a continuous high engine
speed test, but that metal additives reduced wear rates by a
factor of 10 in a test on the same engine with operations alter-
nating between 50 mph and high speed operation.
Overall, the laboratory studies implied that using unleaded
gasoline exclusively in vehicles with unhardened valves designed
for leaded fuel could risk premature valve failure under severe
engine loads. These studies indicated that such severe recession
is most likely to occur in engines operating at high loads or
speeds, which, for light-duty vehicles, would involve vehicle
speeds well above the legal speed limit of 55 mph for extended
periods of time (tens or hundreds of hours).
The evidence indicated that conditioning a vehicle on leaded
gasoline helped to prevent valve recession during subsequent use
of unleaded gasoline for a limited time, but did not lower the
longer-term risk. Giles (1971) measured valve wear during and
after "break-in" periods of an engine running on leaded gasoline.
He demonstrated that recession rates were high initially, even
using leaded gasoline. But, as the leaded gasoline combustion
products built up on the valve seat, recession rates dropped to
very low levels. Giles showed that, after switching the engine
to unleaded gasoline, recession rates continued to be low until
-------
VII-26
the lead deposits wore away (after about 10 hours of high engine
speed operation). Recession then rose again to high rates.
Giles and Updike also showed that vehicles with hardened
valves had no more wear on unleaded gasoline than vehicles with
unhardened valves had on leaded. This result was confirmed by
the Army dynamometer studies.
For engines equipped with hydraulic valve lifters (the vast
majority of on-highway engines), the amount of valve-seat recession
that can be tolerated before serious problems appear is about
0.07 to 0.15 inches. Engines without hydraulic lifters will
require adjustment after a much shorter time. The results of
the laboratory and test-track studies discussed above indicate
that this amount of wear can be experienced in as little as 100
hours of continuous 70 mph freeway driving (7,000 miles) in a
light-duty vehicle with unhardened valve seats. At 60 mph (a
more typical speed for the present day), this limit could still
be reached in as little as 18,000 miles of continuous high speed
operation. Miles accumulated at 55 mph and less are unlikely
to contribute to seat recession. Heavy-duty gasoline engines,
which operate at higher rpm and higher load levels, could be
affected even sooner, but most engines of that type have hardened
valve seats, or valve-seat inserts, and may have speed governors
that restrict engine rpms. These studies were done before
modern engine oils, which contain additives to reduce such wear.
Most of the laboratory studies compared valve wear with
unleaded gasoline to that with leaded at the levels typical of
-------
VII-27
the late 1960s and early 1970s (about 2.3 gplg). Several studies,
however, also tested the effects on valve wear of using gasoline
with a lower lead content. At least four studies concluded that
0.5 gplg of lead would provide sufficient valve-seat protection,
even under severe conditions (Kent and Finnigan, 1971; Pahnke
and Bettoney, 1971; Felt and Kerley, 1971; Fuchs, 1971). Kent
and Finnigan (1971) also found that "as little as 0.2 g/gal of
lead was sufficient to reduce wear to substantially zero."
Only one study examined valve wear at very low lead concen-
trations to discover how little lead was necessary to eliminate
valve recession. Doelling (1971) conducted tests at lead levels
of 0.04, 0.07, and 0.14 gplg for 100 hours each. Focusing on
the maximum recession of any one of the valves, Doelling found
no recession at 0.07 or 0.14 gplg, but found excess wear at 0.04
gplg. He thus concluded that leaded gasoline would protect
exhaust valves beginning at levels between 0.04 and 0.07 grams
of lead per gallon.
Based primarily on concerns raised by these studies, EPA's
proposed rule allowed the continued use of low-lead gasoline
through the mid-1990s. The standard of 0.10 gplg was chosen to
provide a margin of safety in protecting against valve wear.
Vll.C.l.b. Fleet Studies of Valve-Seat Recession
The laboratory tests discussed above suggest that premature
valve-seat recession in some engines with unhardened valve seats
that operate at high speeds could be a serious cost of eliminating
lead in gasoline altogether, though it should not be a significant
-------
VII-28
problem with the 0.10 gplg standard being promulgated. Studies
of engines under normal operating conditions, however, suggest
that even a ban on lead might not have major impacts on valve-seat
durability. The available studies are summarized in Table VII-6.
Several of these fleet studies found little or no incidence
of valve-seat problems with unleaded gasoline (Pahnke and Conte,
1969; Orrin et al., 1972; Gray and Azhari, 1972). Other fleet
studies were inconclusive concerning the relative incidence of
valve-seat problems for unleaded vehicles (Pahnke and Bettoney,
1971; Grouse et al., 1971; Pless, 1974). Wintringham et al.
(1972) also noted that reported incidents of valve problems were
rare among users of unleaded gasoline in the late 1960s (when at
least one major oil company sold a premium unleaded grade). Two
studies, however, cited more valve-seat problems for unleaded
than for leaded vehicles (Wintringham et al., 1972; Felt and
Kerley, 1971). Recently, EPA has become aware of a very large
test by the U.S. Army in the mid-1970s, which found no problems
using unleaded in a wide range of vehicles. These and other
studies are discussed below.
In the middle and late 1960s, Ethyl Corporation carried out
an extensive five-year study of leaded versus unleaded gasoline
use (Wintringham et al., 1972). This study included 64 matched
pairs of cars, owned and driven by Ethyl Corporation employees.
One vehicle in each pair used leaded gasoline, the other used
unleaded exclusively. The cars averaged more than 15,000 miles
per year (an average of 78,749 miles per car for the unleaded
-------
VII-29
TABLE VII-6. Summary of Findings of Consumer and Fleet Studies of Lead Levels
and Valve-Seat Recession
Study
Findings and Conditions
Grouse et al., 1971
Grouse et al., 1971
Felt and Kerley, 1971
Gray and Azhari, 1971
Orrin et al., 1972
Pahnke and Bettoney, 1971
Pahnke and Conte, 1969
Pless, 1974
U.S. Army. 1975
U.S. Post Office, 1983
A 50,000 mile test of matched pairs found an
insignificant decline in valve wear on unleaded.
Lead levels of 2.6 and 0.0 gplg.
A severe service test using a state police
patrol fleet found valve recession after 10 to
15 thousand miles. Lead levels of 3.1 and 0.0
Wintringham et al., 1972
An employee fleet test found more valve problems
at 0.0 than at 0.5 gplg.
No additional valve problems found with employee
fleet test or in a consumer survey. Lead levels
of 2.8 and 0.0 gplg.
No extra valve problems in a study of taxi
fleets. Lead levels of 2.8 and 0 gplg.
A consumer survey found no clear difference but
somewhat more valve problems. Lead levels of
2.3 and 0 gplg.
No additional valve problems for employee cars
in personal use. Lead levels of 2.8, 0.1, and
0.0 gplg.
No severe valve problems, but some valve-stem
wear with unleaded in one of the taxi fleets.
Conversion of six Army bases to unleaded produced
a valve recession rate of 1 per 10 million VMT*
for commercial vehicles. No valve recession in
other vehicles.
Conversion of 1,562 1975-model Ford trucks with
valve seat inserts to unleaded produced valve
recession rate of 1 per 15 million VMT*. 152
International Harvester trucks experienced
no valve-seat failures on unleaded.
An employee fleet test found more expensive valve
problems with unleaded; about 1 per million VMT*.
*Vehicle miles traveled.
-------
VII-30
group over the life of the test). At that time, speed limits on
the interstate highway system were 65 or 70 miles per hour.
Despite this, only four unleaded vehicles (6 percent) required
cylinder head replacements. One vehicle in the leaded group also
required a new cylinder head during the same period. In addition,
the absence of lead showed a beneficial effect in reducing the
number of burned and damaged valves -- only six vehicles in the
unleaded group required valve jobs, compared with sixteen of the
vehicles using leaded gasoline.
Three other studies, conducted about the same time, gave
similar results. Gray and Azhari (1972) reported the results of
a small fleet test and a consumer use survey, neither of which
indicated any particular problems with valve-seat recession.
Overall, engine repair costs for the unleaded group were lower
than for the leaded group, exactly the opposite of what would
have been expected if valve-seat recession were widespread. How-
ever, no details of repair records were provided, so the data
must be interpreted cautiously.
Grouse et al. (1971) provide data on four cars used in a
comparison of leaded and unleaded gasoline effects on lubricants.
The cars were operated on a more-or-less normal schedule, involv-
ing home-to-work driving on weekends and turnpike driving on
weekends. Three cars completed 50,000 miles successfully on this
schedule; the fourth suffered from valve-seat recession and had
to be dropped from the test after 34,000 miles. None of these
cars had operated on anything but unleaded fuel. This is signi-
-------
VII-31
fleant, since the researchers found that preconditioning on leaded
fuel at least doubled the mileage obtained in another test fleet
(operating under very severe patrol-service conditions) before
valve recession became a problem.
Schwochert (1969) operated an experimental catalyst-equipped
car for 50,000 miles on unleaded gasoline in a test cycle that
simulated typical city and highway driving (the Auto Manufacturers
Association's mileage accumulation cycle). Valve-seat recession
in this cycle did not exceed 0.02 inches. Subsequent operation in
a very high-speed cycle (70 to 90 mph) destroyed the valve seats
in less than 12,000 miles.
All of the tests discussed above involved light-duty vehicles.
Heavy-duty vehicles, since they often have lower power-to-weight
ratios and higher rpm at highway speeds, may suffer more severely
from valve recession with unleaded gasoline if they have unhardened
valve seats. These concerns are also applicable to a wide range of
farm, construction, and industrial equipment, much of which also
operates at high average power ratings and rpm.
In this regard, it is instructive to consider tests conducted
by the U.S. Army. These involved some 7,600 vehicles — including
light-duty cars and trucks, heavy-duty trucks, tractors, jeeps,
tactical and combat vehicles, and some motorized heavy equipment
— and lasted for three years, with about half of the vehicles
being added during the last year. Table VII-7 lists the types
of vehicles and other engines involved. The average commercial
vehicle in the study accumulated over 10,000 miles per year, and
many accumulated more. One class-6 truck put on 34,000 miles in
-------
VII-32
the first year alone, and several pick-ups accumulated over 30,000
miles in the first year. Military vehicles accumulated lower
mileage, but generally operate under high load conditions. The
study is documented in a series of reports by the Army Fuels and
Lubricants Research Laboratory (Moffit, 1972; Russel and Tosh,
1973; Tosh et al., 1975; Tosh, 1976). Given the the broad assort-
ment and diverse ages of the vehicles involved, it seems likely
that many of these vehicles did not have hardened valve seats.
The results of this test were negative — no untoward main-
tenance problems that could be attributed to the use of unleaded
gasoline were experienced. Overall, an engine failure rate of
0.5 percent was experienced. This rate was stated as being
comparable to experience with leaded gasoline. Only three cases
of valve-seat recession were reported, all in light-duty vehicles.
This is especially significant because the test was conducted
before the imposition of the 55 mph speed limit, and many of the
posts were located in remote areas, so that considerable highway
driving would be expected. The conclusions of the Army study
are worth quoting:
From the evaluation results, it can be concluded
that commercial, tactical and combat vehicles,
and all other equipment used in this program can
operate satisfactorily during their normal day-
to-day activities without any fuel economy
penalties and with no apparent increases in
vehicle maintenance or operating costs so long as
unleaded gasoline meeting VV-G-00169A Federal
specification is used. (Tosh, 1976, p. 34; emphasis
in original)
The Federal specification cited is essentially that for present-
day commercial unleaded gasoline.
-------
VII-33
TABLE VII-7 Vehicle and Engine Types in U.S.
Commercial Vehicles
Army Unleaded Gasoline Test
Cars 445
Light-Duty Trucks
0-6,000 pounds 1,003
6,000-10,000 pounds 429
Medium and Heavy-Duty Trucks
10,000-14,000 pounds 68
14,000-16,000 pounds 57
16,000-19,500 pounds 163
19,500-26,000 pounds 43
26,000-33,000 pounds 63
33,000 pounds plus 28
Unclassified cars and trucks 411
Buses 87
Tractors 84
Construction and Other Equipment
Cranes 38
Graders 5
Fork Lifts 256
Generators 527
Miscellaneous Construction Equip. 255
Other Vehicles and Engines
Scooters 40
Outboard motors 41
Lawn mowers 225
Motorcycles 7
Tactical Vehicles
1/4 ton trucks* 2785
3/4 ton trucks 8
1 1/4 ton trucks 919
Other tactical trucks 83
*These vehicles did not have
hardened valve seats as of 1971;
status of valve seats for other
tactical vehicles unsure.
-------
VII-34
In 1975, shortly after this test, all branches of the U.S.
Armed Services converted completely to unleaded gasoline wherever
it was available. A monitoring system was set up to detect
subsequent problems and no special vehicle maintenance or other
problems were experienced since this conversion (M. DePara, U.S.
Army, Belvoir Research and Development Center, personal communi-
cation) .
Data provided by the U.S. Postal Service tell a very similar
story for heavy-duty trucks in their service. The Postal Service
has operated some 1,562 1975-model Ford heavy-duty trucks (22,000
pounds) on unleaded gasoline since 1980. These trucks were
originally purchased in 1975, and travel about 50,000 miles per
year, on average. By 1980, most of them were on their second or
third engine rebuild or replacement, so that there was a wide
variety of engine mileages -- from zero to about 100,000 miles --
represented in the fleet. All of the new and rebuilt engines in
the fleet used hardened valve seat inserts (as do most heavy-duty
trucks).
In the 42 months or so since switching to unleaded, the
Postal Service has recorded 69 instances of valve problems (a
valve failure rate of 4.4 percent) and 18 cases of valve-seat
problems (a failure rate of 1.2 percent), while operating these
trucks for an average of about 175,000 miles each on unleaded
gasoline. This would normally include at least one full engine
rebuild (M. Sanders, U.S. Postal Service, personal communication).
For comparison, Ford indicated that its warranty data for the same
types of engines — presumably run primarily on leaded fuel —
-------
VII-35
showed higher valve and cylinder head failure rates (Ford Auto-
motive Emissions Office, personal communication). The Postal
Service has experienced no significant mechanical or operating
problems as a result of using lead-free gasoline in its fleet.
VII.C.I.e. Other Types of Engines
The studies described above generally involved on-road
vehicles (cars and trucks), although the Army study also included
some construction equipment, stationary generators, motorcycles,
and outboard engines. To investigate possible valve-seat damage
in smaller engines, such as those used in lawnmowers, snowblowers,
garden tillers, and snowmobiles, we contacted three manufacturers
of small engines (Briggs, Tecumseh, and Kohler). All said either
that their engines could almost always use either leaded or
unleaded, or that they specifically recommend unleaded. Repre-
sentatives from these companies also stated that they believed
that this would be true for all of the engines that they had
manufactured for at least the last 10 years, and were not aware
of any design changes that would have made this untrue even for
earlier engines.
Marine engines are generally of two types: inboard and
outboard. Inboard engines typically are adapted from automobile
or truck engines, so we would expect the data on light-duty
vehicles to apply to them as well. Outboard engines are almost
all two-stroke engines, for which the fuel is mixed with a special
type of oil. For most such engines, unleaded does not appear to
cause any serious problems; however, for high-output two-stroke
-------
VII-36
engines (125 hp and above), cylinder-wall scoring and premature
bearing failures can occur (Weaver, 1984b). Several solutions
are possible for this problem, if lead is banned. One would be
to allow lead to remain in gasoline for marine use. The other
would be to allow lead or another additive to be added to the oil
that is mixed with gasoline for two-stroke engine use.
Another engine class of potential concern is farm equip-
ment. Although diesel engines now dominate the market for tractors
and other large pieces of farm equipment, there are many older
gasoline-powered engines still in use on farms. The Army tests
involved some farm as well as non-farm tractors that are likely
to be used under conditions similar to those on farms. That
study also included portable generators, which should be similar
to many small engines used on farms to power stationary equipment.
To the extent that the Army data are applicable, it seems that
the Agency's action should not have a significant impact on the
durability of engines used on farms.
Vll.C.l.d. Alternatives to Lead to Avoid Valve Recession
As noted earlier, most engines manufactured over the last
decade have used induction-hardened valve seats, hardened valve-
seat inserts, or other mechanisms to eliminate potential valve-
seat recession problems without lead. It is not feasible to modify
existing engines in those ways, however, except during major
engine rebuilds. Thus, the most promising way of coping with
potential valve-seat durability problems in the total absence of
lead would be alternative additives.
-------
VII-37
Relatively little research has been done on such alternatives,
presumably because there is little incentive to develop and market
them so long as lead remains available. (Because lead is a rela-
tively cheap octane booster, it is a "free" valve lubricant.)
Limited work on the subject, however, suggests several possible
alternatives, the most promising of which is phosphorus.
Several experiments suggest that phosphorus in unleaded
gasoline could reduce or eliminate the risk of valve recession at
high speeds. Specifically, at about 0.06 or 0.07 grams of phos-
phorus per gallon, valve wear proceeds at one-half to one-third
the rate occurring with no additives (Giles and Updike, 1971;
Kent and Finnigan, 1971; Felt and Kerley, 1971; Wagner, 1973).
The tests were run primarily under unusually high loads or speeds,
similar to conditions used in the previously-described studies
of valve recession.
Amoco (Wagner, 1971) reported that its road tests of heavily
loaded 1970-vintage cars, for 20,000 to 30,000 miles at average
speeds of 60 mph (and up to 70 mph), found that 0.07 g/gal of
phosphorus was effective in controlling valve recession for nearly
90 percent of the cars tested. The phosphorus more than halved
the rates of recession for the cars that, without lead or phos-
phorus, had experienced sinkage rates of more than 0.01 inches
per 10,000 miles. Kent and Finnegan found, however, that at lower
load conditions and 2300 rpm for 80 hours, phosphorus was fully
protective against any valve-seat widening or oxidation. Cordera
et al. (1964) found the presence of phosphorus in the gasoline was
critical to exhaust valve-life durability. All of these results
-------
VII-38
indicated that adding phosphorus to unleaded gasoline would
substantially reduce the risk of valve recession for those vehicles
at risk. Because phosphorus has a negative impact on catalysts,
however, it would be necessary either to have a special grade of
unleaded gasoline with phosphorus for older engines, or to make
phosphorus available as a separate additive.
In considering a possible ban on leaded gasoline, EPA is
soliciting comments on phosphorus and other possible additives
to deal with potential valve-seat recession. The Agency is
asking for comments also on other possible approaches, such as
making leaded gasoline available on a very limited basis (e.g.,
at marine terminals).
VII.C.2. Megative Effects of Lead on Engine Durability
Wear in engines may be due either to physical processes
(abrasive wear) or to chemical effects (corrosive wear). Abrasive
wear results from the rubbing contact between two parts. Corrosive
wear is a phenomenon akin to engine rusting — it occurs where
chemicals can attack a surface subject to wear, and either dissolve
it directly, or combine with it to form a less wear-resistant
material.
Lead and its salts are effective solid lubricants. Thus, it
might be expected that engine components exposed to lead deposits
might suffer less abrasive wear. However, discussions with TRW
(H. McCormick, TRW Piston Ring Division, personal communication)
indicate that lead deposits may actually increase abrasive wear
of piston rings. In addition, the acid combustion products of
-------
VII-39
lead scavengers contribute to corrosive wear, especially if water
is present. Hudnall et al. (1969) have commented that corrosive
wear can be much greater than abrasive wear, especially in cold
operation. Heavy-duty engines, however, which are less subject
to rusting due to their higher operating temperatures, also suffer
less from corrosive wear (Hudnall et al., 1969).
Several investigators have compared the levels of wear
observed with leaded and unleaded gasolines. Cordera et al.
(1964) compared wear results with the standard scavenger mixture
containing both chlorine (from EDC) and bromine (from EDB) with
wear using only bromine. They found that eliminating the chlorine
reduced wear rates by about 40 percent. In another test, they
examined the effects of different lead (and lead-scavenger)
concentrations on wear rates. They found that going from 3.0 to
1.5 grams of lead per gallon reduced wear rates by around 40 per-
cent, with a small additional improvement at 0.5 grams per gallon.
Going to unleaded gasoline from 0.5 grams actually increased wear
rates, although wear was still lower than at 3 grams. Gagliardi
and Ghannam (1969) obtained similar results in an 18,000-mile
fleet test. They found that piston ring wear was lowest at 0.5
grams per gallon of lead, and increased slightly for both zero
and 1.5 grams. Wear at 3 grams per gallon was 70 to 200 percent
greater than at 0.5 grams.
The reduction in wear with low-lead gasoline is not
surprising, but the observed increase in wear when going from low-
lead to unleaded gasoline is. One reasonable explanation for this
increase would be the solid lubricating effects of lead deposits
-------
VII-40
on the cylinder walls, which would be present with low-lead
gasoline but not with unleaded. Alternatively, some differences
in combustion or lube-oil chemistry due to the presence of lead
might account for the difference in wear. The available labora-
tory data give conflicting impressions as to the degree of in-
creased wear in changing from low-lead to unleaded gasoline.
Cordera and coworkers found a rather large increase, while
Gagliardi and Ghannem reported only a small effect. To better
evaluate the magnitude of this effect in actual use, it is in-
structive to consider the results of in-use fleet testing.
Orrin et al. (1972) and Carey et al. (1978) have reported
the results of two tests of leaded vs. unleaded gasoline in taxi
fleets. One fleet operated in Oakland, California for 48,000
miles, and the other operated in Montreal, Canada for 80,000
kilometers. These tests would be expected to favor leaded gaso-
line. As taxis generally operate nearly continuously for 8 to 24
hours per day, they spend a comparatively small amount of time in
warm-up and cold operation -- the conditions that tend to favor
corrosion. Despite this, the results of these tests showed a
distinct advantage for unleaded gasoline.
In each case, wear measurements in taxis using leaded fuel
were 70 to 300 percent greater than those for taxis using unleaded.
Piston-ring and cylinder-bore wear, perhaps the most critical
areas, ranged from 70 percent to 150 percent greater with leaded
than with unleaded fuel. Neither fleet experienced any overt
problems with rust, possibly indicating that significant corrosive
wear can occur even in the absence of visible rust. Alternatively,
-------
VII-41
some of the difference might be due to increased abrasive wear
due to lead deposits, as suggested by TRW.
These data, which closely match those of Gagliardi and
Ghannam (1969) in laboratory tests, appear to indicate that the
actual decrease in wear with unleaded gasoline is almost as
great as that found with low-lead fuel. Since lead deposits
appear to form similarly at 0.1 and 0.5 grams per gallon, it is
probable that wear at 0.1 gram per gallon would be similar to or
lower than that at 0.5. The data also indicate that the recent
reduction to 1.10 grams per gallon should have produced a signifi-
cant decrease in corrosive wear, at least with regular oil changes,
However, oil changes are frequently irregular, and blowby volume
in worn engines is much greater than in the new engines on which
these test were conducted. Both of these factors would tend to
increase corrosive wear rates, even at the current 1.10 gram per
gallon level. Thus, either lower-lead (0.5 and 0.1 gplg) or
unleaded fuel could be expected to produce a significant reduction
in wear rates from those observed with regular leaded gasoline,
even at 1.1 gram per gallon.
