United States
Environmental Protection
Agency
Office of
Poicy Analysis
Washington, DC 20460
March 1984
Draft Final Report
EPA-230-03-84-OOE.
Costs and Benefits of
Reducing Lead in Gasoline
-------
COSTS AND BENEFITS
OF REDUCING
LEAD IN GASOLINE
by
Joel Schwartz
Jane Leggett
Bart Ostro
Hugh Pitcher
Ronnie Levin
Draft Final Report
Office of Policy Analysis
Office of Policy/ Planning and Evaluation
U.S. Environmental Protection Agency
Washington, D.C. 20460
March 26, 1984
(minor corrections done 5/84)
-------
ACKNOWLEDGEMENTS
We wish to publically acknowledge the extraordinary support we received
from many people in doing this analysis. Although the list is too long to name
them all, we wish to recognize the special assistance, expertise, and effort of
| certain individuals. First, George Sugiyama, Bob Fegley, and Albert Nichols
§) have provided invaluable assistance, analysis, and general acumen. We are deeply
0 indebted to them.
* For technical assistance, even under strict time strictures, we thank
^ Craig Miller (of Energy and Resource Consultants); Steve Sobotka, Bill Johnson,
] and Terry Higgins (of Sobotka and Company); and Ed Fu.
For secretarial and production support well above the call of duty, we
thank Saundra Womack, Joyce Morrison, Delores Thompson, Sylvia Anderson, and
Ethel Stokes. For research assistance, we thank James Chow, and for help with
the process and general principles, we thank Marty Wagner.
For specialized and professional help, we wish to thank Les Grant,
Barry Nussbaum, Karl Hellman, Dan Salisbury, Susan Martin, and members of the
staff of the Office of Air Quality Planning and Standards. For extensive review
and comments, we thank Valle Nazar of the Office of Policy Analysis and Review.
For reviewing our statistical methodologies we would like to thank Kathleen Knox
and John Warren of EPA's Statistical Policy Staff. We also thank Terry Yosie,
the director of the Science Advisory Board, for arranging for the external peer
review of an earlier draft of this document, and the nine reviewers for their
valuable insights.
-------
TABLE OF CONTENTS
EXECUTIVE SUMMARY
CHAPTER I: INTRODUCTION, FINDINGS, AND QUALIFICATIONS
I. A. Background I. 1
I.E. Approach I. 4
I.B.I. Base Case I. 4
I.B.2. Hypothetical Options I. 5
I.C. Summary of Analysis I. 5
I.C.I. The Costs of Reducing Lead in Gasoline I. 6
I.C.2. The Benefits of Reducing Lead in Gasoline I. 8
I.D. Limitations of the Analysis 1.15
I.E. Quantifying Effects 1.17
CHAPTER II: COSTS OF REDUCING LEAD IN GASOLINE
II.A. Input Assumptions II. 2
II.A.l. Gasoline Volume II. 2
11.A.2. Leaded-Unleaded Split II. 3
II.A.3. Misfueling II. 4
II.A.4. Octane Requirements II. 6
II.B. Reduction in Lead Emissions II. 7
II.C. Cost Estimates II. 8
II.C.I. Incremental Cost of the All Unleaded Case II. 9
II.C.2. Low-lead Case 11.11
II.C.3. Cost of Lead Reductions 11.12
II.D. Price Differentials 11.12
I I.E. Longer Term Projections 11.14
II.F. Refinery Model 11.15
II.F.I. General Description of DOE Petroleum
Refinery Yield Model 11.15
-------
Attachment I: Evolution of DOE Refinery Model and Current Status 11.28
Attachment II: Refinery Processes 11.32
References 11.36
CHAPTER III: BENEFITS FROM REDUCED VEHICLE MAINTENANCE REQUIREMENTS