The economic significance of reduced wear rates would be
considerable. At present, worn-out piston rings and cylinder
bores are one of the major causes of failure in gasoline engines.
They result in poor fuel economy, poor performance, and increased
emissions. Repairing this condition requires an engine overhaul,
at a cost of $500 to more than $1,000, depending on the engine.
Many older vehicles with these problems are simply junked and
not repaired.
-------
VII-42
In new engines, the use of unleaded gasoline can extend
piston-ring lives significantly — by as much as a factor of two
(H. McCormick, TRW Piston Ring Division, personal communication).
It is not clear, however, how to extrapolate from this finding to
estimate the effects on the service lives of engines now in use.
If corrosive wear is the major factor in piston-ring wear, then
one would expect it to get worse over time as blowby rates increase
and, generally, maintenance practices degrade. The general shift
to shorter trips and less annual mileage with increasing age
would also increase corrosion. Thus, a car that had run for
half of the expected lifetime of its piston rings would probably
have accumulated somewhat less than half its lifetime wear, and
a radical decrease in wear might have more than proportional
benefits.
On the other hand, if abrasive wear due to lead deposits is
the dominant factor, these deposits would last for some time
after the switch to unleaded, and would thus result in less than
proportional increases in service life. Overall, the effect of
unleaded gasoline in increasing the remaining service life of the
piston rings is probably best estimated as being linearly propor-
tional to the remaining life. The typical service life for
piston rings in cars using leaded gasoline is about 70,000 to
80,000 miles. A car driven 50,000 miles on leaded gasoline
could expect perhaps another 25,000 miles before needing an
overhaul. Switching to unleaded would probably increase this
by 70 to 150 percent, giving a new expected time-to-overhaul of
about 43,000 to 67,000 miles. Since parts other than piston
-------
VII-43
rings can fail, the actual increase in engine life would probably
be closer to 70 percent than 150 percent. Because a substantial
part of the oil change benefits that have been monetized may
reflect reduced engine wear rather than longer oil change inter-
vals, we have not included any monetized benefits for this
category.
VII.C.3. Summary of Engine Durability Effects
We have made no attempt to monetize the potential engine
durability benefits or costs of reducing or eliminating lead in
gasoline. The net impact of lead on engine durability is unclear.
For most engines, it appears that lead does substantially more
harm than good. For some, however, lead may play an important
role in reducing the risk of valve-seat recession at high loads
and speeds, although tests of vehicles in use suggest that few
engines need lead under normal operating conditions. It appears
that the low valve-seat wear rates in in-use fleets are due both
to the low proportion of time spent at high rpm, and the ability
of engine oil additives to build up a protective coating during
the low rpm use which then protects the engine during intermittent
high speed operation. The 0.10 gplg standard provides a margin
of safety to protect against potential recession at high loads
and speed since it also has been shown to build up a protective
layer during low and moderate rpm use.
For the longer run, when EPA proposes to ban lead in gasoline,
several solutions may be possible for those few engines that need
protection against valve-seat recession. First, phosphorus or
some other additive may prove to provide effective protection.
-------
VII-44
Second, it may be possible to make leaded gasoline available on a
very limited basis, so that its use is restricted to those engines
that truly need it.
If acceptable alternatives are not developed, the Agency may
be forced to accept some increased risk of premature valve-seat
recession in some engines as the price of eliminating the severe
health and environmental consequences associated with lead in
gasoline. As part of its continued deliberations on a possible
ban, the Agency will attempt to develop quantitative estimates
of the magnitude of this problem.
-------
CHAPTER VIII
COST-BENEFIT ANALYSIS OF ALTERNATIVE PHASEDOWN RULES
EPA considered many alternative phasedown schedules before
deciding on the final rule. This chapter compares the costs and
benefits of those alternatives, based on the methods and results
described in earlier chapters. Section A of this chapter sum-
marizes the estimates of benefits and costs, and compares them
to the estimates contained in the Preliminary Regulatory Impact
Analysis (RIA) issued when the rule was proposed in August 1984.
Section B presents the cost and benefit estimates for different
lead standards under various assumptions about the impact of the
rule on misfueling. Section C examines the impact of the proposed
banking rule on the costs and benefits of the final phasedown
rule. Finally, Section D summarizes the conclusions and EPA's
rationale for selecting the final phasedown schedule.
Throughout this chapter, we present benefit estimates with
and without adult blood-pressure-related benefits. The estimation
of a dose-response relationship between blood lead and blood pres-
sure is very recent. The paper presenting those results (Pirkle
et al., 1985) has just been published in a peer-reviewed journal.
A summary of the results of that study and their application to
this rule was also placed in the docket for this rulemaking
several weeks before it closed for public comment. Until the
scientific community has had an opportunity for more intensive
review, however, EPA is not relying on these results to reach
final regulatory decisions on lead in gasoline. As the results
in this chapter show, these blood-pressure-related benefits
-------
VIII-2
greatly increase the total estimated benefits, but even when
they are not included the benefits of the rule exceed the costs
by a large margin.
VIII.A. Summary of Cost and Benefit Estimates
Table VIII-1 summarizes the cost and benefit estimates for
the 0.10 gplg standard in 1986; these estimates assume that the
rule would eliminate all misfueling. As shown in the table, the
benefits total $7.9 billion, while the estimated cost is only
$607 million, resulting in net benefits of $7.3 billion. About
75 percent of the estimated benefits are attributable to reduc-
tions in cardiovascular diseases associated with elevated blood
pressure. Even if these benefits associated with adult health
are excluded from the calculation, however, the benefits still
exceed the costs by more than a three-to-one margin.
Table VIII-1 also presents the cost and benefit estimates
contained in the Preliminary RIA that accompanied the proposed
rule. The most striking difference, of course, is in adult blood
pressure benefits, which were not included in the Preliminary RIA.
All of the other categories show some changes as well,
reflecting changes made in response to comments or to newly
available information. The higher cost estimates reflect several
changes in base-case assumptions. The two most important are:
reduced yields from reformers operated at high severity and
reduced segregation of naphthas for optimal allocation to pro-
cessing units. Partly offsetting those changes are the use of
newer, more efficient catalysts in FCC units, and lower oil
prices, as discussed in Chapter II.
-------
VIII-3
TABLE VIII-1. Costs and Monetized Benefits of 0.10 gplg in 1986,
Assuming No Misfueling: Comparison of Current and
Draft RIA Estimates (millions of 1983 dollars)
MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED
Current
602
5,927
278
933
190
7,930
Draft RIA
271
N.A.
348
840
360
1,819
BENEFITS
TOTAL REFINING COSTS
607
575
NET BENEFITS
7,323
1,244
NET BENEFITS EXCLUDING
BLOOD PRESSURE
1,396
1,244
-------
VIII-4
The category of "Children's Health Effects" is higher for
several reasons. First, the CDC recently reduced the blood lead
and FEP levels that define lead toxicity from 30 ug/dl to 25 ug/dl;
this greatly increases the number of children requiring at least
some follow-up medical testing or treatment. Second, in estimat-
ing medical costs, we have relied on recently published recommend-
ations that call for more extensive testing and treatment than
assumed in the Preliminary RIA. Finally, because of the change
in the CDC definition of lead toxicity and the greater weight
that the most recent draft of the Lead Criteria Document gives
to cognitive effects at blood lead levels in the range of 30 to
50 ug/dl blood lead, we have increased our estimate of the number
of children likely to warrant compensatory education in the
absence of further reductions in gasoline lead.
As discussed earlier, the estimates of the benefits of
reduced emissions of conventional pollutants are the average of
two estimation methods: the value implied by the cost of the
pollution control equipment destroyed by misfueling and a direct
valuation of some of the health and welfare effects associated
with these pollutants. (The second method, direct valuation,
is based on an incomplete quantification of these health and
welfare effects.) The reduction in the overall estimates com-
pared to the Preliminary RIA reflects a decrease in the direct
estimate, as discussed in Chapter VI.
The changes in the maintenance estimates reflect refinements
in the fleet model used to estimate the number of miles traveled
-------
VIII-5
by different classes of vehicles. The reduction in the fuel-
economy estimate is caused by changes in the predicted fuel
density, which in part results from changes in the inputs to the
refinery model.
Table VIII-2 presents some important non-monetary measures
of the estimated benefits of the 0.10 gplg standard in 1986
(again, assuming that the rule eliminates all misfueling). Note
that we estimate that the rule will reduce by 172,000 the number
of children above the new CDC blood-lead limit of 25 ug/dl.
The reductions in the numbers of children at lower, but still
possibly harmful, blood-lead levels are even greater; we estimate
that 1.7 million fewer children will experience blood lead levels
over 15 ug/dl in 1986 as a result of the rule.
If the rule eliminates misfueling, we estimate that it will
eliminate over 2.5 million tons of excess emissions of HC, NOX,
and CO. The current estimates are higher than those made in the
preliminary RIA because we have used the results of the 1983 EPA
tampering and misfueling survey, which show higher rates than the
1982 survey employed earlier, and because of refinements in the
fleet model used to estimate the number of miles driven by dif-
ferent classes of vehicles.
Table VIII-2 also reports estimates of the numbers of reduced
health effects among adults in 1986. As already discussed, these
estimates are restricted to males aged 40 to 59, and the estimates
for myocardial infarctions, strokes, and deaths apply only to
white males in that age range. Despite these limitations, the
estimated benefits are large, ranging from 1.8 million fewer
cases of hypertension to 5,160 fewer deaths from all causes.
-------
VIII-6
TABLE VIII-2. Nonmonetary Measures of Health and Environmental
Benefits of 0.10 gplg in 1986, Assuming No Mis-
fueling: Comparison of Current and Draft RIA
Estimates
Reductions in thousands of
children above selected
blood lead levels
30 ug/dl
25 ug/dl
20 ug/dl
15 ug/dl
Reductions in thousands of
tons of pollutants
Hydrocarbons
Nitrogen oxides
Carbon monoxide
Reductions in adult male
health effects
Thousands of
hypertensives
Myocardial
infarctions
Strokes
Deaths
Current
52
172
563
1,726
305
94
2,116
1,804
5,350
1,115
5,160
Draft RIA
52
172
563
1,726
247
81
1 ,646
N.A.
N.A.
N.A.
N.A.
-------
VIII-7
VIII.E. Comparisons of Alternative Lead Levels
In the August 1984 proposal, EPA discussed a range of alterna-
tive schedules and presented two specific possibilities: (1) a
one-step reduction to 0.10 gplg starting January 1, 1986 and (2)
a phasedown with several steps — 0.50 gplg on July 1, 1985;
0.30 gplg on January 1, 1986; 0.20 gplg on January 1, 1987; and
0.10 gplg on January 1, 1988. The final rule imposes the 0.10
gplg as of January 1, 1986, and also requires a reduction to
0.50 gplg as of July 1, 1985. In addition to these three sched-
ules, however, the Agency considered many other possibilities.
For 1986, the Agency considered levels between 0.1 gplg and
0.5 gplg. Table VIII-3a presents the cost and benefit estimates
for those alternatives, assuming in each case that all misfueling
would be eliminated. Net benefits are maximized at the tightest
of those limits, 0.10 gplg, whether or not adult blood pressure
benefits are included. Table VIII-3b shows the estimates assum-
ing that the rule fails to have any impact on misfueling; again,
net benefits are maximized at 0.10 gplg.
Neither of these two extreme assumptions about misfueling
(that it will be eliminated entirely or that it will continue
unabated, even at very low lead levels) appears realistic. EPA
believes it is more likely that the number of misfuelers is a
declining function of the lead level, primarily because the
manufacturing cost differential between leaded and unleaded
declines and then reverses as the lead limit is tightened. We
expect changes in manufacturing costs to be at least partly
reflected in retail prices. Although we cannot be certain that
-------
VIII-8
TABLE VIII-3a. Costs and Monetized Benefits of Alternative Lead
Levels in 1986, Assuming No Misfueling
(millions of 1983 dollars)
0.50
Lead
0.40
Level (
0.30
Igpig)
0.20
0.10
MONETIZED BENEFITS
Children's health effects 466 504 539 571 602
Adult blood pressure 4,018 4,483 4,955 5,436 5,927
Conventional pollutants 278 278 278 278 278
Maintenance 517 608 706 808 933
Fuel economy 119 119 128 136 190
TOTAL MONETIZED 5,398 5,992 6,606 7,229 7,930
BENEFITS
TOTAL REFINING COSTS 243 305 386 472 607
NET BENEFITS 5,155 5,687 6,220 6,757 7,323
NET BENEFITS EXCLUDING 1,137 1,204 1,265 1,321 1,396
BLOOD PRESSURE
-------
VIII-9
TABLE VIII-3b. Costs and Monetized Benefits of Alternative Lead
Levels in 1986, Assuming Full Misfueling
(millions of 1983 dollars)
Lead Level (gplg)
0.50 0.40 0.30 0.20 0.10
MONETIZED BENEFITS
Children's health effects 403 455 504 550 592
Adult blood pressure 3,328 3,920 4,526 5,144 5,778
Conventional pollutants 00000
Maintenance 186 329 482 642 838
Fuel economy 44 97 97 115 177
TOTAL MONETIZED BENEFITS 3,961 4,801 5,609 6,451 7,385
TOTAL REFINING COSTS 178 260 350 460 627
NET BENEFITS 3,783 4,541 5,259 5,991 6,758
NET BENEFITS EXCLUDING
BLOOD PRESSURE 455 621 733 847 980
-------
VIII-10
unleaded will be priced below leaded at the retail level even if
its manufacturing cost is lower, we do expect the price differential
between leaded and unleaded to decline and that the decline will
cause some misfuelers to switch to unleaded gasoline.
Figure VIII-1 plots three possibilities for how the percent-
age of current misfuelers might decline as a function of the
lead limit. Each of the three curves assumes that 50 percent of
misfueling would be eliminated at 0.25 gplg, the estimated point
at which the manufacturing costs of leaded and unleaded intersect.
They differ, however, in the assumed rates at which misfueling
changes. Curve A assumes that misfueling declines linearly from
0.50 gplg to 0 gplg; at 0.10 gplg, misfueling is 20 percent of
its base level. Curve B also assumes that the decline in misfuel-
ing begins at 0.50 gplg and ends at 0 gplg, but that the rate of
decline is most rapid over the intermediate range. Curve C is
similar to B, but the change is compressed to the range 0.40
gplg to 0.10 gplg; it assumes no misfueling at 0.10 gplg.
Figure VIII-2 plots the net benefits of the alternative lead
limits for all five assumptions about misfueling: no misfueling,
full misfueling, and the three intermediate cases. In all five
cases, not surprisingly given the results in Tables VIII-3a and
VIII-3b, the net benefits peak at 0.10 gplg. Figure VIII-3
presents similar estimates, but excludes the adult blood pressure
benefits; again, net benefits peak at 0.10 gplg, though they are
much smaller than with the inclusion of the adult health effects.
It is impossible to determine which assumption about
-------
c
0)
D
«*-
V)
c
-------
0)
o
.a
S.X
V)
!Z
«4-
0)
c
0)
00
8
7
6
5 -
4 -
3 -
2 -
1 -
0.5
No Misfueling
Full Misfueling
0.4
0.3
Lead Content (gplg)
0.2
<
M
I
M
ro
0.1
FIGURE VIII-2
Net Benefits (Including Blood-Pressure-Related Effects) as
Functions of Lead Level and Misfueling
-------
CO
c
o
-O
N^
OT
.±i
»*-
o>
c
0)
CO
•M
0)
Partial Misfueling
Full Misfueling
H
U)
0.3
Lead Content (gplg)
FIGURE VIII-3. Net Benefits (Excluding Blood-Pressure-Related Effects) as
Functions of Lead Level and Misfueling
0.1
-------
VIII-14
misfueling is most accurate. To do so would require knowing how
gasoline retail prices will respond to changes in production
costs, and how misfuelers will respond to changes in prices.
Thus, most of our estimates in this chapter and the next one
present both extreme possibilities with respect to misfueling.
It is useful, however, to have a standard "partial misfuel-
ing" case for making comparisons. For that case, we have settled
on the simplest of the three curves in Figure VIII-1, curve A,
which assumes that misfueling declines linearly from 100 percent
of its current level at 0.50 gplg and above, to zero at 0 gplg.
Under that assumption, 20 percent of the misfuelers continue to
use leaded gasoline at 0.10 gplg. We believe this is a reasonable
estimate, as some gasoline stations are likely to continue to
sell leaded gasoline at a lower price than unleaded, and some
misfuelers may continue to buy leaded even if it costs more than
regular unleaded, either because they desire higher octane or
because they mistakenly believe that leaded gasoline is better
for their engines. Table VIII-3c presents the cost and benefit
estimates under this assumption, i.e., that misfueling declines
from 100 percent of its current level at 0.50 gplg to 20 percent
at 0.10 gplg.
For 1987, EPA considered two alternative levels: 0.20 gplg
and 0.10 gplg. Tables VIII-4a through VIIl-4c present the cost
and benefit estimates for the no-misfueling, full-misfueling, and
partial-misfueling cases, respectively. As in the earlier tables,
net benefits are maximized at 0.10 gplg, whether or not adult
blood pressure benefits are included.
-------
VIII-15
TABLE VIII-3c. Costs and Monetized Benefits of Alternative Lead
Levels in 1986, Assuming Partial Misfueling
(millions of 1983 dollars)
Lead Level (gplg)
0.50 0.40 0.30 0.20 0.10
MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED BENEFITS
TOTAL REFINING COSTS
NET BENEFITS
403 465 518 563 600
3,328 4,033 4,698 5,319 5,897
0 56 111 167 222
186 385 572 742 914
44 101 109 128 187
3,961 5,039 6,008 6,918 7,821
178
269
364
467
608
3,783 4,770 5,643 6,451 7,213
NET BENEFITS EXCLUDING
BLOOD PRESSURE
455
738
946 1,131 1,316
-------
VIII-16
TABLE VIII-4a. Costs and Monetized Benefits of Alternative
Lead Levels in 1987, Assuming No Misfueling
(millions of 1983 dollars)
MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED
BENEFITS
TOTAL REFINING COSTS
NET BENEFITS
Lead Level
0.20
522
5,262
278
775
132
6,968
452
6,516
(gpig)
0.10
550
5,707
278
880
175
7,590
553
7,037
NET BENEFITS EXCLUDING
BLOOD PRESSURE 1,255 1,330
-------
VIII-17
TABLE VIII-4b. Costs and Monetized Benefits of Alternative
Lead Levels in 1987, Assuming Full Misfueling
(millions of 1983 dollars)
MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED
BENEFITS
TOTAL REFINING COSTS
NET BENEFITS
Lead Level
0.20
501
4,940
0
596
106
6,143
441
5,702
(gpig)
0.10
539
5,543
0
111
150
7,009
578
6,431
NET BENEFITS EXCLUDING
BLOOD PRESSURE 762 888
-------
VIII-18
TABLE VIII-4C. Costs and Monetized Benefits of Alternative
Lead Levels in 1987, Assuming Partial
Misfueling (millions of 1983 dollars)
MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED
BENEFITS
TOTAL REFINING COSTS
NET BENEFITS
Lead Level
0.20
513
5,133
167
703
122
6,638
448
6,191
(gpig)
0.10
547
5,675
222
859
170
7 ,474
558
6,916
NET BENEFITS EXCLUDING
BLOOD PRESSURE 1,058 1,241
-------
VIII-19
For 1985, EPA considered five alternative levels: 1.10 gplg
(i.e., no change), 0.80 gplg, 0.70 gplg, 0.60 gplg, and 0.50 gplg.
Table VIII-5 presents the estimated costs and benefits, assuming
full misfueling. (Estimates are not presented for the no-misfuel-
ing case, as we doubt that standards above 0.50 gplg will have
enough impact on the price of leaded gasoline to make a signifi-
cant difference in misfueling.) Note that the estimates cover
only half a year, as none of the standards considered for 1985
would take effect until the middle of that year (July 1). Again,
net benefits are maximized at the tightest standard discussed in
the August proposal, 0.50 gplg.
The net benefits of the 0.50 gplg standard in 1985 are
substantial: $264 million if blood-pressure-related benefits are
not included, and $2.0 billion if they are. Moreover, as discussed
in Chapter II, all available measures indicate that the refining
industry can comply easily with that portion of the rule; reducing
lead to 0.50 gplg should require minimal adjustments in refinery
operations.
Table VIII-6 compares the present values of the costs and
benefits of three phasedown schedules over the period 1985-1987:
the original primary proposal; the more gradual, illustrative
phasedown presented in the Notice of Proposed Rulemaking; and
the schedule in the final rule. In all cases, the costs and
benefits have been discounted at 10 percent (real) to the begin-
ning of 1985. (The 1985 estimates were discounted for half a
year, the 1986 benefits for a full year, and the 1987 benefits
for two years.) All of the schedules yield substantial
-------
VIII-20
TABLE VIII-5. Costs and Monetized Benefits of Alternative
Lead Levels in 1985, Assuming Full Misfueling
(millions of 1983 dollars)
MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
0.80
124
837
0
37
-5
Lead Level
0.70
159
1,126
0
54
0
(gpig)
0.60
193
1,423
0
80
31
0.50
223
1,724
0
102
35
TOTAL MONETIZED
BENEFITS
993
1,339
1,727
2,084
TOTAL REFINING COSTS
NET BENEFITS
NET BENEFITS EXCLUDING
BLOOD PRESSURE
44 56 75 96
949 1,283 1,652 1,988
112
157
229
264
-------
VIII-21
TABLE VIII-6. Present Values of Costs and Monetized Benefits:
Comparison of Proposed, Alternative, and Final
Schedules for 1985-1987 (millions of 1983 dollars)
With No Misfueling
Costs
Benefits
Net benefits
Net benefits, excluding
blood pressure
With Full Misfueling
Costs
Benefits
Net benefits
Net benefits, excluding
blood pressure
With Partial Misfueling
Costs
Benefits
Net benefits
Proposed
1,009
13,482
12,473
2,368
1 ,048
12,506
11.459
1,625
1,014
13,287
12,273
Alternative
845
14,377
13,532
2,743
774
12,160
11,386
1,547
793
12,932
12,139
Final
1,130
16,095
14,965
2,924
1,139
14,490
13,351
1,876
1,105
15,271
14,166
Net benefits, excluding
blood pressure 2,222 1,985 2,473
-------
VIII-22
net benefits, in excess of $11 billion with blood-pressure-
related benefits and over $1.5 billion without them. Compared
to the other schedules, the final rule has higher costs but even
higher benefits (whether or not blood-pressure-related benefits
are included), with the result that it has the highest net bene-
fits of the three schedules.
Tables VIII-7a through VIII-7c present year-by-year estimates
of the costs and benefits of the final rule over the period 1985
to 1992, under the three assumptions about misfueling. The costs
fall from 1986 to 1992 because we project that the demand for
leaded gasoline would fall even in the absence of the rule, as a
result of retirement of older cars. For that same reason, most
of the estimated annual benefits also decline over time. The
major exception is conventional pollutants, because we expect the
amount of misfueling to increase in the absence of the rule, as
the number and average age of catalyst-equipped vehicles increase.
Table VIII-8 shows the present values of the final rule
under the different assumptions about misfueling. The estimated
net benefits, not including blood-pressure-related benefits, range
from $4.1 billion if the rule has no impact on misfueling, to
$6.7 billion if the rule eliminates all misfueling. Under the
more realistic "partial misfueling" assumption, the present value
of the net benefits is $5.9 billion. If the adult blood pressure
benefits are included, the present value of the net benefits is
much higher, $33.4 billion under the partial misfueling assumption,
-------
VIII-23
TABLE VIII-7a. Year-by-Year Costs and Monetized Benefits of Final Rule, Assuming
No Misfueling (millions of 1983 dollars)
1985 1986 1987 1988 1989 1990 1991 1992
MONETIZED BENEFITS
Children's
health effects 251 602 550 504 455 417 371 361
Adult blood
pressure 2,033 5,927 5,707 5,484 5,227 5,008 4,722 4,736
Conventional
pollutants 140 278 278 280 282 288 299 310
Maintenance 252 933 880 840 811 792 780 776
Fuel economy 68 190 175 124 144 145 182 175
TOTAL MONETIZED
BENEFITS 2,744 7,930 7,590 7,232 6,919 6,649 6,354 6,358
TOTAL REFINING
COSTS 127 607 553 530 502 468 442 440
NET BENEFITS 2,617 7,323 7,037 6,702 6,417 6,181 5,912 5,918
NET BENEFITS
EXCLUDING BLOOD
PRESSURE 584 1,396 1,330 1,218 1,190 1,174 1,190 1,182
-------
VIII-24
TABLE VIII-7b. Year-by-Year Costs and Monetized Benefits of Final Rule, Assuming
Full Misfueling (millions of 1983 dollars)
1985 1986 1987 1988 1989 1990 1991 1992
MONETIZED BENEFITS
Children's
health effects 223 592 539 494 445 406 361 350
Adult blood
pressure 1,724 5,778 5,543 5,303 5,031 4,798 4,521 4,512
Conventional
pollutants
Maintenance 102 838 777 730 694 668 650 640
Fuel economy 35 177 150 70 96 113 131 122
TOTAL MONETIZED
BENEFITS 2,084 7,385 7,009 6,597 6,266 5,985 5,663 5,624
TOTAL REFINING
COSTS 96 627 578 539 514 485 451 443
NET BENEFITS 1,988 6,758 6,431 6,058 5,752 5,500 5,212 5,181
NET BENEFITS
EXCLUDING BLOOD
PRESSURE 264 980 888 755 721 702 691 669
-------
VIII-25
TABLE VIII-7c. Year-by-Year Costs and Monetized Benefits of Final Rule, Assuming
Partial Misfueling (millions of 1983 dollars)
1985 1986 1987 1988 1989 1990 1991 1992
MONETIZED BENEFITS
Children's
health effects 223 600 547 502 453 414 369 358
Adult blood
pressure 1,724 5,897 5,675 5,447 5,187 4,966 4,682 4,691
Conventional
pollutants 0 222 222 224 226 230 239 248
Maintenance 102 914 859 818 788 767 754 749
Fuel economy 35 187 170 113 134 139 172 164
TOTAL MONETIZED
BENEFITS 2,084 7,821 7,474 7,105 6,788 6,517 6,216 6,211
TOTAL REFINING
COSTS 96 608 558 532 504 471 444 441
NET BENEFITS 1,988 7,213 6,916 6,573 6,284 6,045 5,772 5,770
NET BENEFITS
EXCLUDING BLOOD
PRESSURE 264 1,316 1,241 1,125 1,096 1,079 1,090 1,079
-------
VIII-26
TABLE VIII-8. Present Values of Costs and Benefits of Final
Rule, 1985-1992 (millions of 1983 dollars)
No Full Partial
Misfueling Misfueling Misfueling
MONETIZED BENEFITS
Children's
health effects 2,582 2,506 2,546
Adult blood
pressure 27,936 26,743 27,462
Conventional
pollutants 1,525 0 1,114
Maintenance 4,331 3,634 4,077
Fuel
economy 856 643 788
TOTAL
MONETIZED
BENEFITS 37,231 33,526 35,987
TOTAL REFINING
COSTS 2,637 2,678 2,619
NET BENEFITS 34,594 30,847 33,368
NET BENEFITS EXCLUDING
BLOOD PRESSURE 6,658 4,105 5,906
-------
VIII-27
VIII.C. Impact of Banking on Costs and Benefits of Final Rule
The analysis thus far in this chapter has assumed that
refineries will follow the phasedown schedule being promulgated
in this final rule. EPA has proposed the use of "banking" of
lead rights during 1985 for use in 1986 and 1987. Under banking,
refineries would have the option of using less lead than allowed
in 1985 and "banking" it for use in 1986 or 1987. As discussed
in Chapter II, this provision would give individual refineries
extra flexibility in reducing lead, without increasing the total
allowable amount of lead in gasoline between 1985 and 1987.
Although refiners would be under no obligation to use the right
to bank, EPA expects that most would, because the marginal value
of lead to refiners will be higher in 1986 and 1987 than in 1985,
for the reasons discussed in Chapter II. This section briefly
examines the impact that banking may have on the costs and bene-
fits of the phasedown rule. This analysis is for illustrative
purposes only, as EPA has determined that the final rule is
feasible and not unduly expensive without banking, and should
be promulgated independent of the banking rule.
Table VIII-9 compares the schedule without banking to the
two possible alternatives with banking that were presented in
Chapter II. Both alternatives are based on the partial misfuel-
ing assumption; i.e., that misfueling continues unabated through
1985 and is reduced by 80 percent in 1986 and 1987.
Alternative 1 assumes that refiners would not start banking
until the second quarter of 1985, at which point they would use
an average of 0.60 gplg, thus banking 0.50 grams, on average, for
-------
VIII-28
TABLE VIII-9. Alternative Phasedown Patterns with Banking (gplg)
1985(by quarter 1986 1987
Alternative I II III-IV
Without Banking 1.10 1.10 0.50 0.10 0.10
With Banking
Alternative 1 1.10 0.60 0.40 0.25 0.19
Alternative 2 0.80 0.60 0.45 0.30 0.21
-------
VIII-29
each gallon of leaded produced.* Under alternative 1, we also
assume some banking in the last half of the year, with leaded
gasoline averaging 0.40 gplg, slightly below the limit of 0.50
gplg. A total of 7.0 million grams of lead (about 22 percent
of the total allowed in 1985) would be banked during 1985, allow-
ing refiners to average 0.25 gplg in 1986 and 0.19 gplg in 1987-
Note that shaving 0.10 gplg from the annual average in 1985
translates into a larger per gallon increase in 1986 or 1987,
because the amount of leaded gasoline produced in the later years
is smaller; the total amount of lead use allowed over the three
years, however, would be the same as without banking.
Alternative 2 assumes that some refiners would be able to
reduce lead more quickly, so that banking would begin in the
first quarter of 1985. Compared to Alternative 1, those extra
banked rights from the first quarter would be used to reduce
banking slightly in the last half of 1985 and to achieve slightly
higher lead levels in 1986 and 1987. The amount banked would be
9.1 million grams. Again, the total amount of lead used would
be the same as without banking.
Table VIII-10 presents the year-by-year cost and benefit
estimates for the two alternatives with banking. Compared to
the estimates without banking (see Table VIH-7c), the benefits
and costs would be higher in 1985, but lower in 1986 and 1987.
* Note that in California, where a state-imposed rule will limit
leaded gasoline to 0.8 gplg in the first half of 1985, refiners
producing 0.60 gplg leaded gasoline would only be able to bank
0.20 gplg in the second quarter; refiners would not be allowed
to take credit for reductions mandated by state laws.
-------
VIII-30
TABLE VIII-10. Costs and Monetized Benefits of Alternative (Banking) Phasedown
Patterns, Assuming Partial Misfueling (millions of 1983 dollars)
Alternative 1*
1985
MONETIZED BENEFITS
Children's
health effects 347
Adult blood
pressure 2,745
Conventional
pollutants 0
Maintenance 110
Fuel economy 95
TOTAL MONETIZED
BENEFITS 3,297
TOTAL REFINING
COSTS 176
NET BENEFITS 3,121
NET BENEFITS
EXCLUDING BLOOD
PRESSURE 376
1986
550
5,122
222
716
123
6,733
420
6,313
1,191
1987
522
5,244
222
752
126
6,866
463
6,403
1,159
Alternative 2*
1985
395
3,008
0
97
87
3,587
170
3,417
409
1986
532
4,870
222
661
114
6,399
378
6,021
1,151
1987
515
5,162
222
732
126
6,757
452
6,305
1,143
*See Table VIII-9 for description of alternatives.
-------
VIII-31
Table VIII-11 compares the present values of the costs and bene-
fits with banking to those without. In present value terms,
banking reduces refining costs by $173 million under alternative
1 and $226 million under alternative 2, a reduction of over 20
percent. As discussed in Chapter II, this is likely to be an
underestimate of the real savings that can be achieved with bank-
ing, as it does not account for the extra flexibility banking
allows in meeting unexpected problems. It is also interesting
to note that the present value of the cost of the final rule with
banking is actually lower than the cost of the August proposed
rule without banking (cf. Table VIII-6), despite the fact that
the final rule eliminates significantly more lead.
Banking has a slight negative impact on the present value of
estimated benefits. This reduction reflects several factors. The
most important of these is that our estimates of the relationship
between blood pressure and blood lead employs the logarithm of
blood lead; thus it predicts slightly higher benefits from concen-
trating the reduction in lead in 1986 and 1987 rather than
spreading it over 1985-1987. The same holds true for mainten-
ance benefits, because exhaust system corrosion appears to fall
most sharply as lead is reduced at low levels. In contrast,
children's health benefits are higher with banking than without,
primarily because the lead reductions are achieved earlier with
banking.
Overall, banking has virtually no impact on net benefits.
The small differences shown in Table VIII-11 are well within the
"noise" of the estimates and should be regarded as insignificant.
-------
VIII-32
TABLE VIII-11. Present Values of Costs and Benefits of
Alternative Phasedown Patterns, 1985-87,
Assuming Partial Misfueling
(millions of 1983 dollars)
MONETIZED BENEFITS
Children's
health effects
Adult blood
pressure
Conventional
pollutants
Maintenance
Fuel
economy
TOTAL
MONETIZED
BENEFITS
Without
Banking
1,210
11,693
385
1,638
344
15,271
With
Alt. 1
1,262
11,605
385
1,377
307
14,936
Banking
Alt. 2
1,285
11,558
385
1,298
291
14,818
TOTAL REFINING
COSTS 1,105 932 879
NET BENEFITS 14,166 14,007 13,939
NET BENEFITS EXCLUDING
BLOOD PRESSURE 2,473 2,399 2,381
-------
VIII-33
Coupled with the important, but nonmonetized, cost saving that
banking provides in the form of flexibility to meet unexpected
refining problems, these results indicate that banking is a
desirable policy.
VIII.D. Summary
The results presented in this chapter indicate that the
final phasedown rule being promulgated has the highest net
benefits of the alternatives considered. This conclusion holds
whether or not the recently developed estimates of blood-
pressure-related benefits are included, and whether or not it is
assumed that the rule will eliminate misfueling. Although many
of the estimates are subject to uncertainty, the magnitude of the
estimated monetized benefits relative to the costs indicates that
this conclusion is very robust. Moreover, the monetary benefit
estimates represent an incomplete tabulation of the benefits
likely to result from the rule; in short, the benefits are
under-estimated.
Three limitations deserve particular notice. First, we have
not estimated any benefits for children at blood lead levels
below the CDC cutoff of 25 ug/dl, and our monetized estimates for
children above that level cover only medical and compensatory
education costs. Second, the direct estimates of benefits
associated with reduced conventional pollutants omit several
important categories, including benefits associated with ozone's
effects on nonagricultural vegetation, chronic health effects
related to ozone, and the effects of nitrogen oxides and carbon
-------
VIII-34
monoxide on health. Last, the estimates of adult health benefits
cover only blood-pressure-related effects, and only males aged
40-59. In the case of myocardial infarctions, strokes, and
deaths, only whites males in this age range are included. (This
is because of limited data on the cardiovascular risks associated
with high blood pressure in nonwhites.) These unquantified bene-
fits add further strength to the conclusion that rapid reductions
of lead in gasoline are amply justified.
-------
REFERENCES
Abell RF. Operation and application of hydraulic valve lifters.
SAE Paper #690347, Society of Automotive Engineers, Warrendale,
PA 1969.
Adams R, McCarl B. Assessing the benefits of alternative oxidant
standards on agriculture: the role of response information.
Prepared for U.S. EPA, 1983 September.
Adams R, et al. Economic effects of ozone on agriculture.
Final Report to U.S. EPA, 1984.
Adams WC, Schelegle ES. Ozone toxicity effects consequent to
prolonged, high intensity exercise in trained endurance
athletes. J App Physiol Respir Environ Exercise Physiol
1983; 55: 805-12.
Agerty HA. Lead poisoning in children. Med Clin North Am 1952; 36:
1587-97.
American Petroleum Institute. Unpublished Carcinogenicity Study
on Unleaded Gasoline in Vapor in Fischer 344 Rats and B6C3F1
Mice. 1982.
Amoco, submission to EPA Central Docket EN-84-05.
Anderson E, et al. Effect of low-level carbon monoxide exposure
on onset and duration of angina pectoris: a study on 10 patients
with ischemic heart disease. Annals of Internal Medicine 1973;
79: 46-50.
Anderson G, et al. Development and application of methods for
estimating: effects of industrial emission controls on air
quality impact of reactive pollutants. U.S. EPA 1984 July.
Angle CR, Mclntire MS, Swanson MS, Stohs SJ. Erythrocyte
nucleotides in children - increased blood lead and
cytidine triphosphate. Pediatric Research 1982; 16: 331-4.
Annest JL, Pirkle JL, Makuc D, Neese JW, Bayse DD, Kovar MG.
Chronological trends in blood lead levels between 1976 and
1980. New England Journal of Medicine 1983; 308: 1373-7.
Armstrong JA. Investigations into the emissions effects of
vehicle misfueling. Presented to: The North American Motor
Vichicle Emission Control Conference. New York, 1984 April
2.
Aub JC, Fairhall Lt, Minot AS, Reznikoff P, Hamilton A. Lead
Poisoning in Baltimore, MD. The Williams and Wilkins Company
(Medicine monographs: V.7.)
-------
-2-
The Australian Therapeutic Trial in Mild Hypertension:
Report by the Management Committee. Lancet 1980; 2: 1261-7.
Barrett LB, Waddell T. The Cost of Air Pollution Damages: A
Status Report. U.S. EPA; Research Triangle Park, NC 1973
Barton JC, Conrad ME, Nuby S, Harrison L. Effects of
iron on the absorption and retention of lead. Journal of
Laboratory and Clinical Medicine 1978; 92: 536-47.
Bascunana J, Stahman R. Impact of gasoline characteristics on
fuel economy and its measurement. Report 76-10, Technology
Assessment and Evaluation Branch, ECTD, EPA, 1976 December.
Bates D, Sizto R. Relationship between air pollution levels and
hospital admissions in southern Ontario. Canadian J of Pub
Health 1983 March/April; 79-
Batuman V, Landy E, Maesaka JK, Wedeen RP. Contribution of lead to
hypertension with renal impariment. N Engl J Med 1983; 309:
17-21.
Beevers DG, Erskine E, Robertson M, Beattie AD, Campbell
BC, Goldberg A, Moore MR, Hawthorne VM. Blood-lead and
hypertension. Lancet 1976; 2(7975): 1-3.
Beevers DG, Cruickshank JK, Yeoman WB, Carter GF, Goldberg A,
Moore MR. Blood-lead and cadmium in human hypertension.
J Environ Pathol Toxicol 1980; 4: 251-60.
Bell AG, Keene JA, Reders K. Road antiknock performance of low-
lead and nonleaded gasoline in european cars. SAE Paper #710625,
Society of Automotive Engineers, Warrendale, PA 1971.
Bellinger D, Needleman H, Leviton A, Waternaux C, Rabinowitz M,
Nichols M. Early sensory-motor development and prenatal exposure
to lead. Neurobehavioral Toxicology and Teratology 1984; 6:
387-402.
Benignus VA, Otto DA, Muller KE, Seiple KJ. Effects of age and
body lead burden on CNS function in young children: II.
EEG Spectra. Electroencephalog Clin Neurophysiol 1981;
52:240-8.
Bennett PA. Engine rusting in service and dynamometer tests.
Lubrication Engineering 1960 November: 529.
Benson JD. Some factors which affect octane requirement increases.
SAE Paper #750933, Society of Automotive Engineers, Warrendale,
PA 1975.
-------
-3-
Benson JD. Manganese fuel additive (MMT) can cause vehicle prob-
lems. SAE Paper #770655, Society of Automotive Engineers,
Warrendale, PA 1977.
Bernard JP. Short-trip engine oil rust testing. SAE Paper #710839,
Society of Automotive Engineers, Warrendale, PA 1971.
Bert JA, Gething JA, Hansel TJ, Newhall HK, Peyla RJ, Voss PA.
Gasoline additive concentrate removes combustion chamber depos-
its and reduces vehicle octane requirement. SAE Paper #831709,
Society of Automotive Engineers, Warrendale, PA 1983.
Betts PR, Astley R, Raine DN. Lead intoxication in children in
Birmingham. Br Med J 1973; 1(5850): 402-6.
Bhalla AK, Amento EP, Clemens T, Holick MF, Krane SM. Specific
high-affinity receptors for 1,25-dihydroxyvitamin 03 in human
peripheral blood mononuclear cells: presence in monocytes and
induction in T lymphocytes following activation. J Clin
Endocrinol Metab 1983; 57: 1308-10.
Bigley HA Jr, Keller BD, Kloppe MG. Octane requirement increase in
1971 model cars with leaded and unleaded gasolines. SAE Paper
#710675, Society of Automotive Engineers, Warrendale, PA 1971.
Bigley HA Jr, Benson JD. Octane requirement increases in 1971 model
cars - with and without lead. SAE Paper #730013, Society of
Automotive Engineers, Warrendale, PA 1973.
Billick IH, Curran AS, Shier DR. Analysis of pediatric blood
lead levels in New York City for 1970-1976. Environmental
Health Perspectives 1979; 31:183-90.
Billick IH, et al. Predictions of pediatric blood lead levels
from gasoline consumption. U.S. Department of Housing and
Urban Development 1982.
Blank FM, et al. Valuation of aesthetic preferences: A case
study of the economic value of visibility. Draft report
to the Electric Power Research Institute, Palo Alto, Ca 1977.
Blomquist G. Valuation of Life: Implications of Automobile
Seat Belt Use. Ph.D. Dissertation. University of Chicago, 1977.
Brennan MJW, Cantrill RC. y-Aminolevulinic Acid is a potent
agonist for GABA auto receptors. Nature 1979; 280: 514-5.
Brookshire DS, et al. The valuation of aesthetic preferences.
J Environ Econ Manag 1976; 3: 325-46.
Brown C. Equalizing differences in the labor market.
Quarterly Journal of Economics 1980; 94.
-------
-4-
Bryant EC. Statistical Analyses (2nd Edition) New York:
McGraw Hill 1966.
Bryce-Smith D, Deshpunde RR, Hughes J, Waldron HA. Lead and
cadmium levels in still births. Lancet 1977; 1: 1159.
Bull RJ, Stanaszek PM, O'Neill JJ, Lutkenhoff SD. Specificity of
the effects of lead on brain energy metabolism for substrates
donating a cytoplasmic reducing equivalent. Environ Health
Perspect 1975; 12:89-95.
Bull RJ. Effects of trace metals and their derivatives on the
control of brain energy metabolism. Lee SD ed. Biochemical
Effects of Environmental Pollutants. Ann Arbor, Mi: Ann Arbor
Science 1977: 425-40.
Bull RJ, Lutkenhoff SD, McCarty GE, Miller RG. Delays in the
postnatal increase of cerebral cytochrome concentrations in
lead-exposed rats. Neuropharmacology 1979; 18: 83-92.
Bull RJ. Lead and energy metabolism. Singhal PL, Thomas JA eds.
Lead Toxicity. Baltimore: Urban and Schwarzenberg, Inc 1980:
119-68.
Bull RJ, McCauley DT, Tayler DH, Crofton KM. The effects of
lead on the developing central nervous system of the rat.
Neurotoxicology 1983; 4(1): 1-17.
Bureau of Food and Drug Administration Compliance Program Report
of Findings, Total Diet Studies, 1973-1980 reports.
Byers RK. Lead poisoning, review of the literature and report
on forty-five cases. Pediatr 1959; 23:585-603.
Caracciolo F, Sperot JA. The effects of phosphorus-containing
engine oil additives on exhaust oxidation catalyst degradation.
SAE Paper #760562, Society of Automotive Engineers, Warrendale,
PA 1976.
Carey LR, Stover WH, Murray DW. Extended drain passenger car
engine oils. SAE Paper #780952, Society of Automotive Engineers,
Warrendale, PA 1978.
Champion Spark Plug Co. Champion Ingnition and Engine Performance
Conferences. Volumes 1971-1976.
Chance RL, Ceselli RG. Corrosiveness of exhaust gas condensates.
SAE Paper #830585, Society of Automotive Engineers, Warrendale,
PA 1983.
-------
-5-
Cherrie JM. Factors influencing valve temperature in passenger-car
engines. SAE Paper #650484, Society of Automotive Engineers,
Warrendale, PA 1965.
Chesney RW, Rosen JF, DeLuca HF. Disorders of calcium metabolism
in children. Chiumello, Sperling M eds. Recent Progress in
Pediatric Endocrinology. New York, NY: Raven Press 1983: 5-24.
Children living in the vicinity of a primary lead smelter.
Environmental Research 1980; 22: 81-94.
Chisolm JJ Jr, Harrison HE. The exposure of children to lead.
Pediatrics 1956; 18: 943-58.
Chisolm JJ Jr. The use of chelating agent in the treatment of acute
and chronic lead intoxication in children. J Pediatr (St. Louis)
1968; 73: 1-38.
Chisolm JJ Jr, Barltrop D. Recognition and management of children
with increased lead absorption. Arch Dis Child 1979; 54: 249-62.
Chock D, et al. Effect of NOX emission rates on smog formation in
the California south coast air basin. Environmental Science
and Technology 1981 August; 15(8).
Cohen C, et al. Respiratory symptoms, spirometry, and oxidant air
pollution in nonsmoking adults. Amer Rev Resp Disease 1972;
105: 251-61.
Cohen GJ, Ahrens WE. Chronic lead poisoning: a review of seven
years' experience at the Children's Hospital, District of Colum-
bia. J Pediatr (St Louis) 1959; 54: 271-84.
Colyer C, Harding K. The 1980s — a decade of challenge for engine
oils. SAE Paper # 811410, Society of Automotive Engineers,
Warrendale, PA 1981.
Cooper WC, Gaffey WR. Mortality of lead workers. Cole JF ed.
Proceedings of the 1974 Conference on Standards of Occupational
Lead Exposure; February 1974; Washington, DC. J Occup Med 1975;
17: 100-7.
Cooper WC. Mortality in employees of lead production facilities
and lead battery plants, 1971-1975. Lynam Dr, Piantanida LG,
Cole JF Eds. Environmental Lead; Proceedings of the second
international symposium on environmental lead research;
Cincinnati, OH, 1978 December. New York: Academic Press 1981.
Cordera FJ, et al. TEL scavengers in fuel affect engine performance
and durability. SAE Paper #877A, Society of Automotive Engineers,
Warrendale, PA 1964.
-------
-6-
Council on Environmental Quality. Environmental Quality - 1980.
Government Printing Office, 1980 December.
Graver RJ, Podiak RS, Miller RD. Spark-plug design factors and
their effect on engine performance. SAE Paper #700081, Society
of Automotive Engineers, Warrendale, PA 1970.
Crocker T, Vaux H. Some economic consequences of ambient oxidant
impacts on a natural forest. Prepared for U.S. EPA, Office of
Policy Analysis, 1983 August.
Crofton KM, Taylor DH, Bull RJ, Snulka DJ, Lutkenhoff SD. Develop-
mental delays in exploration and locomotor activity in male rats
exposed to low level lead. Life Science 1980; 26: 823-31.
Grouse WW, Johnson RH, Reiland WH. Effect of unleaded fuel on
lubricant performance. SAE Paper #710584, Society of Automotive
Engineers, Warrendale, PA 1971.
Danis LJ. Engine valve cooling. SAE Paper #730055, Society of
Automotive Engineers, Warrendale, PA 1973.
De la Burde B, Choate MS Jr. Does asymptomatic lead exposure in
children have latent sequelae. Journal of Pediatrics 1972;
81: 1088-91.
De la Burde B, Choate MS Jr. Early asymptomatic lead exposure
and development at school age. Journal of Pediatrics 1975;
87: 638-42.
Delves HT. A micro-sampling method for the rapid determination of
lead in blood by atomic absorbtion spectrophotometry. Analyst
1970: 431-3.
DePalma TV, Bailey CH. Application of catalytic converters to
heavy-duty gasoline powered trucks. SAE Paper #750902, Society
of Automotive Engineers, Warrendale, PA 1975.
De Para M, Belvoir Research and Development Center, U.S. Army.
personal communication.
Detels R, et al. The UCLA population studies of chronic obstructive
respiratory disease. American Journal of Epidemiology 1979; 109-
Dingwall-Fordyce I, Lane RE. A follow-up study of lead workers.
Br J Ind Med 1963; 20: 313-5.
Doelling RP. Engine's definition of unleaded gasoline. SAE Paper
8401, Society of Automotive Engineers, Warrendale, PA 1971.
Dresner DL, Ibrahim NG, Mascarenhas BR, Levere RD. Modulation of
bone marrow heme and protein synthesis by trace elements. Environ
Res 1982; 28: 55-66.
-------
-7-
DRI. The Data Resources Inc review of the U.S. economy. Various
issues, 1983 and 1984. DRI, Lexington, MA.
Drinkwater B, et al. Air pollution, exercise, and heat stress.
Archives of Environmental Health 1974; 28: 177-81.
Durand D. Stable Choas. Morristown, NJ: General Learning
Corporation 1971.
Eccleston BH, Hurn RW. Comparative emissions from some leaded and
prototype lead-free automobile fuels. Report of Investigation
#7390, U.S. Department of the Interior, Bureau of Mines
Petroleum Research Center, Bartlesville, OK 1970.
Ellis JC. Future automotive fuels. SAE Paper #720617, Society of
Automotive Engineers, Warrendale, PA 1972.
Ellis JC. Gasoline for low-emission vehicles. SAE Paper #730616,
Society of Automotive Engineers, Warrendale, PA 1973.
Energy and Environmental Analysis, Inc. Historical and projected
emissions conversion factor and fuel economy for heavey-duty
trucks 1962-2002. Motor Vehicle Manufacturers Association,
Detroit, Michigan 1983.
Ennis JM, Harrison HE. Treatment of lead encephalopathy with BAL
(2,3-dimercaptopropanol). Pediatrics 1950; 5: 853-68.
Erickson MM, Pokus A, Canter CE, Dickson AW, Hillman LS. Tissue
mineral levels in victims of sudden infant death syndrome. I.
toxic metals - lead and cadmium. Pediatric Research 1983;
17: 784-99.
ETA Engineering, Inc. Assessment of benefits from new source
performance standards for volatile organic compounds. Prepared
for U.S. EPA, Office of Policy Analysis, 1983 September.
Ethyl Corporation. Monthly Retail Gasoline Sales Bulletins,
1976-1982.
Evans, et al. Ozone measurement from a network of remote sites.
Journal of the Air Pollution Control Association 1983 April;
32(4).
Expert Committee on Trace Metal Essentiality. Independent peer
review of selected studies by Drs. Irchgessner and Reichlmayr-
Lias concerning the possible nutritional essentiality of lead:
official report of findings and recommendations of an interdis-
ciplinary expert review committee, 1983. Available for inspection
at: U.S. Environmental Protection Agency, Environmental Criteria
and Assessment Office, Research Triangle Park, NC.
-------
-8-
Exxon Memo, Re: Gulf/East Coast Gasoline. 1978 January.
Fachetti S, Geiss F. Isotopic lead experiment status report.
Luxembourgh: Commission of the European Communities 1982;
Publication No. EUR8352EN.
Faggan JE, et al. An evaluation of manganese as an antiknock in
unleaded gasoline. SAE Paper 1750925, Society of Automotive
Engineers, Warrendale, PA 1975.
Fahim MS, Fahim Z, Hall DG. Effects of subtoxic lead levels on
pregnant women in the state of Missouri. Res Commun Chemical
Pathol Pharmacol 1976; 13(2): 309.
Faucett Associates, Draft Report: Review and Critique of Previous
OMSAPC Cost-Effectiveness Analysis, 1983 March.
Felt AE, Kerley RV. Engines and effects of lead-free gasoline.
SAE Paper #710367, Society of Automotive Engineers, Warrendale,
PA 1971.
Ferris B Jr. Health effects of exposure to low levels of regulated
air pollutants. Journal of the Air Pollution Control Association
1978 May; 28(5) .
Fisher RA. Statistical Methods for Research Workers. New York:
Hafner Press 1970.
Florey C du V, et al. The relation between respiratory illness
in primary schoolchildren and the use of gas for cooking. III.
nitrogen dioxide, respiratory illness and lung infection.
Int J Epidemiol 1979; 8: 347-53.
Folinsbee LJ, et al. Effect of 0.62 ppm N02 on cardiopulmonary
function in young male non-smokers. Environmental Research
1978; 15: 199-205.
Folinsbee LJ, Bedi JF, Horvath SM. Pulmonary function changes
in trained athletes following 1-hour continuous heavy exercise
while breathing 0.21 ppm ozone. J Appl Physiol Respir Environ
Exercise Physiol (in press) 1984.
Ford Automotive Emissions Office, personal communication.
Forthofer RN. Investigation of non-response bias in NHANES
II. American Journal Of Epidemiology (Bult.) 1983
Fouts PJ, Page IH. The effect of chronic lead poisoning on arterial
blood pressure in dogs. Am Heart J 1942; 24: 329-31.
Fouts PJ, Page IH. The effect of chronic lead poisoning on
arterial blood pressure in dogs. American Heart Journal 1942;
24: 329-31.
-------
-9-
Fowler BA. General subcellular effects of lead, mercury, cadmium,
and arsenic. Environ Health Perspect 1978; 22: 37-41.
Fowler BAf Kimmel CA, Woods JS, McConnell EE, Grant LD. Chronic
low-level lead toxicity in the rat: III. an integrated assessment
of long-term toxicity with special reference to the kidney.
Toxicol Appl Pharmacol 1980; 56: 59-77.
Freas W. A digital solution to city-specific ekma isopleth diagrams
in ozone regulatory analysis. U.S. EPA, Research Triangle Park
1983 July.
Freeman AM III. Air and Water Pollution Control: A Benefit-
Cost Assessment. New York: John Wiley and Sons 1982.
Freeman R. Reversible myocarditis due to chronic lead poisoning in
childhood. Arch Dis Child 1965; 40: 389-93.
Friedlander SK. Chairman, Clean Air Scientific Advisory Committee
(CASAC). Review and Closure of the OAQPS Staff Paper for Nitro-
gen Oxides, 1982 July 6.
Fuchs EJ. Unleaded versus leaded fuel results in laboratory engine
tests. SAE Paper #710676, Society of Automotive Engineers,
Warrendale, PA 1971.
Gagliardi JC. Effect of fuel antiknock compounds and deposits on
exhaust emissions. SAE Paper #670128, Society of Automotive
Engineers, Warrendale, PA 1967.
Gagliardi JC, Ghannam F. Effects of tetraethyl lead concentration
on exhaust emissions in customer-type vehicle operation. SAE
Paper #690015, Society of Automotive Engineers, Warrendale, PA
1969.
Gallopoulos NE. Projected lubricant requirement of engines oper-
ating with lead-free gasoline. SAE Paper #710585, Society of
Automotive Engineers, Warrendale, PA 1971.
General Motors Corporation. General Motors Corporation comments
on EPA report 'Costs and Benefits of Reducing Lead in Gasoline1
letter from T.M. Fisher (GM) to R.D. Morgenstern (EPA/OPA), 1984
June 18.
Gergel WC, Sheahan TJ. Maximizing petroleum utilization through
extension of passenger car oil drain periods - what's required?
SAE Paper #760560, Society of Automotive Engineers, Warrendale,
PA 1976.
Gerking S, Stanley L, Weirick W. An economic analysis of air
pollution and health: the case of St. Louis. Prepared for
U.S. EPA, Office of Policy Analysis, 1983 July.
-------
-10-
Giles WS. Valve problems with lead free gasoline. SAE Paper
#710368, Society of Automotive Engineers, Warrendale, PA 1971.
Giles WS, Updike SH. Influence of low lead fuels on exhaust
valve performance. SAE Paper #710674, Society of Automotive
Engineers, Warrendale, PA 1971.
Glasson WA, Tuesday CS. Environ Sci Technol 1970; 4: 37.
Glasson W. Effect of Hydrocarbons and NOX on photochemical smog
formation under simulated transport conditions. Journal of the
Air Pollution Control Association 1981 November 11; 31(11).
Godfrey D, Courtney RL. Investigation of the mechanism of exhaust'
valve seat wear in engines run on unleaded gasoline. SAE Paper
#710356, Society of Automotive Engineers, Warrendale, PA 1971.
Goldberg AM, Meredith DA, Miller S, Moore MR, Thompson CG. Hepatic
drug metabolism and heme biosynthesis in lead-poisoned rats.
British Journal of Pharmacology 1978; 62: 529-36.
Goldstein BD. The relation between respiratory illness in school-
children and the use of gas for cooking. II. Factors affecting
nitrogen dioxide levels in the home. Int J Epidemiol 1979;
8: 339-45.
Goldstein E, et al. Fate and distribution of inhaled nitrogen
dioxide in rhesus monkey. Am Rev Respir Dis 1977; 115: 403-12.
Goldstein E, Moderator. Photo Chemical Air Pollution. Inter-
departmental Conference sponsored by the Department of Medicine,
University of California, School of Medicine; Davis, California
1982.
Goth G. Recent developments in hardfacing alloys for internal
combustion engines. SAE Paper #831287 in State-of-the-Art on
Design and Performance of Diesel Power Cylinder Components,
SP-552, Society of Automotive Engineers, Warrendale, PA 1983.
Goyer RA, Rhyne BC. Pathological effects of lead. Internal Rev
Exp Pathol 1973; 12: 1-77.
Gray DS, Azhari AG. Saving maintenance dollars with lead-free
gasoline. SAE Paper #720084, Society of Automotive Engineers,
Warrendale, PA 1972.
Green AES, et al. An interdisciplinary study of the health, social,
and environmental economics of the sulfur oxide pollution in
Florida. Interdisciplinary Center for Aeronomy and (other)
Atmospheric Sciences, Univ of Florida; Gainesville, FL 1978.
-------
-11-
Gross-Selbeck E, Gross-Selbeck M. Changes in operant behavior of
rats exposed to lead at the accepted no-effect level. Clinical
Toxicology 1981; 18: 1247-56.
Gunter EW, et al. Laboratory Procedures Used by the Clinical
Chemistry Division, Centers for Disease Control for the Second
National Health and Nutrition Examination Survey. Atlanta
Centers for Disease Control 1981.
Hackney JDr et al. Experimental studies on human health effects
on air pollutants. II. Four hour exposure to ozone alone and
in combination with other pollutants. Arch Environ Health
1975a; 33: 379-84.
Hackney JD, et al. Experimental studies on human health effects
on air pollutants. III. Two hour exposure to ozone alone and
in combination with other gases. Arch Environ Health 1975b;
30: 176-81.
Hackney JD. Effects of atmospheric pollutants on human physiologic
function. Final report, U.S. EPA 1976.
Hackney JD, et al. Experimental studies on human health effects
of air pollutants. IV. Short-term physiological and clinical
effects of nitrogen dioxide exposure. Arch Environ Health 1978;
33: 176-81.
Hall CA, Felt AE, Brown WJ. Evaluating effects of fuel factors on
stabilized exhaust emissions levels. SAE Paper # 690014, Society
of Automotive Engineers, Warrendale, PA 1969.
Hammer D, et al. Los Angeles student nurse study. Daily symptom
reporting and photochemical oxidants. Archives of Environmental
Health 1974; 28.
Hare CT, Springer KJ, Huls TA. Exhaust emissions from farm
construction, and industrial engines and their impact. SAE Paper
#750788, Society of Automotive Engineers, Warrendale, PA 1975.
Harlan WR, Landis JR, Schmouder RL, et al. Relationship of
blood lead and blood pressure in the adolescent and adult
U.S. population. JAMA 1985 January 25.
Harrison. Principles of Internal Medicine. Ninth Edition.
Hartunian NS, Smart CN, Thompson MS. The Incidence and Economic
Costs of Major Health Impariments. Lexington, MA: Lexington
Books 1981.
-------
-12-
Harvey P, Hamlin M, Kumar R. The Birmingham blood lead study.
Presented at: annual conference of the British Psychological
Society, Symposium on lead and health, 1983. Available for
inspection at: U.S. Environmental Protection Agency, Environ-
mental Criteria and Assessment Office, Research Triangle Park,
N.C.
Hasselblad V. Modeling dose response relations for health effects
data. Environmetrics 81: Selected Papers, SIAM, Philadelphia,
1981.
Hasselblad V, Svendsgaard D. Reanalysis of the Los Angeles student
nurse study. Unpublished paper, U.S. EPA, Health Effects Research
Lab, Research Triangle Park, NC 1975 August.
Heck W, et al. Nitrogen dioxide: time-concentration model to
predict acute foliar injury. U.S. EPA, Corvallis, Oregon.
Pub. No. 600/3-79-057, 1979-
Heck W, et al. A reassessment of crop loss from ozone.
Environmental Science and Technology 1983; 17(12).
Heintz HT, Hershaft A, Horak GC. National Damages of Air and
Water Pollution. Report submitted to U.S. EPA 1976.
Hickling Partners, Inc. Final Report on the Assessment of the
Economic Impact on the Automotive Parts/Service Industry of
Proposed Gasoline Lead Content Regulations, submitted to Policy
Planning and Assessment Directorate, Environment/Canada 1981
March.
Hernberg S, Nikkanen J. Enzyme inhibition by lead under normal
urban conditions. Lancet 1970; 1: 63-4.
Holtzman D, Shen Hsu J. Early effects of inorganic lead on immature
rat brain mitochondrial respiration. Pediatric Res 1976; 10:
70-5.
Holtzman D, Shen Hsu J, Desautel M. Absence of effects of lead
feedings and growth-retardation on mitochondriae and microsomal
cytochromes in the developing brain. Toxicol Appl Pharmacol
1981; 58: 48-56.
Hoppenbrouwers T, et al. Seasonal relationships of sudden infant
death syndrome and environmental pollutants. American Journal
of Epidemiology 1981; 113(6).
Hornbeck DD, et al. Advantages of lead in gasoline for european
cars - the lead road bonus. SAE Paper #750936, Society of
Automotive Engineers, Warrendale, PA 1975.
-------
-13-
Horvath S, et al. Maximal aerobic capacity at different levels
of carboxyhemoglobin. Journal of Applied Physiology 1975;
38: 300-3.
Horvath SM, Folinsbee LJ. Effects of pollutants on cardiopulmonary
function. Report to U.S. EPA 1979.
Hospadaruk V. Fundamental machanisms of automobile corrosion. SAE
Paper #780909, Society of Automotive Engineers, Warrendale, PA
1978.
Hudnall JR, et al. New gasoline formulations provide protection
against corrosive engine wear. SAE Paper #690514, Society of
Automotive Engineers, Warrendale, PA 1969.
Hudson PJ. Fitting segmented curves whose join points have to be
estimated. Journal of the Amercian Statistical Association
1966; 61: 873-80.
Hypertension Detection and Follow-up Program Cooperative Group.
The Effect of Treatment on Mortality in Mild Hypertension.
N Eng J Med 1983; 307: 976-80.
lannacone A, Carmigmani M, Boscolo P. Cardiovascular reactivity
in the rat following chronic exposure to cadmium and lead. Ann
1st Supr Sanrta 1981; 17: 655-60.
Innes WB. Environmental Science and Technology 1981; 15: 933.
Isringhaus EA Jr, Henderson BM. Engine rusting in short-run
passenger-car service. SAE Paper #650868, Society of Automotive
Engineers, Warrendale, PA 1965.
Jacobs KM. Energy usage and other comparisons between gasoline
and diesel medium duty trucks. SAE Paper #770757, Society of
Automotive Engineers, Warrendale, PA 1977.
/Janney A. The relationship between gasoline lead emissions and
blood poisoning in Americans. Prepared for U.S. EPA, Office of
Policy Analysis, 1982.
Joumard R, et al. Mathematical models of the uptake of carbon
monoxide on hemoglobin at low carbon monoxide levels.
Environmental Health Perspectives 1981; 41.
Journal of Pediatrics (Editorial). New approaches to screening
for iron deficiency. 1977; 90: 678.
Kagawa J, et al. Photochemical air pollution: its effects on
respiratory function of elementary school children. Arch
Environ Health 1975; 30: 117-22.
-------
-14-
Kakalik JS, et al. The Cost of Special Education. Rand Corporation
Report N-1791-ED, 1981.
Kammholz LP, Thatcher LG, Blodgett FM, Good TA. Rapid protoporphy-
rin quantitation for detection of lead poisoning. Petiatrics
1972; 50: 625-31.
Keller BD, et al. ORI of today's vehicles. SAE Paper #760195,
Society of Automotive Engineers, Warrendale, PA 1976.
Keller MD, et al. Respiratory illness in household using gas and
electricity for cooking. II. Symptoms and objective findings.
Environ Res 1979; 19: 504-15.
Kent WL, Finnigan FT. Effect of some fuel and operating parameters
on exhaust valve seat wear. SAE Paper #710673, Society of Auto-
motive Engineers, Warrendale, PA 1971.
Kerr HD, et al. Effects of nitrogen dioxide on pulmonary function
in human subjects. Environmental Chamber Study. U.S. EPA
Research Triangle Park, NC 1978.
Kerr JCS, et al. Effects of nitrogen dioxide on pulmonary function
in human subjects. Environmental Chambers Study. Environ Res
1979; 19: 392-404.
Killus JP, et al. Application of a regional oxidant model to the
northeast United States. Presented at the International
Conference on Long Range Transport Models for Photochemical
Oxidants and Their Precursor. U.S. EPA; Research Triangle Park,
NC, 1983 April 12-14.
Kim Y, Nickola R. A heat-resistant aluminized steel for high-
temperature applications. SAE Paper #800316, Society of Auto-
motive Engineers, Warrendale, PA 1980.
Kinosian J. Ozone precursor relationships from EKMA diagrams.
Environmental Science and Technology 1982; 16 (12).
Kipp KL, et al. Ability of gasoline additives to clean engines and
reduce exhaust emissions. SAE Paper #700456, Society of Automo-
tive Engineers, Warrendale, PA 1970.
Klein JP, et al. Hemoglobin affinity for oxygen during short-term
exhaustive exercise. Journal of Applied Physiology 1980; 48.
Kleinbaum DG, Morgenstern H, Keyser LL. Epidemiologic Research.
1982. °=
Kopp R, Vaughan W. Agricultural benefit analysis: alternative
ozone and photochemical oxidant standards. Resources for the
Future 1983 June 30; and discussion with authors, 1984.
-------
-15-
Kopp SJ, Glonek T, Erlanger M, Perry EF, Perry HM Jr, Barany M.
Cadmium and lead effects on myocardial function and metabolism.
J Environ Pathol Toxicol 1980; 4: 205-27.
Kromhout D, Coulande CL. Trace metals and CHD risk indicators
in 152 elderly men (the Zutphen Study). Eur Heart J 1984;
5 (Abstr suppl 1) : 101
Kuroishi N. Recent technology of sintered metals for automotive
use. SAE Paper #811416, Society of Automotive Engineers,
Warrendale, PA 1981.
Lamola A-A, Joselow M, Yamane T. Zinc protoporphyrin (ZPP): a
simple, sensitive, fluorometric screening test for lead poison-
ing. Clin Chem (Winston Salem, NC) 1975a; 21: 93-7.
Laraola A-A, Piomelli S. Poh-Fitzpartick MB, Yamane T, Barber LC.
Erythropoietic protoporphyria and lead intoxication: the
molecular basis for difference in cutaneous photosensitivity:
II. different binding of erythrocyte proteporphyrin to hemo-
globin. J Clin Invest 1975b; 56: 1528-5.
Landrigan PJ, Baker EL, Feldman RG, Cox DH, Eden KV, Orenstein WA,
Mather JA, Yankel AJ, von Lindern IH. Increased lead absorption
with anemia and slower nerve conduction in children near a lead
smelter. Journal of Pediatrics 1976; 89: 904-10.
Larson J, of Weyerhauser. personal communication, 1984.
Lauwerys R, Buchet JP, Roels H, Hubermont G. Placenta I transfer
of lead, mercury, cadmium, and carbon monoxide in women.
I. Comparison of the frequency distributions of the biological
indices in maternal and umbilical cord blood. Environmental
Research 1978; 15: 278-89.
Lave L, Seskin E. Air Pollution and Human Health. Baltimore:
Johns Hopkins University Press 1977.
Lax D. Analysis of fuel economy and vehicle use data for the 1977
truck inventory and use survey. SAE Paper #810388, Society of
Automotive Engineers, Warrendale, PA 1981.
Layland D, Cole H. A review of recent applications of the SAI
urban airshed model. U.S. EPA; Research Triangle Park,
NC 1983 December.
Lebowitz MD, et al. The effect of air pollution and weather on
lung function in exercising children and adolescents. American
Review of Respiratory Diseases 1974; 109.
-------
-16-
Lebowitz MD, O'Rourke MK, Dodge R. Holberg CJ, Gorman G, Hoshaw
RW, Pinnas JL, Barbe RA, Sweller MR. The adverse health
effects of biological aerosols, other aerosols, and indoor
microclimate on asthmatics and nonasthmatics. Environ Int
1982; 8: 375-80.
Lebowitz MD, Holberg CJ, Dodge RR. Repiratory effects on population
from low level exposures to ozone, presented at: 34th annual
meeting of the Air Pollution Control Association, 1983 June,
Atlanta, GA. Pittsburgh, PA: Air Pollution Control Association,
paper no 83 - 12.5.
Lebowitz, MD. The effects of environmental tobacco smoke exposure
and gas stoves on daily peak flow rates in asthmatic families.
Eur T Respir Dis 1984; 64 (133): 90-7.
Leighton J, Shehadi A, Wolcott R. Aggregate benefits of air pollu-
tion control, prepared for U.S. EPA, Office of Policy Analysis,
by Public Interest Economics Foundation; Washington, DC 1983
June.
Leikkanen HE, Beckman EW. Effect of leaded and unleaded gaso-
lines on exhaust emissions as influenced by combustion chamber
deposits. SAE Paper #710843, Society of Automotive Engineers,
Warrendale, PA 1971.
Lenane DL. MMT - a further evaluation. SAE Paper #770656, Society
of Automotive Engineers, Warrendale, PA 1977.
Levy RI. The decline in coronary heart disease mortality: Status
and perspective on the role of cholesterol. American Journal
of Cardiology 1984 August 27; 54 (4).
Lewis BW, Collin RJ, Wilson HS. Seasonal incidence of lead poisoning
in children in St. Louis. Southern Medical Journal 1955; 48: 298-301
Lewis RA, Newhall HK, Peyla RJ, Voss DA, Welstand JS. A New concept
in engine deposit control additives for unleaded gasoline. SAE
Paper #830938, Society of Automotive Engineers, Warrendale, PA
1983.
Lilienfeld AM, Lilienfeld DE. Foundations of Epidemiology.
New York, 1980.
Lincoln RH. written statement before the Senate Committee on
Environmental and Public Works in Opposition to Senate Bill 2609,
1984.
Lin-Fu JS. Vulnerability of children to lead exposure and toxicity:
parts one and two. New England Journal of Medicine 1973; 289:
1229-33, 1289-93.
-------
-17-
Linn WS, et al. Respiratory function and symptoms in urban office
workers in relation to oxidant air pollution exposure. Amer
Rev Resp Disease 1976; 114: 477-83.
Linn W, et al. Human respiratory effects of heavy exercise in
oxidant-polluted ambient air. American Review of Respiratory
Disease 1981; 123 (4) .
Lippmann M, et al. Effects of ozone on the pulmonary function of
children. Lee S, et al. eds, The Biomedical Effects of Ozone and
Related Photochemical Oxidants. Princeton Scientific Publishers,
Inc; Princeton, NJ: Advances in Modern Environmental Toxicology
1983; V: 423-46.
Litman DA, Corriea MA. L-tryptophan: a common denominator of
biochemial and neurological events of acute hepatic porphyrias?
Science 1983; 222: 1031-3.
Loehman E, et al. Distributional Analysis of regional benefits and
cost of air quality control. Journal of Environmental Economics
and Management 1979 Steptember; 6(3).
Love GJ, et al. Acute respiratory illness in families exposed
to nitrogen dioxide ambient air pollution in Chattanooga, TN.
Arch Environ Health 37: 75-80.
MacLean DC. Stickstoffoxide ALS phytotoxis CHE Luftuerurein 16
UNGEN. Staub-Reinhalt Luft 1975; 35(5).
Mage D, et al. A sensitivity analysis of the effects of CO time
patterns on computer blood carboyhemogologin levels, presented
at: Air Pollution Control Association Meetings 1983, Atlanta, Ga.
Mahaffey KR, Annest JL, Roberts J, Murphy MS. National estimation
of blood lead levels: United States (1976-1980). New England
Journal of Medicine; 307: 573-9.
Mahaffey KR, Rosen JF, Chesney RW, Peeler JT, Smith CM, Deluca HF.
Association between age, blood lead concentrations and serum
1,25-dihydroxycholecalciferol levels in children. Amercian Jour-
nal of Clinical Nutrition 1982; 35: 1327-31.
Makino K, Mizoguchi I. Symptoms caused by photochemical smog.
Japan Journal of Public Health 1975; 22(8).
/ /Manton WI. Sources of lead in blood: identification by stable
' ' isotopes. Archives of Environmental Health 1977; 32:149-59.
McBride WG, Black BP, English BJ. Blood lead levels and behavior of
400 preschool children. Med Journal Aust 1982; 2: 26-9.
-------
-18-
McCauley PT, Bull RJ. Lead-induced delays in synaptogenesis in the
rat cerebral cortex. Fed Proc Fed Am Soc Exp Bio 1978; 37: 740.
McCormick H. TRW Piston Ring Division, personal communication.
McDonnel TF. The effects of engine oil additives on vehicle fuel
economy, emissions, emission control components and engine wear.
SAE Paper #780962, Society of Automotive Engineers, Warrendale,
PA 1978.
McDonnell WF, Hortsmann DH, Hazucha DH, Hazucha MJ, Seal E Jr, Haak
ED, Salaam S, House DE. Pulmonary effects on ozone exposure
during exercise: dose-response characteristics. J Appl Physiol
Respir Environ Exercise Physiol 1983; 54: 1345-52.
McGartland A, Ostro B. Benefits Analysis of New Source Performance
Standards: The control of HC and NOX. U.S. EPA, Office of Policy
Analysis, Draft reported 1984 October.
McGartland A, Ostro B. Benefit Analysis of New Source Performance
Standards: The Control of VOC and NOX. U.S. EPA, Office of
Policy analysis, 1985.
McGee D, Gordon T. The results of the Framingham Study applied to
four other U.S. - based epidemeologic studies of coronary heart
disease. The Farmingham Study. Section 31. DHEW Pub No. (NIH)
76-1083. National Institutes of Health, Washington, D.C. U.S.
Government Printing Office 1976.
McLaughlin S, et al. Measuring effects of air pollution stress on
forest productivity. Tapp Journal 1984; 67: 74.
Melia RJW, et al. Association between gas cooking and respiratory
disease in children. Br Med J 1977; 2: 149-52.
Melia RJW, et al. Differences in N02 levels in kitchens with gas or
electric cookers. Atm Environ 1978; 12: 149-52.
Melia RJW, et al. The relation between respiratory illness in
primary schoolchildren and the use of gas for cooking. I -
Results from a national survey. Int J Epid 1979; 8: 333.
Mellins RB, Jenkins CD. Epidemiological and psychological study of
lead poisoning in children. Journal American Medical Association
1955; 158: 15-20.
Mentzer WC. Differentration of iron deficiency from thalessemia
trait. Lancet 1973; 1: 882.
Menzel DB. The role of free radicals in the toxicity of air pollu-
tants (nitrogen oxides and ozone). Bryor WA ed. Free Radicals
in Biology, vol III. New York: Academic 1976: 181-202.
-------
-19-
Meredith PA, Moore MR, Campbell BC, Thompson GG, Goldberg A. Delta-
animolaevulinic acid metabolism in normal and lead-exposed humans,
Toxicology 1978; 9: 1-9.
Milar CR, Schroeder SR, Mushak P, Dolcourt J, Grant LD. Contribu-
tions of the caregiving environment to increased lead burden of
children. American Journal Mental Deficiency 1980; 84: 339-44.
Milar CR, Schroeder SR, Mushak P, Boone L. Failure to find hyper-
activity in preschool children with moderately elevated lead bur-
den. Journal Pediatric Psychology 1981; 6: 85-95.
Moffitt JV. Compatability of military standard engines with leaded
and low-leaded fuels. AFLRL-19, Southwest Research Institute,
San Antonio, Texas (NTIS IAD-756 511) 1972.
Monthly Energy Review. U.S. Department of Energy, various issues,
1978 - 1982.
Moore MR, Meredith PA. The association of delta aminolevulinic acid
with the neurological and behavioral effects of lead exposure.
Hemphill DD ed. Trace Substances in Environmental Health - X.
Columbia, MO: University of Missouri - Columbia 1976.
Moran JB. The environmental implications of manganese as an alter-
native antiknock, SAE Paper #750926, Society of Automotive
Engineers, Warrendale, PA 1975.
Morbidity and Mortality Weekly Reports, various issues, 1976-1981,
U.S. Public Health Service.
Morbidity and Mortality Weekly Reports: Reports of the Lead Point
Poisoning Prevention Screening Program 1973-1981.
Morgan JM. Hyperkalemia and acidosis in lead neuropathy. South
Med J 1976; 69(7): 881-3.
Moreau T, Orssaud G, Juguet B, Busquet G. Plombemie et press ion
arterielle: premiers resultats d'une enquete transversable de 431
sujets de sexe masculin. [Blood lead levels and arterial pres-
sure: initial results of a cross sectional study of 431 male
subjects.] [Letter] Rev Epidemol Sante Fpulique 1982; 30: 395-7.
Morris WE, Rogers JD Jr, Poskitt RW. 1971 cars and the 'new1 gaso-
lines. SAE Paper #710624, Society of Automotive Engineers,
Warrendale, PA 1971.
Motor Vehicle Manufacturing Association. Motor Vehicle Facts and
Figures "83". Detroit 1983.
Motor Vehicle Manufacturing Association. Incentives for Proper
Usage of Unleaded Fuel. Detroit, 1984.
-------
-20-
Multiple Risk Factor Intervention Trial: Risk factor changes and
mortality results. Multiple Risk Factor Intervention Trial
Research Group. JAMA 1982; 248: 1465-77.
National Academy of Sciences. Lead: airborne lead in perspective.
Washington, DC: National Academy of Sciences 1972. (Biologic
effects of atmospheric pollutants.)
National Academy of Sciences. Air quality and automobile emission
control, Vol 4. prepared for the Committee on Public Works, U.S.
Senate, U.S. Government Printing Office 1974.
National Academy of Sciences. Nitrogen oxides: medical and bio-
logic effects of environmental pollutants. National Research
Council, Washington, DC 1976.
National Academy of Sciences. Nitrogen oxides. National Academy of
Sciences, Washington, DC 1977: 197-214.
National Center for Health Statistics. National Hospital Discharge
Survey, various issues.
National Center for Health Statistics. Plan and Operation of the
Second National Health and Nutrition Examination Survey 1976-
1980. National Center for Health Statistics 1981 (Vital and
Health Statistics Series 1, No. 15)
National Petroleum Council. Refinery Flexibility, an interim report
of the National Petroleum Council, 1979 December.
National Research Council. Air quality and automobile emisison
control, Vol 4: The costs and benefits of automobiles emission
control. A report by the coordinating Committee on Air Quality
Studies of the National Academy of Sciences and the National
Academy of Engineering. Prepared for the Committee on Public
works, United States Senate, 93rd Congress, 2nd Session.
Washington, DC: U.S. Government Printing Office 1974.
National Research Council. Nitrogen oxides: medical and biologi-
cal effects of environmental pollutants. National Academy of
Sciences, Washington, DC 1977: 1-333.
National Research Council. Nitrates, an environmental assessment.
National Academy of Sciences, Washington, DC 1978: 723.
National Research Council. Acid deposition: atmospheric processes
in eastern North America. NRC, National Academy Press, Washing-
ton DC 1983.
Needleman HL, Gunnoe C, Leviton A, Reed R, Peresie H, Maher C,
Barret P. Deficits in psychological and classroom performance of
children with elevated dentine lead levels. New England Journal
of Medicine 1979; 300: 689-95.
-------
-21-
Needleman HLf Rabinowitz M, Leviton A, Linn S, Schoenbaums S. The
relationship between prenatal exposure to lead and congenital
anomalies. Journal of the American Medical Association 1984; 251:
2956-9.
New approaches to screening for iron deficiency. Editorial.
Journal of Pediatrics; 90: 678.
Nisbet ICT. Sulfates and acidity in precipitation: their relation-
ship to emissions and regional transport of sulfur oxides. Air
quality and stationary source emission control, a report by the
Commission on Natural Resources of the National Research Council.
prepared for the Committee on Public Works, U.S. Senate, 94th
Congress, 1st Session. Washington, DC: U.S. Government Printing
Office 1975.
Odenbro A, Greenberg N, Vroegh K, Bederka J, Kihlstrom J-E. Func-
tional disturbances in lead-exposed children. Ambio 1983; 12:
40-4.
Orehek J, et al. Effect of short-term, low-level nitrogen dioxide
exposure on brochial sensitivity of asthmatic patients. J Clin
Invest 1976; 57:37
Orrin DS, Miner WR, Kipp KL. Unleaded gasoline - lubricant require-
ments and fuel additive performance. SAE Paper #720689, Society
of Automotive Engineers, Warrendale, PA 1972.
Otto DA, Benignus VA, Muller KE, Barton CN. Effects of age and body
lead burden on CNS function in young children. I: slow cortical
potentials. Electroencephalog Clin Neurophysiol 1981; 52: 229-39.
Otto, et al. Effects of low to moderate lead exposure on slow
cortical potentials in young children: two year follow up study.
Neurobehav Toxicol Teratol 1982; 4:733-7.
Otto D, Rovinson G, Baumann S, Schroeder S, Kleinbaum D, Barton C,
Mushak P, Boone L. Five-year follow-up study of children with
low-to-moderate lead absorption: electrophysiological evaluation.
presented at second international conference on prospective lead
studies 1984 April; Cincinnati, OH. available for inspection at:
U.S. Environmental Protection Agency, Environmental Criteria and
Assessment Office, Research Triangle Park, NC.
Page WP. Estimation of Economic losses to the agricultural sector
from airborne residuals in the Ohio River Basin. Journal of the
Air Pollution Control Association, 1982; 32: 151-4.
Paglia DE, Valentine WN, Dahlgren JG. Effects of low level lead
exposure on pyrimidine 5' nucleotidase and other erythrocyte
enzymes: possible role of pyrimidine 5' nucleotidase in the
pathogenesis of lead induced anemia. Journal of Clinical Inves-
tigation 1975; 56:1164-9.
-------
-22-
Pahnke AJ, Conte JF. Effects of combustion-chamber deposits and
driving conditions on vehicle exhaust emissions. SAE Paper
#690017, Society of Automotive Engineers, Warrendale PA 1969.
Pahnke AJ, Bettoney WE. Role of lead antiknocks in modern gasoline.
SAE Paper #710842, Society of Automotive Engineers Warrendale, PA
1971.
Patterson WR. Materials, design and corrosion effects on exhaust-
system life. SAE Paper #780921, Society of Automotive Engineers,
Warrendale, PA 1978.
Perlstein MA, Attala R. Neurologic sequelae of plumbism in children.
Clinic Pediatric (Philadelphia) 1966; 5: 292-8.
Perry HM Jr, Erlanger MW. Pressor effects of chronically feeding
cadmium and lead together. Hemphill DD ed. Trace Substances
in Environmental Health-XII; [proceedings of University of
Missouri's 12th annual conference on trace substances in environ-
mental health]; 1978 June, Columbia, MO. Columbia MO: University
of Missouri-Columbia: 268-75.
Pitts JN Jr, et al. Comment on effect of nitrogen oxide emissions
on ozone levels in metropolitan regions effects of NOX emission
rates on smog formation in the California south coast air basin,
and effects of hydrocarbon and NOX on photochemical smog forma-
tion under simulated transport conditions. Environ Sci Technol
1983; 54-7.
Piomelli S, Davidow B, Guinee VF, Yound P, Gay G. The FEP (free
erythrocyte porphyrins) test: a screening micromethod for lead
poisoning. Pediatrics 1973; 51: 254-9.
Piomelli S, Seaman C, Zullow D, Curran A, Davidow B. Metabolic
evidence of lead toxicity in "normal" urban children. Clinic Res
1977; 25: 459A.
Piomelli S, Seaman C, Zullow D, Curran A, Davidow B. Threshold for
lead damage to heme synthesis in urban children. Proceeding of
the National Academy of Sciences in 1982 May; 79: 3335-9.
Piomelli S, Rosen JF, Chisolm JJ, Graef JW. Management of child-
hood lead poisoning. Journal of Pediatrics 1984; 4:105.
Pirkle JL, Annest JL. Blood lead levels 1976-1980 - Reply (letter).
N Eng J Med 1984; 310(17): 1125-6.
Pirkle JL, Schwartz J, Landis JR, Harlan WR. The relationship
between blood lead levels and blood pressure and its cardiovas-
cular risk implications. American Journal of Epidemology, 1985-
121(2): 246-58
-------
-23-
Platts Oilgram Price Report, Price Average Supplement. December
1983 Monthly Averages. 1984 January 26.
Pless LG, Bennett PA. Effects of some engine, operating, and oil
variables on engine rusting in short-trip service. Lubrication
Engineering 1969 October; 384.
Pless LG. Effects of some engine, fuel, and oil additive factors on
engine rusting in short-trip service. SAE Paper #700457, Society
of Automotive Engineers, Warrendale, PA 1970.
Pless LG. Interactions among oil additive and engine operating
parameters affecting engine deposits and wear. SAE Paper #720686,
Society of Automotive Engineers, Warrendale, PA 1972.
Pless LG. A study of lengthened engine oil-change intervals. SAE
Paper #740139, Society of Automotive Engineers, Warrendale, PA
1974.
Pless LG. Deposits, wear, and catalyst performance with low ash
and ashless engine oils. SAE Paper #750900, Society of Automotive
Engineers, Warrendale, PA, 1975.
Pocock SJ, Ashby D. Environmental lead and children's intelli-
gence: a review of recent epidemiological studies. Statician (in
press).
The Pooling Project Research Group. Relationship of blood pres-
sure, serum cholesterol, smoking habit, relative weight and ECG
abnormalities to incidence of major coronary events: Final
report of the pooling project. J Chron Dis 1978; 31: 201-306.
Port CD. A comparative study of experimental and spontaneous emphy-
sema. J Toxicol Environ Health 1977; 2:589-604.
Portney P, Mullahy J. Ambient ozone and human health: an epidemio-
logical analysis, prepared for U.S. EPA, Office of Air Quality
Planning and Standards, 1983 September.
Portney P, et al. A multinomial model of the air pollution respira-
tory health relationship. Resources for the Furture 1984.
Portney P, Mullahy J. Urban air quality and acute respiratory
illness. Journal of Urban Economics forthcoming 1985.
Posin C, et al. Nitrogen dioxide and human blood biochemistry.
Arch Environ Health 1978 Nov/Dec; 318-24.
Provenzano G. The social costs of excessive lead-exposure during
childhood. HL Needleman ED. Low Level Lead Exposure; The
Clinical Implications of Current Research. New York: Raven
Press 1980.
-------
-24-
Provvedini DM, Tsoukas CD, Deftos LJ, Manolagas SC. 1,25-dihydroxy-
vitamin 03 receptors in human lukocytes. Science (Washington,
DC) 1983; 221: 1181-2.
Public Use Data Tape Documentation, Hematology and Biochemistry.
Second National Health and Nutrition Examination Survey 1976-
1980. Catalog number 5411, U.S. Public Health Service, National
Center for Health Statistics.
Rabinowitz WB, Needleman H. Petrol lead sales and umbilical cord
blood lead levels in Boston, Massachusetts. Lancet 1983.
Rabinowitz WB, Wetherill GW, Kopple JD. Kinetic analysis of lead
metabolism in human health. Journal of Clinical Investigation
1976; 58:260-70.
Ramirez-Cervantes B, Embree JW, Hine CH, Nelson KM, Varner MO,
Putnam RD. Health assessment of employees with different body
burdens of lead. Journal Occupational Medicine 1978; 20: 610-7.
Randall A, et al. Bidding games for valuation of aesthetic
environmental improvements. J Environ Econ Manag 1974; 1:
132-49.
Rasmussen H, Waisman DM. Modulation of cell function in the calcium
messenger system. Rev Physiol Biochemical Pharmacol 1983; 95:
111-48.
Raven P, et al. Effect of carbon monoxide and peroxyacetylitrate on
man's maximal aerobic capacity. Journal of applied physiology
1974; 36.
Raymond L. Today's fuels and lubricants and how they got that way.
SAE Paper #801341, Society of Automotive Engineers, Warrendale,
PA 1980.
Richet G, Albahary C, Morel-Maroqer L, Guillaume P, Galle P. Les
alterations renales dans 23 cas de saturnisme professionnel.
[Renal changes in 23 cases of occupational lead poisoning] Bull
Memorial Society Medical Hop Paris 1966; 117: 441-66.
Roels H, Buchet JP, Lauwerys R, Hubermont G, Braux P, Claeys-
Thoreau F, LaFontaine A, Van Overschelde J. Impact of air pollu-
tion by lead on the heme biosynthetic pathway in schoolage
children. Arch Environ Health 1976; 31: 310-6.
Rogers JJ, Kabel RH. The sequence IID engine oil test. SAE Paper
#780931, Society of Automotive Engineers, Warrendale, PA 1978.
Rosen JF, Chesney RW, Hamstra A, De Luca HF, Mahaffey KR. Reduction
in 1,25-dihydroxy vitamin D in children with increased lead
absorption. New England Journal of Medicine 1980a; 302: 1128-31.
-------
-25-
Rosen JF, Chesney RW, Hamstra A, Deluca HF, Mahaffey KR. Reduction
in 1,25-dihydroxyvitamin D in children with increased lead
absorption. Brown SS, Davis DS ed. Organ Directed Toxicity
Chemical Indices and Mechanisms. Pergamon Press 1980b.
Rosen JF, Chesney RW. Circulating calcitriol concentrations in
health and disease. Journal Pediatric (St. Louis) 1983; 103: 1-7.
Rowley DW. Exhaust system considerations for heavy duty trucks.
SAE Paper #770893, Society of Automotive Engineers, Warrendale,
PA 1977.
Royal Comission on Environmental Pollution. Ninth report - lead
in the environment. Cmnd 8852, Her Majesty's Stationery Office,
London, England 1983.
Rowe R, Chestnut L. Ozone and asthmatics in Los Angeles: a benefits
analysis. Report for Office of policy Analysis by Energy and
Resource Consultants, 1984.
Rummo JH, Routh DK, Rummo NJ, Brown JF. Behavioral and neurological
effects of symptomatic and asymptomatic lead-exposure in children.
Arch Env He 1979; 34(2): 120-4
Russell JA, Tosh JD, Johnson AA. Performance of army engines with
leaded and unleaded gasoline. AFLRL-21, Southwest Research
Institute, San Antonio, Texas (NTIS IAD-766 760) 1973.
SAI. Simulations of the Regional Air Quality Impacts of
Industrial Emission Controls. Final report to U.S. EPA,
Economic Analysis Division 1984.
Sanders M. U.S. Postal Service, personal communication.
SAS Users Guide. Statistical Analysis System, inc. Gary,
NC 1982.
Sassa S, Garnick JL, Garnick S, Kappas A, Levere RD. Studies in
lead poisoning. I: Microanalysis of erythrocyte protoporphyrin
levels by spectrofluorometry in the detection of chronic lead
intoxication in the subclinical range. Biochem Medical 1973;
8: 135-48.
Schwartz J, Leggett J, Ostro B, Pitcher H, Levin R. Costs and
Benefits of Reducing Lead in Gasoline. U.S. EPA, Office of
Policy Analysis 1984a.
/Schwartz J, Janney A, Pitcher H. The relationship between gasoline
lead and blood lead. U.S. EPA, Office of Policy Analysis 1984b.
Schwartz J. Blood lead and blood pressure. Memo to EPA Central
Docket EN-84-05, 1984c September 7.
-------
-26-
Schwartz J. Change in aromatics emissions due to lead phasedown.
Memo to EPA Central Docket EN-84-05, 1984d September 17.
Schwartz J. Availability and severity of catalytic reforming
equipment. Memo to EPA Central Docket EN-84-05, 1984e.
September 11.
Schwochert HW. Performance of a catalytic converter that operates
with nonleaded fuel. SAE Paper #690503, Society of Automotive
Engineers, Warrendale, PA 1969.
Seigneur C, Saxena P, Roth P. Preliminary results of acid rain
modeling, submitted at a Specialty Conference on Atomspheric
Deposition sponsored by the Air Pollution Control Association;
Detroit, Michigan 1982 November 7-10.
Seigneur C, et al. Modeling studies of sulfate and nitrate chemis-
try: the effect changes in sulfur dioxide, nitrogen oxide, and
reactive hydrocarbon levels. final report 1984.
Shah BV, Surregr; Standard Errors of Regression Coefficients
from Sample Survey Data. Research Traingle Park, NC,
Research Triangle Institute 1982.
Shepard CD. Instrumentation and computer techniques aid automotive
exhaust system design for low noise levels; materials to reduce
system corrosion. SAE Paper #690006, Society of Automotive
Engineers, Warrendale, PA 1969.
Sherwin RP, et al. Sequestration of exogenous peroxidase in the
lungs of animals exposed to continuous 0.5 ppm nitrogen oxide.
Fed Proc 1977; 36:1091.
Shier DR, Hall A. Analysis of housing and data collected in a lead
based paint survey in Pittsburgh. National Bureau of Standards
1977.
Shlossman M, Brown M, Shapiro E, Dziak R. Calcitonin effects on
isolated bone cells. Calcif Tissue Int 1982; 34: 190-6.
Shy CM, et al. The Chattanooga school children study: effects of
community exposure of nitrogen dioxide. I Methods, description
of polllutant exposure and results of ventilatory function test-
ing. J Air Pollut Control Assoc 1970; 20 (8):539-45.
Shy CM, Love GJ. Recent evidence on the human health effects of
nitrogen dioxide. Proceedings of the Symposium on Nitrogen
Oxides; Honolulu, Hawaii, 1979 April 4-5.
Sigsby J, et al. Automotive emissions of ethylene dibromide.
SAE Paper #820786. Society of Engineers, Warrendale, PA 1982.
-------
-27-
Silbergeld EK, Adler HS. Subcellular mechanisms of lead neuro-
toxicity. Brain Research 1978; 148: 451-67.
Silbergeld EKf Hruska RE, Miller LP, Eng N. Effects of lead ir\ vivo
and in vitro on GABAergic neurochemistry. J Neurochem 1980;
34:1712-8.
Silver W, Rodriguez-Torres R. Electrocardiographic studies in
children with lead poisoning. Pediatrics 1968; 41: 1124-7.
Smith M, Delves T, Lansdown R, Clayton B, Graham P. The effects of
lead exposure on urban children: the Institute of Child Health/
Southampton study. United Kingdom Department of the Environment
1983.
Smith RS. The feasibility of an injury tax approach to occupational
safety. Law and Contemporary poblems 1974; 38: 730-44.
Smith RS. The Occupational Safety and Health Act. Washington, DC:
American Enterprise Institute for Public Policy Research, 1976.
Society of Automotive Engineers. Fuel Economy Measurement - Road
Test Procedure. SAE Standard J 1082, Warrendale, PA 1980
September.
Soloman and Associates. Letter to the National Petroleum
Refiners Association. Available in the U.S. EPA Central
Docket EN-84-05, 1984.
Sorem SS. Effects of fuel factors on emissions. SAE Paper #710364.
Society of Automotive Engineers, Warrendale, PA 1971.
Speizer FE, et al. Respiratory disease rates and pulmonary function
in children associated with NC>2 exposure. Am Rev Resp Dis 1980;
121:3-10.
Spengler JD, et al. Sulfur dioxide and nitrogen dioxide levels
inside and outside homes and the implications on health effects
research. Environ Sci Techn 1979; 13:1271-6.
Spengler JD. Nitrogen dioxide inside and outside 137 homes and
implications for ambient air quality standards and health effects
research. Environ Sci Technol 1983; 17:164-8.
SRI International. An estimate of the nonhealth benefits of meeting
the Secondary National Ambient Air Quality Standards, a final
report to the National Commission on Air Quality 1981.
Stambaugh RL, Kopko RJ, Franklin TM. Effects of unleaded fuel and
exhaust gas recirculation on sludge and varnish formation. SAE
Paper #720944, Society of Automotive Engineers, Warrendale, PA
1972.
-------
-28-
Sugiura K. Present status of Japanese motor oils and extension of
oils drain intervals. SAE Paper #780953, Society of Automotive
Engineers, Warrendale, PA 1978.
Supplemental Analysis of Refining Costs. Submitted to U.S. EPA
Central Docket EN-84-05, 1984 September 19.
Tera, et al. Identification of gasoline lead in childrens1 blood
using isotopic analysis. Arch Env Health 1985 January.
Thaler R, Rosen S. The value of saving a life: evidence from the
labor market. Taleckji NE ed. Household Production and Consump-
tion. New York: Columbia University Press 1976.
Thomas HV, et al. Lipoperoxidation of lung lipids in rats exposed
to nitrogen dioxide. Science 1968; 159:532-4.
Thompson CR, et al. Effects of continous exposure of navel oranges
to NO2. Atm Environ 1970; 4:349-55.
Tosh JD, Johnson AA, Frame EA. Performance of army engines with
leaded and unleaded gasoline. AFLRL-54, Southwest Research
institute, San Antonio, Texas (NTIS #AD-A005 577) 1975.
Tosh JD. Performance of army engines with unleaded gasoline—
field study. AFLRL-82, Southwest Research Institute, San Antonio,
Texas (NTIS #AD-A032 075) 1976.
Trijonis JC. Empirical Relationships Between Atmospheric Nitrogen
Dioxide and Its Precursors. U.S. EPA, Office of Research and
Development. Environmental Sciences Research Laboratory,
Research Triangle Park, NC, 1978 February.
Trijonis JC, et al. The Relationship of Ambient NC<2 to Hydrocarbon
and NOX Emissions. Draft report from Technology Service Corpor-
ation to U.S. EPA under Contract NOX 68-02-2299. U.S. EPA,
Office of Research and Development. Triangle Park, NC 1979.
Turner, Mason and Associates. Reply on behalf of the Lead Indus-
tries Association, July 1984. Available in the EPA Central
Docket EN-84-05.
Union Carbide, personal communication, 1984.
Unzelman GH, Michalski GW. Octane improvement update — refinery
processing, antiknocks, and oxygenates. NPRA Paper #AM-84-43,
National Petroleum Refiners Association, Washington, DC 1984.
Urban CM, Springer KJ, McFadden JJ. Emissions control of gasoline
engines for heavy-duty vehicles. SAE Paper #750903, Society of
Automotive Engineers, Warrendale, PA 1975.
-------
-29-
Urban CM, Kaupert AW. Evaluation of emission control technology
approaches for heavy-duty gasoline engines. SAE Paper #780646,
Society of Automotive Engineers, Warrendale, PA 1978.
U.S. Department of Agriculture. Agricultural Statistics, 1980.
Washington, DC: U.S. Government Printing Office 1982.
U.S. Department of Energy, Energy Information Administration.
Annual Report to Congress. 1983.
U.S. Department of Energy, Energy Information Administration.
Annual Report to Congress. 1984a.
U.S. Department of Energy, Energy Information Administration,
Office of Oil and Gas Refinery Evaluation Modeling System (REMS),
model documentation. 1984b July.
U.S. Department of Energy. Refinery Evaluation Modeling System
(REMS), 1983 Analysis of Verification. 1985 January.
U.S. Department of Health, Education, and Welfare. Limitations of
activity and mobility due to chronic conditions. Public Health
Service, Vital and Health Statistics Series 10, 1973a; (96).
U.S. Department of Health, Education, and Welfare. Prevalance of
selected chronic respiratory conditions: United States - 1970.
Public Health Service, Vital and Health Statistics Series 10,
1973b; (84).
U.S. Department of Health and Human Services, Public Health Ser-
vice. Blood carbon monoxide levels in persons 3-74 years of age:
United States, 1976-80. Advance Data 1982 March 17; (76).
U.S. Environmental Protection Agency. Ozone SIP Data Base and
Summary Report. Office of Air Quality Planning and Standards,
Research Triangle Park, NC.
U.S. Environmental Protection Agency. Air Quality Criteria for
Ozone and Other Photochemical Oxidants. Office of Research and
Development, 1978 April.
U.S. Environmental Protection Agency, Carcinogen Assesmment Group.
The Carcinogen Assessment Group's Final Report on Population Risk
to Ambient Benzene Exposure. 1979a January 10.
U.S. Environmental Protection Agency, Office of Mobile Source Air
Pollution Control. Draft Regulatory Analysis: Heavy-Duty Diesel
Particulate Regulations. U.S. EPA; Ann Arbor, Michigan 1979b.
U.S. Environmental Protection Agency. Regulatory Impact Analysis
of the National Ambient Air Quality Standards for Carbon Monoxide,
Office of Air Quality Planning and Standards, 1980.
-------
-30-
U.S. Environmental Protection Agency. The Costs of Controlling
Emissions of 1981 Model Year Automobiles. Office of Mobile
Source Air Pollution Control, 1981 June.
U.S. Environmental Protection Agency. 1982 NCLAN Annual Report.
Environmental Research Lab, Corvallis, Oregon 1982a.
U.S. Environmental Protection Agency. Air Quality Criteria for
Oxides of Nitrogen (Criteria Document). Research Triangle Park,
NC 1982b.
U.S. Environmental Protection Agency. Review of the National
Ambient Air Quality Standards for Nitrogen Oxides: Assessment of
Scientific and Technological Information. OAQPS, NC, 1982c August.
U.S. Environmental Protection Agency. NAAQS Environmental Impact
Statement for N02. Research Triangle Park, NC 1982d.
U.S. Environmental Protection Agency. Cost and Economic Assessment
of Regulatory Alternatives for NO2 NAAQS (DRAFT). Research Tri-
angle Park, NC 1982e.
U.S. Environmental Protection Agency. National Air Pollutant
Emissions Estimates, 1940 - 1980. Monitoring and Data Analysis
Division 1982f January.
U.S. Environmental Protection Agency. Regulation of Fuel and Fuel
Additives. 47 Federal Register 49382, 1982g October 28.
U.S. Environmental Protection Agency. Motor Vehicle Tampering
Survey 1982. National Enforcement Investigation Center, 1983a
April.
U.S. Environmental Protection Agency, OAR, QMS. Anti-Tampering and
Anti-Misfueling Programs to Reduce In-Use Emissions from Motor
Vehicles. 1983b December 31.
U.S. Environmental Protection Agency. The API Study and Its
Possible Human Health Implications, memo from Al Lorang (Chief,
Technical Support Staff) to Charles Gray, Jr. (Director, Emission
Control Technology Division), 1983c May 16.
U.S. Environmental Protection Agency, Environmental Criteria and
Assessment Office. Air Quality Criteria for Lead. 1983d.
U.S. Environmental Protection Agency, Office of Mobile Source Air
Pollution Control, Standards Development and Support Branch.
Issues Analysis - Final Heavy-Duty Engine HC and CO Standards.
U.S. EPA; Ann Arbor, Michigan 1983e.
U.S. Environmental Protection Agency. Preliminary Draft Revised
Air Quality Criteria for Ozone and Other Photochemical Oxidants.
Office of Researach and Development 1983f.
-------
-31-
U.S. Environmental Protection Agency. The Feasibility, Cost, and
Cost-Effectiveness of Onboard Vapor Control, prepared by Glenn
Passavant, 1984a March.
U.S. Environmental Protection Agency. VOC/Ozone Relationships from
EKMA. memo from Warren Freas (Air Management Technology Branch)
to Alan McGartland (Benefits Branch), 1984b January 27.
U.S. Environmental Protection Agency. Response to Public Comments
on EPA's Listing of Benzene Under Section 112 and Relevant Proce-
dures for the Regulation of Hazardous Air Pollutants. OAQPS,
1984c.
U.S. Environmental Protection Agency, Office of Air, Noise and
Radiation. Regulation of Fuels and Fuel Additives: Lead Phase
Down. Federal Register 1984d August 2; 31032.
U.S. Environmental Protection Agency. Motor Vehicle Tampering
Survey 1983. National Enforcement Investigation Center, 1984e.
U.S. Environmental Protection Agency, Office of Policy Analysis.
Preliminary Regulatory Impact Analysis of Proposed Rules Limiting
the Lead Content of Gasoline. 1984f July 23.
U.S. Environmental Protection Agency. Regulatory Impact Analysis
Guidelines. 1984g.
U.S. Environmental Protection Agency. Draft Revised Air Quality
Criteria for Ozone and Other Photochemical Oxidants. Office of
Research and Development 1984h.
U.S. Environmental Protection Agency. Regulation of Fuels and Fuel
Additives; Banking of Lead Rights. 50 Federal Register 718, 1985
January 4.
U.S. Office of Management and Budget. Interim Regulatory Impact
Analysis Guidance. 1981, p. 4.
Victery W, Vander AJ, Shulak JM, Schoeps P, Juluis S. Lead, hyper-
tension, and the renin-angiotensin system in rats. Journal Labor-
tory Clinic Medical 1982; 99: 354-62.
Violette, Chestnut. Valuing Reductions in Risks: A Review of the
Empirical Estimates. 1983.
Viscusi WK. Labor market valuations of life and limb: empirical
evidence and policy implications. Public Policy 1978; 26: 359-86
Von Nieding G, et al. Minimum concentration of N02 causing acute
effects on the respiratory gas exchange and airway resistance in
patients with chronic bronchitis. Int Arch Arbeitsmed 1971;
27:338-48. U.S. Environmental Protection Agency, Researach
Triangle Park, NC.
-------
-32-
Von Nieding G, et al. Studies of the acute effect of NC-2 on lung
function: influence on diffusion, perfusion and ventilation in
the lungs. Int Arch Arbeitsmed 1973; 31:61-72.
Von Nieding G, et al. Acute effects of ozone on lung function of
men. VDI-Ber 1977; 270:123-9.
Wade WA III, et al. A study of indoor air quality. J Air Pollut
Control Assoc 1975; 25-93-939.
Wagner TO. Motor gasoline and automotive air pollution. SAE Paper
#710485, Society of Automotie Engineers, Warrendale, PA 1971.
Wagner TO, Russum LW. Optimum octane number for unleaded gasoline.
SAE Paper #720552, Society of Automotive Engineers, Warrendale,
PA 1973.
Ware JH, et al. Passive smoking, gas cooking and respiratory
health of children living in six cities. America Review of
Respiratory Diseases 1984; 129.
Weaver C. Particulate control technology and particulate standards
for heavy-duty diesel engines. SAE Paper #840174, Diesel Particu-
late Traps: 140, Society of Automotive Engineers, Warrendale, PA
1984.
Weaver CS. The Effects of Low-Lead and Unleaded Fuels on Gasoline
Engines. Prepared for: U.S. Environmental Protection Agency,
Office of Policy Analysis. Energy and Resource Consultants, Inc,
1984b September 28.
Webb RC, Winquist RJ, Victery W, Vander AJ. In vivo and in vitro
effects on lead on vascular reactivity in rats. AM J Physiol
1981; 24: H211-6.
Wedeen RP, Maesaka JK, Weiner B, Lipat GA, Lyons MM, Vitale LF,
Joselow MM. Occupational lead nephropathy. American Journal of
Medicine 1975; 59: 630-41.
Weekly Petroleum Status Report. 1984, various issues.
Weinstein MC, Stason WB. Allocation of Resources to Manage Hyper-
tension. N Eng J Med 1977; 296(13): 732-9.
White JM, Harvey DR. Defective synthesis of A and B globin chains
in lead poisoning. Nature (London) 1972; 236: 71-3.
Whittemore A, Korn EL. Asthma and air pollution in the Los Angeles
area. American Journal of Public Health 1980; 70.
Wibberly DG, Khera AK, Edwards JH, Rushton DI. Lead levels in human
placentae from normal and malformed births. Journal of Medical
Genetics 1977; 14: 339.
-------
-33-
Williams BJ, Griffith WH III, Albrecht CM, Pirch JH, Hejtmancik MR
Jr. Effects of chronic lead treatment on some cardiovascular
responses to norepinephrine in the rat. Toxicology Appl Pharma-
col 1977; 40: 407-13.
Williams BJ, Goldman D, Hejtmancik MR, Ziegler MG. Noradrenergic
effects of lead in neonatal rat. Pharmacology 1978; 20(3): 186.
Williams BJ, Hejtmancik M. Time and level of perinatal lead-
exposure for development of norepinephrine cardiotoxicity. Res
Comm CP 1979; 24(2): 367-76.
Willits N, Ott W. Modeling the dynamic response of an automobile
for air pollution exposure studies. Environraetrics, Summaries of
Conference Presentations 1981; 81: 104-5.
Winneke G, Brockhaus A, Baltissen R. Nerve behavioral and systemic
effects of longterm blood lead elevation in rats. Arch Toxicol
1977; 37: 247-63.
Winneke G. Impaired intelligence in children from environmental
lead (letter). Mun Med Woe 1979; 121(26): 865.
Winneke G, Lilienthal H, Werner W. Task dependent nerve behavioral
effects of lead in rats. Arch Toxicol 1982a; Supp 5: 84-93.
Winneke G, Hrdina K-G, Brockhaus A. Neuropsychological studies in
children with elevated tooth-lead concentrations, part 1: pilot
study. Int Arch Occupational and Environmental Health 1982b;
51: 169-83.
Winneke G, Kramer U, Brockhaus A, Ewers U, Kujanek G, Lechner H,
Janke W. Neuropsychological studies in children with elevated
tooth lead concentrations. Part II: extended study. Int Arch
Occup Environ Health 1983; 51: 231-52.
Winneke G, Beginn U, Ewert T, Havestadt C, Kramer U, Krause C,
Thron HL, Wagner HM. Studie zur erfassung subklinischer blei-
wirkungen auf das nervensystem bei kindern mit bekannter prana-
taler exposition in Nordenham. [Study on the determination of
subclinical lead effects on the nervous system of Nordenham
children with known pre-natal exposure.] BGA-Berichte 1984.
Wintringham JS, et al. Car maintenance expense in owner service
with leaded and nonleaded gasolines. SAE Paper #720499, Society
of Automotive Engineers, Warrendale, PA 1972.
Wong GL. Actions of parathyroid hormone and 1,25-dihydroxcholecal-
ciferol on citrate decarboxylation in osteoblast-like bone cells
differ in calcium requirement and in sensitivity to trifluopera-
zine. Calcif Tissue Int 1983; 35: 426-31.
-------
-34-
World Health Organization, United Nations Environmental Program;
Lead: Environmental Health Criteria 3, Geneva, Switzerland 1977.
Worthen RP, Tunnecliffe TN. Temperature controlled engine valves.
SAE Paper #820501, Society of Automotive Engineers, Warrendale,
PA 1982.
WR Grace, Davison Chemical Division. Davison Catalagram #69, 1984.
Yip R, Norris TN, Anderson AS. Iron status of children with ele-
vated blood lead concentrations. Journal Pediatr (St. Louis)
1981; 98: 922-5.
Young CT, Grimes DA. Erosion mechanisms of automotive spark plug
electrodes. SAE Paper #780330, Society of Automotive Engineers,
Warrendale, PA 1978.
Yule W, Landsdown R, Millar IB, Urbanowicz MA. The relationship
between blood lead concentrations, intelligence and attainment in
a school population: A pilot study. Dev Medical Child Neurology
1981; 23:567-76.
Yule w, Lansdown R. Lead and children's development: recent
findings, presented at international conference: Management
and Control of Heavy Metals in the Environment, 1983 September;
Heidelberg, West Germany.
Yule W, Urbanowicz M-A, Lansdown R, Millar IB. Teacher's ratings
of children's behaviour in relation to blood lead levels.
British Journal of Developmental Psychology 1984.
Zagraniski R, Leaderer B, Stolwuk J. Ambient sulfates, photochemical
oxidants and acute adverse health effects: an epidemiologic study.
Environmental Research 1979; 19.
Ziskind RA, et al. Carbon Monoxide Intrusion in Sustained-Use
Vehicles, #SAI-068-80-535. U.S. EPA; Research Triangle Park,
NC, 1979 November.
-------
APPENDIX A
REFINERY PROCESSES
In refining, crude oil is first separated by molecular size
into fractions, each of which can be blended directly into final
petroleum products or processed further. In the downstream pro-
cessing operations, the molecular size and structure of petroleum
fractions are altered to conform to desired characteristics of
refined products. Table li-l in the text classifies the various
refinery processes according to their principal functions. The
actual processing configuration will depend on the characteristics
of the crude oil processed and on the desired final product mix.
These major processing steps are described briefly below.
Fluid Catalytic Cracking uses high temperature in the pres-
ence of a catalyst to convert or "crack" heavier fractions into
lighter products, primarily gasoline and distillates. Feed is
brought to process conditions (1000°F and 20 pounds per square
inch pressure [psi]) and then mixed with a powdered catalyst in
a reaction vessel. In the reactor, the cracking process is
completed and the hydrocarbon products pass to a fractionating
section for separation.
Coke, a coal-like by-product, is formed on the catalyst as a
result of the cracking reaction. Coked catalyst is transferred
from the reactor to a regenerator vessel where air is injected
to burn the coke to CO and CO2. The regenerator flue gases are
passed through cyclones and, sometimes, electrostatic precipita-
tors, to remove entrained catalyst. They are then vented to the
-------
A-2
atmosphere or sent to a CO boiler where carbon monoxide is burned
to produce CC>2. The regenerated catalyst is returned to the
reactor.
Hydrotreating (also known as hydrodesulfurization) is a
catalytic process designed to remove sulfur, nitrogen, and
heavy metals from petroleum fractions. Feed is heated to process
temperatures (650° to 705°F), mixed with hydrogen, and fed to
a reactor containing a fixed bed of catalyst. The primary reac-
tions convert sulfur compounds in the feed to hydrogen sulfide
(H2S) and the nitrogen compounds to ammonia. The B^S and
ammonia are separated from the desulfurized product; the t^S
is sent to sulfur recovery facilities.
Catalytic reforming is used to upgrade low-octane naphtha
to produce high-octane gasoline blending stocks. The flow pat-
tern is similar to that of hydrotreating except that several
reactor vessels are used. The required temperature is about
1000"F and the required pressure is about 200 pounds per square
inch. Reforming catalysts are readily poisoned by sulfur,
nitrogen, or heavy metals, and therefore the feed is normally
hydrotreated before being charged to the reforming unit.
In hydrocracking the cracking reaction takes place in the
presence of hydrogen. The process produces high quality desul-
furized gasoline and distillates from a wide variety of feed-
stocks. The process employs one or more fixed bed reactors and
is similar in flow to the hydrotreating process. Process
conditions are 800°F and 2000 psi. Like hydrotreating, hydro-
-------
A-3
cracking produces by-product H2S, which is diverted to sulfur
recovery.
Coking is another type of cracking which does not employ a
catalyst or hydrogen. The process is utilized to convert heavy
fuel oils into light products and a solid residue (coke). Feed
is brought to process conditions (900°P and 50 psi) and fed to
the coking vessel. Cracked products are routed to a fractionation
section. Coke accumulates in the vessel and is drilled out about
once a day. In one version of the coking process, fluid coking,
a portion of the coke is used for process fuel and the balance
is removed as small particles.
Acid gas treating and sulfur recovery units are used to
recover hydrogen sulfide (H2S) from refinery gas streams and
convert it to elemental sulfur. Sour gas containing H2S is
produced in several refinery units, particularly cracking and
hydrotreating. In the acid gas treating units, f^S is removed
from the fuel gas by absorbing it in an alkaline solution. This
solution, in turn, is heated and steam-stripped to remove the
H2S to form sulfur and water. Sulfur recovery is high but
never 100%. The remaining sulfur is incinerated and discharged
to the atmosphere or removed by a tail gas treating unit.
The purpose of the tail gas treating unit is to convert any
remaining sulfur compounds from the sulfur recovery unit to
elemental sulfur. There are several processes available, the
most common of which are the Beavon and SCOT processes. In both
processes, sulfur compounds in the sulfur unit tail gas are
-------
A-4
converted to H2S. The Beavon process converts H2S to sulfur
through a series of absorption and oxidation steps. The SCOT
process concentrates the f^S and returns it to the sulfur
recovery facilities. In both processes, the treated tail gas
is virtually free of sulfur compounds when released to the
atmosphere.
-------
APPENDIX B
THE FLEET MODEL
We developed a fleet model to predict the number of cars and
gasoline powered trucks (for six weight classes) on the road; the
model also forecasts the number of miles driven per year for each
vehicle type for the years 1985-1992. In this document we have
used our fleet model for several different purposes. Estimates
of total miles driven per year by vehicle type, age and fuel
consumed are used to estimate the maintenance savings attributable
to the use of lower lead or unleaded fuel. Estimates of lifetime
vehicle-miles traveled by new misfuelers in each year provide
the basis for estimating the value of the reduced emissions due
to reduced misfueling.
B.I Vehicles on the Road
The extrapolation procedure for total vehicle miles per year
by age and type of vehicle relies on two basic sets of data:
1) Vehicles on-the-road for 1968-1983 as published annually by
the Motor Vehicle Manufacturers Association(MVMA) in Motor
Vehicle Facts and Figures. Data are available for cars
(Table B-l) and trucks (Table B-2) ages 1 to 15 years, as
well as 16 and older. No breakdown of trucks by weight and
engine type (diesel versus gasoline) is available, however.
2) Data Resources, Inc. (DRI) projections of the total fleet
size for cars, as well as annual sales for cars and trucks.
Using these data sets, the fleet model projected how many cars
and trucks (by age) would be on the road for each of the years,
1985 through 1992. For trucks, the projections reflect only
-------
B-2
gasoline-powered vehicles and the fleet is divided into six
weight classes.
The model required two sets of inputs: an inventory of
vehicles on the road by age and a set of survival rates for
determining how many vehicles in each age category would still be
on the road the next year. The initial inventory for vehicles
aged "New1 through 15 years of age is taken from the 1983 column
in Tables B-l and B-2. The category of vehicles aged 16 and over
is broken into 15 additional age categories. The procedures we
used are discussed in sections B.I.a for cars and B.l.b. for
trucks.
The survival rates used in the forecast are based on an
analysis of historical survival rates. These historical survival
rates are derived by dividing the number of vehicles in a model
year cohort in one year by the number of vehicles in the same
cohort the previous year. For cars, the historical survival
rates are given in Table B-3, while Table B-4 presents the rates
for trucks. As both tables show, there is significant variation
in the rates over time. A statistical analysis of the data
showed that both the overall scrappage rate (defined as total
vehicle retirements divided by fleet size) and the fraction of
the total fleet that each model year comprised were important
explanatory variables. So, for the forecast, we adjusted the
survival rates to reflect changes in the scrappage rates and the
relative size of the cohort in each year. As explained below,
the scrappage rates used for cars are derived from the DRI
-------
B-3
Table B-l Cars on the Road 1968-1983
Year
Age of
Car
New
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 +
Total
1968
6182
8122
8836
8939
7667
7058
6183
4657
4615
3347
1709
1990
1612
1496
743
623
1517
75358
1969
6467
8927
8054
8798
8855
7532
6829
5804
4087
3726
2452
1188
1421
1139
1063
525
1578
80449
1970
6288
9299
8816
7878
8538
8506
7116
6268
5058
3267
2776
1692
799
996
794
753
1583
80449
1971
5927
8888
9280
8802
7772
8313
8171
6651
5624
4274
2525
2035
1183
563
730
580
1804
83138
1972
7169
8915
8851
9122
8596
7499
7930
7583
5920
4713
3343
1824
1413
805
389
526
1813
86439
1973
7988
10158
8715
8612
8881
8291
7120
7333
6715
4963
3698
2470
1268
967
548
274
1780
89805
1974
6433
11269
10147
8622
8493
8615
7931
6624
6531
5710
3976
2824
1813
901
682
391
1621
92608
1975
4684
9763
11332
10098
8549
8341
8339
7556
6113
5796
4825
3234
2229
1407
689
523
1742
95241
Year
1976
1977 1978 1979 1980 1981 1982
1983
Age of
Car
New
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 +
6472
7683
9746
11130
9872
8249
7966
7774
6856
5361
4888
3923
2578
1740
1083
526
1943
7177
9557
7477
9594
10854
9563
7866
7449
6963
5859
4416
3887
3023
1969
1315
818
2093
7426
10382
9483
7291
9431
10559
9140
7326
6784
6087
4917
3589
3093
2369
1545
1021
2496
7288
10699
10219
9203
6990
9004
9965
8431
6573
5909
5034
3999
2862
2460
1874
1223
2930
5868
10402
10483
9931
8900
6682
8499
9151
7544
5653
4939
4049
3172
2280
1969
1516
3514
5140
8818
10245
10290
9758
8735
6463
8050
8458
6791
4929
4238
3369
2635
1910
1654
4346
4399
8280
8825
10075
10155
9661
8471
6190
7498
7629
5989
4243
3581
2822
2208
1609
5220
5044
7429
8273
8749
10014
10038
9434
8195
5867
6885
6798
5239
3632
3052
2395
1869
6037
Total 97818 99904 102957 104677 104564 105839 106867 108961
-------
B-4
Table B-2 Trucks on the Road 1968-1983
Year
Age of
Truck
NEW
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 +
TOTAL
Year
Age of
Truck
NEW
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 +
1968
1058
1452
1512
1380
1208
1035
889
692
749
682
461
508
514
537
370
425
2182
15685
1976
1893
2148
2732
2799
2346
1697
1635
1731
1345
1220
1191
1024
828
642
503
351
2455
1969
1262
1581
1447
1495
1357
1177
1004
860
661
706
632
421
459
459
480
330
2242
16586
1977
2177
2746
2109
2689
2752
2291
1639
1573
1645
1267
1129
1096
922
736
566
442
2422
1970
1263
1881
1536
1428
1483
1339
1154
975
826
621
658
583
383
417
414
432
2278
17686
1978
2533
3240
2743
2076
2656
2681
2227
1567
1509
1554
1189
1043
998
832
663
506
2531
1971
1193
1736
1872
1496
1398
1441
1298
1112
927
774
585
610
532
347
376
369
2383
18465
1979
2402
3541
3231
2679
2006
2589
2587
2140
1501
1435
1459
1111
958
906
749
595
2687
1972
1637
1784
1744
1858
1468
1372
1409
1260
1066
877
721
525
544
471
303
328
2380
19773
1980
1362
3765
3663
3332
2750
2046
2641
2609
2163
1540
1429
1452
1082
919
856
688
2948
1973
1883
2385
1753
1709
1825
1446
1336
1363
1208
1005
817
663
474
487
419
268
2356
21412
1981
1244
2225
3808
3633
3318
2722
2005
2567
2516
2078
1438
1337
1348
999
842
779
3192
1974
1834
2829
2396
1742
1698
1804
1418
1305
1309
1145
941
752
603
427
434
370
2289
23312
1982
1291
2099
2183
3817
3621
3297
2689
1950
2469
2399
1942
1364
1258
1256
930
779
3628
1975
1326
2739
2848
2384
1730
1668
1779
1395
1273
1256
1085
884
697
554
388
391
2393
24813
1983
1564
2214
2106
2160
3779
3566
3230
2615
1887
2368
2247
1848
1288
1184
1177
870
4026
TOTAL 26560 28222 30565 32583 35268 36069 36987 38143
-------
B-5
Table B-3 Year to Year Survival Rates for Cars(historical)
Year
Age of
Car
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16+
Year
Age of
Car
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16+
1968
-1969
1.242
.992
.996
.991
.982
.968
.939
.878
.807
.733
.695
.714
.707
.711
.707
2.533
1976
-1977
1.260
.973
.984
.975
.969
.954
.935
.896
.855
.824
.795
.771
.764
.756
.755
3.979
1969
-1970
1.284
.988
.978
.970
.961
.945
.918
.871
.799
.745
.690
.673
.701
.697
.708
3.015
1977
-1978
1.238
.992
.975
.983
.973
.956
.931
.911
.874
.839
.813
.796
.784
.785
.776
3.051
1970
-1971
1.411
.998
.998
.987
.974
.961
.935
.897
.845
.773
.733
.699
.705
.733
.730
2.396
1978
-1979
1.261
.984
.970
.959
.955
.944
.922
.897
.871
.827
.813
.797
.795
.791
.792
2.870
1971
-1972
1.161
.996
.983
.977
.965
.954
.928
.890
.838
.782
.722
.694
.680
.691
.721
3.126
1979
-1980
1.299
.980
.972
.967
.956
.944
.918
.895
.860
.836
.804
.793
.797
.800
.809
2.873
1972
-1973
1.238
.978
.973
.974
.965
.949
.925
.886
.838
.785
.739
.695
.684
.681
.704
3.384
1980
-1981
1.309
.985
.982
.983
.981
.967
.947
.924
.900
.872
.858
.832
.831
.838
.840
2.867
1973
-1974
1.315
.999
.989
.986
.970
.957
.930
.891
.850
.801
.764
.734
.711
.705
.714
5.916
1981
-1982
1.293
1.001
.983
.987
.990
.970
.958
.931
.902
.882
.861
.845
.838
.838
.842
3.156
1974
-1975
1.470
1.006
.995
.992
.982
.968
.953
.923
.887
.845
.813
.789
.776
.765
.767
4.455
1982
-1983
1.241
.999
.991
.994
.988
.977
.967
.948
.918
.891
.875
.856
.852
.849
.846
3.752
1975
-1976
1.188
.998
.982
.978
.965
.955
.932
.907
.877
.843
.813
.797
.781
.770
.763
3.715
-------
B-6
Table B-4 Truck Year to Year Survival Rates (historical;
Year
Age of
Truck
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16+
Year
Age of
Truck
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16+
1968
-1969
1.151
.997
.989
.983
.974
.970
.967
.955
.943
.927
.913
.904
.893
.894
.892
5.275
1976
-1977
1.123
.982
.984
.983
.977
.966
.962
.950
.942
.925
.920
.900
.889
.882
.879
6.900
1969
-1970
1.273
.972
.987
.992
.987
.980
.971
.960
.939
.932
.922
.910
.908
.902
.900
6.903
1977
-1978
1.149
.999
.984
.988
.974
.972
.956
.959
.945
.938
.924
.911
.902
.901
.894
5.726
1970
-1971
1.278
.995
.974
.979
.972
.969
.964
.951
.937
.942
.927
.913
.906
.902
.891
5.516
1978
-1979
1.132
.997
.977
.966
.975
.965
.961
.958
.951
.939
.934
.919
.908
.900
.897
5.310
1971
-1972
1.133
1.005
.993
.981
.981
.978
.971
.959
.946
.932
.897
.892
.885
.873
.872
6.450
1979
-1980
1.414
1.034
1.031
1.027
1.020
1.020
1.009
1.011
1.026
.996
.995
.974
.959
.945
.919
4.955
1972
-1973
1.191
.983
.980
.982
.985
.974
.967
.959
.943
.932
.920
.903
.895
.890
.884
7.183
1980
-1981
1.187
1.011
.992
.996
.990
.980
.972
.964
.961
.934
.936
.928
.923
.916
.910
4.640
1973
-1974
1.171
1.005
.994
.994
.988
.981
.977
.960
.948
.936
.920
.910
.901
.891
.883
8.541
1981
-1982
1.233
.981
1.002
.997
.994
.988
.973
.962
.953
.935
.949
.941
.932
.931
.925
4.657
1974
-1975
1.333
1.007
.995
.993
.982
.986
.984
.975
.960
.948
.939
.927
.919
.909
.901
6.468
1982
-1983
1.153
1.003
.989
.990
.985
.980
.972
.968
.959
.937
.952
.944
.941
.937
.935
5.168
1975
-1976
1.123
.997
.983
.984
.981
.980
.973
.964
.958
.948
.944
.937
.921
.908
.905
6.279
-------
B-7
forecast of sales and fleet size for cars. Since DRI does not
forecast fleet size for trucks, we modified the adjustment pro-
cedure for truck year to year survival rates to use car scrappage
rates. A detailed description of the adjustment procedure is
given in the next two sections.
B. 1.a Projection Model for Cars on the Road
We begin with a description of the projection procedure for
cars. Let C^j denote cars of age i in year j. Thus C^o,1985
represents the number of ten year old cars in 1985. Then,
letting S^j be the probability of a car of age i in year j still
being on the road in year j + 1, we have C^ + i ,j + i= C^jX S-[j.
Further, we have
RJ = SUM Cj_jx(l - Sij )
i
where RJ is total retirements in the fleet in year j. However,
we also have
R-J = Fj + 1 - Fj + N-J,
where Fj is the size of the fleet in year j and Nj is new car
sales in that year. Thus, we have two equations which determine
retirements and these must be consistent if the model is to
operate properly. We used the DRI forecast as the source of the
overall loss rates and adjusted year to year survival rates until
they gave the same total losses as the DRI forecast.
We adjusted age specific survival rates by using the
regression coefficient of the scrappage rate in the model
(B.I) Sij = a0 + bSRj + c(Cij/SUM Cij) - e
where SR-; is the overall scrappage rate for year j and
-------
B-8
Ci;j/SUM Vjj is the fraction of the fleet that vintage i is of the
fleet in year j. The survival rates were varied by changing the
scrappage rate in equation B.I until the total losses from the
survival rates were within 0.01% of those in the DRI forecast.
In equation B.I the scrappage rate value necessary to achieve
equality in total losses with the DRI forecast was typically
larger than the DRI scrappage rate. Thus, the survival rate
model overpredicts the size of the fleet in the next year when
compared to the DRI forecast. However, the advantage of adjusting
using the model is that the survival rates maintain a consistent
pattern by vehicle age as model forecasts of retirements from
each cohort in the fleet are changed in order to obtain equality
with the DRI forecast.
For the car fleet, the adjustment equations were estimated
using sixteen years of data from the MVMA for cars on the road in
July of each year. The resulting estimates are given in Table
B-5. For forecasting purposes, the results of the regression
analysis were not used for the first two age groups. Because new
cars on the road on July 1st of the year fluctuated sharply as a
percentage of total car sales for the model year, we assumed they
were equal to 75% of total sales for that model year. One year
old cars were assummed to be 99.9 percent of the appropriate new
car model year sales and two year old cars were assumed to be
99.5 percent of the original sales level. One and two year old
cars show little variation in survival rates and the correction
mechanism occassionally gave survival rates greater than one —
therefore these rates were imposed.
-------
B-9
Table B-5.
Regression Coeficients for the Car Year to Year
Survival Rates
Survival
Rate
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
Constant
1.426
1.009
.9854
1.019
.9967
.9857
1.013
1.013
.8523
.5212
.5580
.6533
.6067
.6810
.7057
Scrappage
Rate
-.000239
-.003030
-.003828
-.004151
-.004715
-.004911
-.006537
-.006537
-.01218
-.001298
-.004541
-.008412
-.000991
-.005657
-.004773
Cohort
Weight
-1.880
.0721
.2983
-.05617
.1383
.1298
-.3060
-.7581
1.619
5.467
5.969
5.550
6.832
7.202
7.637
-------
B-10
The other two pieces of information necessary to start the
forecast procedure were a 1983 inventory of vehicles and an
initial set of survival rates. The initial inventory data were
extended from the 15 years available in the data to 30 years.
The extension was necessary because consumers have been keeping
their cars longer, and consequently, older vehicle have been
steadily growing as a fraction of the fleet. As Tables B-l and
B-2 indicate, the cohort of vehicles aged 16 years old and
greater has been growing in size during the last 15 years, in
both absolute and relative terms.
We began the inventory extension by noting, as Tables 3 and
4 show, that the survival rates for vehicles aged 11 through 15
were similar in size and showed a consistent pattern of change
from year to year. Therefore, we assumed that the survival rate
for older vehicles was equal to the rate for 14 to 15 year-old
cars and that it changed in the same way as this rate did in
response to a change in the overall scrappage rate. Using this
assumption, we constructed our estimates of 15 through 30 year
old cars in 1983 iteratively. We began in 1968 with 15 year old
cars and estimated the number of 16 year old cars in 1969 using
the survival rate for 14 to 15 year old cars for 1968 to 1969.
Thus, in 1969 we had an estimate for 15 and 16 year old cars. We
then repeated the same process for two cohorts to get estimates
of 15, 16 and 17 year-old cars in 1970. A similar process was
carried on through 1983 to give us estimates of cars 15 through
30 years of age. In 1983, the total number of cars resulting
from our estimate of the individual cohorts was 5 percent less
-------
B-ll
than the actual total for cars over 15 years of age. We corrected
for this error by adjusting each age group by
(B.2) (1 + r)i - 15,
where i is the age of the cohort and r was chosen to yield the
desired equality. This adjusts older vehicles more than younger
vehicles to reflect their longer extrapolation period. The value
for r was .0181 for cars. The largest correction, for the oldest
cohort, was a 31 percent increase.
The other required set of initial information was the
1983-1984 survival rates. These were the 82 to 83 set with
rates for age 15 and older cohorts set to 0.84. This was the
rate for 14 year old cars in 1983. This was was used because the
assumption had been used in the inventory extension process and
had worked quite well .
With these two inputs determined, the model could forecast
the total number cars on the road for each year from 1985 through
1992; the results are in Table B-6. As the table shows the
number of older vehicles continues to grow. By 1992, cohorts
aged 16 and older exceed 12 million cars and are nearly 10
percent of the fleet. While this is to some extent an artifact
of the extrapolation procedure, the overall size of the pool of
cars over 15 years of age is consistent with existing behavior
within the fleet and the DRI forecast of cars on the road.
B.l.b Modification of the Model for Trucks
For trucks, it was necessary to modify the car forecasting
procedure because DRI does not provide a forecast of the size of
-------
B-12
Table B-6 Predicted Number of Cars on the Road 1984-1992
Year 1984 1985 1986 1987 1988
1989
1990 1991 1992
Age
Car
NEW
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
of
7950
9191
7392
8117
8526
9620
9545
8741
7345
4988
5993
5715
4317
3070
2525
1973
1539
1326
979
861
650
422
279
172
86
67
36
18
26
21
20
7950
10589
9145
7256
7915
8196
9154
8852
7842
6257
4342
5042
4716
3650
2543
2082
1626
1268
1092
807
709
536
348
230
142
71
55
30
15
21
17
7875
10589
10536
9016
7109
7651
7844
8554
8007
6787
5456
3675
4208
3993
3046
2111
1727
1349
1052
906
669
588
445
289
191
118
59
46
25
12
17
8100
10490
10536
10408
8852
6888
7341
7355
7765
6978
5923
4631
3083
3565
3344
2535
1756
1437
1122
875
754
557
489
370
240
159
98
49
38
21
10
8400
10789
10438
10396
10207
8565
6600
6870
6663
6741
6087
5019
3874
2611
2980
2779
2106
1459
1194
932
727
626
463
406
307
199
132
81
41
32
17
8625
11189
10735
10288
10182
9862
8194
6164
6211
5760
5878
5150
4186
3279
2178
2472
2304
1746
1210
990
773
603
519
384
337
255
165
109
67
34
27
8850
11489
11133
10558
10053
9812
9409
7625
5550
5328
5018
4959
4271
3541
2725
1801
2043
1904
1443
1000
818
639
498
429
317
278
211
136
90
55
28
9000
11788
11432
10936
10303
9673
9346
8737
6850
4740
4640
4226
4100
3612
2937
2250
1486
1685
1570
1190
825
675
527
411
354
261
229
174
112
74
45
9075
11988
11729
11219
10661
9901
9202
8664
7835
5830
4126
3903
3485
3466
2990
2421
1853
1224
1388
1293
980
679
556
434
338
292
215
189
143
92
61
Tot 111510 112498 113950 115769 117741 119876 122011 124188 126232
-------
B-13
the truck fleet. Rather than attempt to develop an independent
estimate of the total size of the fleet, we assumed that the year
to year survival rates were a function of the scrappage rates
observed in the car fleet. Since these can he derived from the
DRI forecast and from historical data, we can estimate the size
of the fleet by adding trucks surviving from the previous year to
new trucks delivered. The precise details are given below.
We regressed truck survival rates on car scrappage rates
and the fraction of the fleet that the truck cohort of age i was
of all trucks. Thus,
(B.3) SijT = a + bSRjfCars + c(Tij/SUM TIj) + e
i
Because the truck fleet data for 1980 reflect the reclassifi-
cation of all passenger vans from cars to trucks, we excluded
the survival rates from 1979 to 1980 from the data base used for
the regressions. The regression results are presented in Table
B-7. The regression coefficients for SRj are used to adjust
the survival rates in forecast years for each truck cohort.
Aggregate fleet size is the sum of the individual cohorts plus
new truck sales for the year. As in the car model, new trucks
on the road were assumed to be 75 percent of model year sales
and one year-old trucks were assumed to be 99.9 percent of orig-
inal sales. Trucks on the road were extended to 30 years of
age using the same process as for cars. The extrapolation
resulted in a 17 percent overestimate of the fleet between 16
and 30 years of age. The adjustment coefficient (r) in equation
B.2 was -.019, which resulted in a 25 percent reduction in the
oldest cohort.
-------
B-14
Table B-7 Regression Coefficients for Truck Year to Year
Survival Rates
Survival
Rate
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
Constant
1.227
1.018
1.000
.9843
.9504
.9824
.9799
.9589
.9509
.9503
.8771
.8680
.8721
.8759
.8784
Scrappage
Rate
.007742
-.004245
-.000509
-.000364
.001117
-.000491
-.000722
.000908
-.002385
-.001290
-.001353
-.002685
-.001531
-.002938
-.003051
Cohort
Weight
-1.079
.1060
-.08927
.05054
.2416
-.02511
-.06452
.1289
.2915
-.07566
1.326
1.813
1.483
1.756
1.781
-------
B-15
The initial set of survival rates for trucks for 1983-84 was
taken to be the same as the 82-83 rates. However, the rates for
cohorts age 16 and greater were not set equal to the 14-15 rate
for 1983 for two reasons. First, in the inventory extrapolation,
it was necessary to revise the older cohorts downward significantly
to make them equal to the number of trucks 16 years of age and
older in 1983. Secondly, there was no overall forecast to con-
strain the year to year survival rates, so it was essential to
reflect the information from the extrapolation process in the
survival rates. The rates were set to decline from 0.93 to 0.82
as cohorts ranged from 16 to 30 years of age. As seen in Table
B-8, even with this reduction, older trucks become about one
sixth of the fleet by 1992. Further, the total number of trucks
increases substantially relative to cars during the forecast
period.
Once total trucks on the road were determined, several
additional adjustments to the data were made. First, total
trucks on the road had to be disaggregated by weight class.
Trucks were divided into six weight classes, derived from the
usual eight classes by combining classes three through five into
one class.' We did this because classes three(10,000 to 14,000
pounds), four (14,000 to 16,000 pounds) and five (16,000 to
19,500 pounds) are very small. Currently, they constitute less
than one tenth of 1 percent of all trucks sold.
The adjustment procedure required the weight composition of
new trucks on the road in every year from 1954 to 1992. For the
years from 1968 through 1983, we used published data on truck
-------
B-16
Table
Year
B-8
1984
Predi.ted Trucks on the Road 1984-1992
198F
1986
1987
1988
1989
1990
1991
1992
Age of
Truck
New
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
3139
3139
2189
2084
2130
3700
3464
3122
2500
1764
2249
2114
1736
1202
1103
1090
801
731
591
525
415
314
232
173
120
117
95
59
61
60
334 )
40 8 i
310.1
2165
2058
2084
3591
3345
2975
2333
1673
2109
1981
1613
1115
1017
1000
731
663
533
471
371
277
202
149
103
99
79
48
50
3275
4299
4034
3069
2137
2014
2024
3469
3192
2779
2214
1571
1978
1845
1500
1030
935
914
664
600
480
421
328
243
175
127
87
82
65
39
3468
4243
4250
3991
3029
2092
1956
1955
3312
2982
2638
2080
1474
1843
1716
1386
947
855
831
601
539
429
372
286
210
149
107
72
68
53
3670
4499
4194
4205
3939
2964
2031
1890
1867
3094
2830
2477
1952
1373
1714
1586
1274
866
111
751
540
482
379
325
248
179
126
90
60
55
3830
4757
4447
4150
4150
3856
2879
1962
1804
1744
2936
2658
2325
1818
1277
1584
1458
1165
788
703
676
483
426
332
281
212
151
105
74
48
3995
4963
4702
4400
4096
4062
3744
2780
1873
1685
1655
2757
2494
2165
1690
1180
1456
1333
1059
712
632
604
427
373
286
240
179
126
87
60
4148
5171
4905
4652
4343
4009
3944
3616
2654
1749
1599
1554
2587
2323
2012
1561
1084
1330
1211
957
640
565
534
373
322
245
203
149
104
71
4295
5368
5111
4853
4592
4250
3892
3809
3451
2479
1660
1501
1458
2409
2158
1859
1435
991
1209
1095
861
572
499
467
323
275
206
169
123
85
Tot 41020 43360 45590 47938 50440 53078 55812 58614 61453
-------
B-17
sales by weight class. These data were converted to percentages
and applied to new trucks on the road to yield estimates of the
fleet by weight. For the years before 1968, we assumed a constant
composition of the fleet at 1968 levels. For the years after
1983, we used DRl's decomposition of new truck sales into
light trucks (those under 14000 pounds) and medium and heavy
trucks (those over 14,000 pounds). We broke the light truck
category into two classes light duty trucks classes 1 and 2. We
used the 1981-1983 average ratio of class 1 to class 2 trucks and
applied it to the DRI light trucks forecast to derive these two
categories of trucks. We assumed there were no trucks between 10
and 14,000 pounds (1983 sales were 145). Similarly, we applied
the average ratios for the other four weight categories to get a
breakdown of the DRI forecast of medium and heavy trucks. The
resulting percentage composition of the fleet is presented in
Table B-9.
The other adjustment that we made in the data was to estimate
the fraction of each class of truck that was diesel; we removed
these from the fleet. For the years 1968 through 1983, we esti-
mated the fraction of truck sales that were diesel from actual
data. Because it is difficult to find a coherent set of data on
all aspects of truck sales, we used total factory sales of diesels
for the numerator and total sales of trucks for the denominator;
this may understate the diesel fraction somewhat. For the years
before 1968 we used the data in column two of Table B-10 as the
fraction of each of the six truck classes that were diesel. These
correspond to the actual ratios in 1965.
-------
B-18
Table B-9 Fraction of Truck Fleet in Each Weight Class
Year
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
1974
1973
1972
1971
1970
1969
1968
1967
1966
1965
1964
1963
1962
1961
1960
1959
1958
1957
1956
1955
1954
LDTl
.4874
.4874
.4874
.4865
.4862
.4848
.4833
.4815
.4814
.4824
.4879
.4899
.4850
.4901
.4543
.4415
.3929
.3409
.3746
.4331
.4684
.5635
.5825
.5781
.5828
.5615
.5830
.5992
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
LDT2
.4448
.4448
.4448
.4440
.4437
.4424
.4410
.4395
.4393
.4403
.4453
.4471
.4453
.4276
.4309
.4368
.4865
.5467
.5172
.4602
.4048
.2682
.2556
.2389
.2369
.2373
.2106
.2035
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
10,000
-19,500
.0007
.0007
.0007
.0007
.0008
.0008
.0008
.0008
.0009
.0008
.0007
.0007
.0005
.0011
.0014
.0027
.0065
.0209
.0128
.0172
.0142
.0097
.0239
.0336
.0442
.0452
.0514
.0536
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
19,500
-26,000
.0182
.0182
.0181
.0186
.0188
.0195
.0203
.0212
.0212
.0207
.0179
.0169
.0172
.0197
.0365
.0402
.0451
.0398
.0469
.0502
.0675
.0823
.0682
.0744
.0644
.0736
.0766
.0745
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
26,000
-33,000
.0199
.0199
.0198
.0204
.0205
.0213
.0222
.0232
.0232
.0227
.0196
.0185
.0219
.0278
.0261
.0262
.0153
.0105
.0082
.0073
.0098
.0126
.0142
.0173
.0177
.0227
.0173
.0221
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
33,000
plus
.0291
.0290
.0290
.0298
.0300
.0312
.0324
.0338
.0340
.0331
.0286
.0270
.0301
.0337
.0509
.0526
.0536
.0413
.0404
.0320
.0354
.0636
.0557
.0577
.0539
.0597
.0610
.0472
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
-------
B-19
Table B-10. Fraction of Trucks that are Diesel by Weight Class
Weight
Class
0-6,000
6,000-10,000
10,000-19,500
19,500-26,000
26,000-33,000
33,000+
Before
1968
0.0
0.001
0.05
0.06
0.45
0.80
After
1983
0.02
.10 to .135
.00
.24
.54
.975
For the years after 1983, we used the values given in column 3
of Table B-10 as the fraction of the fleet that will be diesel.
These forecasts are the result of an inspection of historical
data on diesel sales ratios and an overall forecast that the
price of gasoline and diesel fuel relative to other goods is not
going to rise significantly. Class two trucks are the only
class that gives any indication of a growth in diesel penetration
at the current time. Therefore, we increased diesel penetration
rates by 0.05 per year for this class.
B.2 Miles Driven per Year by Vehicle Class
Data for cars and light duty trucks on miles driven per year
are taken directly from the Transportation Energy Data Book
(Department of Energy, 1982). For heavier trucks, only average
miles per year for all trucks in the class are available in this
reference. To develop a representation of how miles per year
declined with the age of these vehicles, we used data from
-------
B-20
Jambekar and Johnson (1978), and normalized this to the average
found in the DOE report.
An important division of gasoline fueled vehicles for our
purposes was the extent to which vehicles designed for unleaded
fuel were misfueled with leaded gasoline. This was important to
compute both the maintenance benefits and the benefits due to the
reduction of emissions from engines that would have been misfueled
in the absence of the rule. To calculate this, we estimated the
extent of misfueling by vehicle age from EPA's 1983 survey of
tampering and misfueling. This data was quite variable when
broken down by age of vehicle, so we smoothed it by fitting
a regression. The resulting regression is
MSFR = .0326 + .02314 T
where MSFR is the misfueling rate by vehicle age and T is the age
of the vehicle (T = 0 for new vehicles). The truck sample size
in the survey was too small to support a separate estimate for
them. Therefore, we used the same regression for trucks and
cars although the available data suggests that trucks misfuel at
a higher rate than cars.
B.3 Total Miles per Year for Maintenance Benefits
For maintenance benefits we had to compute the total mileage
driven by both misfuelers and legal leaded-gasoline users in a
given year. Miles driven by misfuelers were estimated by com-
puting the fraction of misfuelers in each model year using the
regression given above. Cars and light duty trucks over 10 years
of age were assumed to experience no further increase in misfueling.
-------
B-21
For legal leaded users, we used estimates of the legal leaded
fleet and the corresponding estimates of miles diriven per year.
In Table B-ll, we present the estimated miles per year traveled
by cars and by light duty trucks classes 1 and 2 by leaded,
misfueled, and unleaded status. Because we computed no maintenance
benefits for heavier trucks or other gasoline powered vehicles,
no estimates of usage are provided in this table. The results
reflect the decline in the size of the fleet of legal leaded-gas-
oline users; the results also show how miles traveled by misfuelers
are projected to grow over time.
B.4 Miles per Year for Conventional Pollutant Emissions
Emissions due to misfueling constitute the other major source
of benefits calculated on the basis of the fleet model results.
For conventional benefits estimates, we had to calculate the
increase in misfueling for each model year in a given calendar
year. For each year, the incremental increase in misfueling
for a given model year is given by the slope coefficient. Since
we restricted the increase in overall misfueling so that misfueling
was constant for vehicles 10 years of age and older, there is no
increase in misfueling for cars more than 10 years of age and,
thus, no conventional pollutant benefits are attributed to these
vehicles. We then computed the expected number of miles the
additional misfuelers will travel each year for the next 20
years. These estimates are corrected to reflect expected loss
rates and reductions in miles traveled as the vehicle ages. The
resulting estimates are present valued to the year of initial
-------
B-22
Table B-ll
Mileage Estimates by Misfueling Status and Type of
Vehicle (millions of miles per year)
Year
1984
1985
1986
1987
1988
1989
LEGAL LEADED
CARS
LDT1
LDT2
MISFUELED
CARS
LDT1
LDT2
UNLEADED
CARS
LDT1
LOTS
Year
LEGAL LEADED
CARS
LDT1
LDT2
MISFUELED
CARS
LDT1
LDT2
UNLEADED
CARS 1
LDT1
LDT2
312.454
50.772
77.076
119.747
13.264
6.612
861.141
110.594
70.709
1990
76.682
22.638
55.804
181.770
27.194
17.965
162.103
199.305
136.531
250.001
44.513
71.901
132.948
15.494
8.413
921.376
128.196
84.593
1991
59.441
19.645
54.239
189.069
29.746
19.874
1200.156
213.001
146.108
199.925 159.257
39.015 34.124
67.040 62.740
144.940 155.805
17.742 20.020
10.312 12.299
974.934 1025.857
142.936 157.037
96.027 106.681
1992
46.191
17.092
53.122
195.837
32.328
21.703
1233.052
226.255
154.995
125.563 98.420
29.796 26.000
59.971 57.729
165.381 173.989
22.351 24.724
14.185 16.062
1074.616 1120.383
171.227 185.365
116.523 126.642
-------
B-23
misfueling using a 10 percent discount rate. Total lifetime
miles driven by new misfuelers in each year are given in Table
B-12. The total miles are then multiplied by estimates of the
increase in emissions due to poisoning of the catalyst. These
increases vary by model year and are given in Table VI-1. The
resulting increases in total emissions are given in Table VI-3.
-------
B-24
Table B-12 Discounted Mileage Traveled in Following Twenty Years
by Vehicles that Initially Misfuel in Given Year
(millions of vehicle miles)
Year
1985
1986
1987
1988
1989
1990
1991
1992
Cars
132,104
135,764
139,178
143,644
147,955
152,671
157,174
161 ,523
LDT1
17,917
19,309
20,812
22,398
23,899
25,710
27,478
29,134
LDT2
12,032
13,509
14,973
16,392
16,948
17,986
19,046
20,076
-------
APPENDIX C
REGRESSION RESULTS
This appendix consists of four sectiions. Part C.I contains
of a description of the variables used in the regressions reported
in later sections. Part C.2 contains logistic regressions for
white and black preteens, obtained using PROC LOGIST, a statistical
software package available within the Statistical Analysis System
(SAS). The third section, Part C.3, presents linear regression
results for whites and blacks obtained by application of SURREGR,
a statistical routine available as an adjunct of SAS which corrects
regression results for the complex sample survey design used in
the NHANES data collection. Part 4 of appendix C contains regres-
sion results where the solder content of cans has been added to
the variables used in the gas lead blood lead regressions reported
in Chapter Three .
C.I Variable Descriptions
In addition to the regressions shown in Chapter Three, we
have used the regressions presented in this appendix in our
forecasts of child health effects. We have used the following
variables in these regressions:
Variable Name Description
Gaslead Lead Used in gasoline, in hundreds of
metric tons a day, lagged one month.
Poor 1 if Income l(see below); 0 otherwise
Age 1 1 is age >^ 6 months and < 2 years;
0 otherwise
Age 2 1 if age ^> 2 years and < 4 years;
0 otherwise
-------
C-2
Age 3
Age 4
Age 5
Age 6
Age 7
Teen
Income 1
Income 2
Male
Teen Male
Adult Male
Small City
Rural
Drinker
Heavy Drinker
Northeast, Mideast,
South
1 if age _> 4 years and < 6 years;
0 otherwise
1 if age >_ 6 years and < 8 years;
0 otherwise
1 if age >_ 8 years and < 10 years;
0 otherwise
1 if age _> 10 years and < 12 years;
0 otherwise
1 if age ^> 12 years and < 14 years;
0 otherwise
1 if age _> 14 years and < 18 years;
0 otherwise
1 if family income < $6,000; 0
otherwise
1 if family income > $6,000 and <
$15,000; 0 otherwise
1 if gender is male; 0 otherwise
1 if gender is male and age >^14 and
< 18 years; 0 otherwise
1 if gender is male and age >^18 years;
0 otherwise
1 if residence is in city with
population _> 1,000,000; 0 otherwise
1 if residence is in a rural area as
defined by the Bureau of the Census;
0 otherwise
1 if alcohol consumption is _>. 1 drink/
week and < 1 drink/day; 0 otherwise
1 if alcohol consumption is _>. 1
drink/day; 0 otherwise ~~
1 if in this Census Region; 0 otherwise
-------
C-3
Education 0 if the person never completed grade
school; 1 if grade school was the highest
level completed; 2 if high school was
the highest level completed; and 3 if
college was completed
Kid 1 if age < 6; 0 otherwise
-------
C-4
C.2 Logistic Regression Results
Black Preteens = 8-13 years old, 112 observations
Dependent variable; 1 if blood lead is over 20 ug/dl; 0 otherwise
Model Chi-square =6.42 with 4 D.F-
Variable
Intercept
Gaslead
Poor
Age 5
Age 6
Beta
-6.0148
0.9786
0.2356
0.6158
0.2397
Std. Error
2.4044
0.4943
0.5289
0.6304
0.6208
Chi-Square
6.26
3.92
0.20
0.95
0.15
P-Value
0.0124
0.0477
0.6560
0.3286
0.6994
Fraction of concordant pairs of predicted probabilities and responses = 0.656
White Preteens = 8-13 years old, 660 observations
Dependent variable: 1 if blood lead is over 20 ug/dl; 0 otherwise
Model Chi-square = 21.35 with 4 D.F.
Variable
Intercept
Gaslead
Poor
Age 5
Age 6
Beta
-8.9395
1.0674
0.8355
1.4199
1.2041
Std. Error
1.6782
0.3374
0.4883
0.5810
0.5904
Chi-Square
28.38
10.01
2.93
5.97
4.16
P-Value
0.0000
0.0016
0.0871
0.0145
0.0414
Fraction of concordant pairs of predicted probabilities and responses = 0.710
All logistic regression results were run using PRDC LOGISTIC within the Statistical
Analysis System (SAS). This procedure uses individual data where the dependent
variable is 1 if the individual is above the threshold, and 0 otherwise.
-------
C-5
APPENDIX C.3. Linear Regression Results'
Whites :
Dependent variable:
Variable
Intercept
Gaslead
Income 1
Income 2
Age 1
Age 2
Age 3
Age 4
Teen
Male
Teen Male
Adule Male
Small City
Rural
Drinker
Heavy DrinKer
Northeast
Midwest
South
Education Level
individual
Beta
5.4436
2.1835
0.7675
0.3381
3.2352
4.0452
3.2020
2.1818
-0.7386
0.5763
1.7556
3.9812
-0.8490
-1.3215
0.8582
2.0871
-1.0908
-1.2243
-1.0598
-0.9440
blood lead levels
Std. Error
1.1842
0.0345
0.0553
0.0288
0.2015
0.1713
0.1267
0.2118
0.0519
0.1040
0.2150
0.1203
0.1080
0.1188
0.0296
0.0889
0.1302
0.1631
0.2493
0.0182
F-Statistic
—
138.19
10.65
3.97
51.95
95.51
80.91
22.48
10.52
3.19
14.34
131.72
6.67
14.70
24.84
48.97
9.14
9.19
4.51
48.90
P-Value
—
0.0000
0.0026
0.0548
0.0000
0.0000
0.0000
0.0000
0.0028
0.0834
0.0006
0.0000
0.0146
0.0006
0.0000
0.0000
0.0049
0.0048
0.0416
0.0000
* Procedure used is SURREGR in SAS.
-------
06
Whites:
Dependent variable:
Variable
Intercept
Gas lead
Income 1
Income 2
Kid
Teen Male
Rural
Small City
Teen
Male
Adult Male
Age 4
Age 5
Age 6
Age 7
Drinker
Heavy Drinker
Northeast
Midwest
South
Education Level
individual
Beta
5.4593
2.1821
0.7542
0.3386
3.2344
2.0860
-1.3350
-0.8443
-1.5987
1.1333
3.4231
1.8952
0.5581
0.4784
0.3958
0.8672
2.0789
-1.0823
-1.2414
-1.0619
-0.9461
blood lead levels
Std. Error
1.1766
0.0344
0.0559
0.0284
0.0926
0.2093
0.1221
0.1098
0.0910
0.0348
0.0504
0.2205
0.1126
0.1629
0.0727
0.0303
0.0894
0.1312
0.1663
0.2504
0.1808
F-Statistic
—
138.53
10.17
4.04
112.97
20.79
14.59
6.49
28.08
36.90
232.33
16.29
2.77
1.41
2.15
24.92
48.35
8.92
9.27
4.50
49.51
P-Value
—
—
0.0000
0.0032
0.0531
0.0000
0.0001
0.0006
0.0159
0.0000
0.0000
0.0003
0.1060
0.2445
0.1520
0.0000
0.0000
0.0054
0.0046
0.0417
0.0000
-------
C-7
Blacks:
Dependent variable:
Variable
Intercept
Gaslead
Income 1
Income 2
Age 1
Age 2
Age 3
Age 4
Teen
Male
Adule Male
Drinker
Heavy Drinker
Education Level
individual
Beta
4.8847
1.9342
1.1457
1.0941
6.1030
8.8867
6.6989
4.8920
0.6352
1.8280
4.2469
1.0359
1.4088
-0.8329
blood lead levels
Std. Error
2.4116
0.1432
2.2593
0.2902
1.3729
0.5052
0.4592
0.7706
0.1869
0.3413
0.6157
0.4713
1.2531
0.0874
F-Statistic
—
26.12
5.06
4.13
27.13
156.32
97.73
31.06
2.16
9.79
29.29
2.28
1.58
7.93
P-Value
—
0.0000
0.0328
0.0522
0.0000
0.0000
0.0000
0.0000
0.1533
0.0042
0.0000
0.1429
0.2190
0.0090
-------
C-8
Blacks:
Dependent variable:
Variable
Intercept
Gas lead
Income 1
Income 2
Kid
Teen
Male
Adule Male
Age 4
Age 5
Age 6
Age 7
Drinker
Heavy Drinker
Northeast
Midwest
South
Education Level
individual
Beta
4.795
2.041
1.016
1.063
7.204
-0.806
1.860
4.061
4.869
2.494
2.215
0.417
1.063
1.386
-1.460
0.145
-0.1173
-0.826
blood lead levels
Std. Error
2.48
0.12
0.26
0.33
0.29
0.35
0.24
0.48
0.81
1.10
0.44
0.59
0.44
1.17
0.84
1.05
0.501
0.086
F-Statistic
—
33.84
3.90
3.44
180.91
1.84
14.39
34.08
29.22
5.67
11.07
0.30
3.03
1.64
2.53
0.02
0.03
7.91
P-Value
—
0.0000
0.0587
0.0748
0.0000
0.1857
0.0008
0.0000
0.0000
0.0246
0.0025
0.5910
0.0933
0.2117
0.1230
0.8885
0.8695
0.0091
-------
C-9
APPENDIX C.4. Regression Results: Whites with Solder Used in Cans
Effect Coefficient
Intercept
Gasoline Lead
Low Income
Moderate Income
Child
Solder
Teenager
Male
Male Teenager
Adult Male
Small City
Rural
Drinker
Heavy Drinker
Northeast
South
Midwest
Education Level
•6U.S. GOVERNMENT PRINTING OFFICE: 1965 526 259 30280
Standard Error
P-Value
4.16
2.16
0.78
0.36
3.47
0.74
-0.32
0.68
1.38
3.89
-0.82
-1.32
0.83
2.09
-1.02
-1.15
0.88
-0.93
—
0.04
0.06
0.03
0.11
1.01
0.05
0.08
0.13
0.11
0.11
0.12
0.03
0.09
0.13
0.17
0.29
0.02
—
0.0000
0.0024
0.0383
0.0000
0.4651
0.1667
0.0251
0.0005
0.0000
0.0212
0.0007
0.0000
0.0000
0.0091
0.0084
0.1093
0.0000
------- |