III.A. Maintenance Savings in the All Unleaded Case in. 5
III.A.I. Sources of Data III. 5
III.A.2. General Garments on the Method III. 6
III.A.3. Fewer Replacements of Exhaust Systems ill. 8
III .A.4. Better Performance or Less Frequent
Spark Plug Changes III.12
III.A.5. Extended Oil Change Intervals III.16
III.A.6. Improved Fuel Economy III.23
III.B. Maintenance Savings for the Low-Lead Case III.26
III.B.I. Exhaust System Savings III.27
III.B.2. Spark Plug Savings III.29
III.B.3. Oil Change Savings III.30
III.B.4. Sum of Maintenance Savings for the
Low-Lead Case III.30
III.C. Risk of Valve Recession III.31
III.C.I. How Much Lead is Required to
Protect Valves III.36
III.C.2. Alternatives to Lead to Avoid
Valve Recession III.37
III.D. Summary III.42
References III.46
CHAPTER IV: BENEFITS OF AVOIDING EXCESS HC, NOx AND CO EMISSIONS
IV.A. Value by the Costs of "Next-Step" Regulations w. 6
IV.B. The Value of Preserving Pollution Control Equipment IV. 8
-------
IV.C. Benefits Estimated Directly from Health and Welfare
Improvements IV.10
IV.C.I. Benefits of Reducing Ozone IV.10
IV.C.I.a. Ozone Health Effects IV.16
IV.C.l.b. Ozone Agricultural Benefits IV.29
IV.C.I.e. Nonagricultural Vegetation
Benefits of Reduced Ozone IV.31
IV.C.l.d. Ozone Materials Benefits IV.32
IV.C.2. Benefits of Reducing NOx Emissions IV.33
IV.C.3. Reducing Emissions of Hydrocarbons IV.35
IV.C.4. Reducing Emissions of Carbon Monoxide IV.37
IV.D. Summary of Health and Welfare Benefits IV.41
IV.E. Summary of HC, CO, and NOx Benefits IV.42
Technical Appendix IV.44
References IV. 56
CHAPTER V: BENEFITS OF REDUCING LEAD: CHILDREN WITH HIGH BLOOD LEVELS
V.A. The Relationship between Gasoline Lead
and Blood Lead V. 3
V.B. Medical Benefits of Reducing High Blood Lead Levels V. 6
V.C. Cognitive and Behavioral Effects V.ll
V.D. Estimating Avoided Costs of Reduced Cognitive Ability V.12
V.E. Statistical Methodologies V.15
V.E.I. The NHANES II Data V.15
V.E.2. Reduction in Number of Children
Below Critical Thresholds V.20
V.E.3. Incidence Versus Prevalence V.24
V.E.4. Assessing the Accuracy of our
Forecasting Procedures V.28
V.F. Conclusion V.30
Technical Appendix V.33
References V.41
-------
CHAPTER VI: BENEFITS OF AVOIDING HEALTH EFFECTS OF BLOOD LEAD LEVELS
BELOW 30 ug/dl
VI.A. Pathophysiological Effects VI. 3
VLB. Hematological Effects of Lead VI. 9
VI.B.I. Effects on Blood Cell Volume and
Hemoglobin Content VI.10
VLB.2. The Relationship Between Blood
Lead and FEP VI.13
VLB.3. The Relationship Between FEP
Levels and Anemia VI. 15
VI.C. Fetal Effects VI.19
VI.D. Neurological Effects VI.20
VI.D.I. Cognitive and Behavioral Effects VI.23
V.D.I.a Assessment of the Relationship Between
IQ or Cognitive Function and Low Blood
'sad Levels VI.23
V.D.I.b Policy Implications of Significance Tests VI.30
VI.D.2. Estimating Avoided IQ Loss Associated with
Reduced Blood Lead Levels VI.32
VI.D.3. Threshold for Effects of Blood Lead on IQ
and the Size of the Affected Population VI.34
VLB. Estimating the Reduction in the Number of Children at Risk VI.38
VI.F.I. Distributional Aspects of Lead Exposure VI.40
VI.F. Conclusion VI.41
References VI.44
-------
TABLES, EXHIBITS AND FIGURES
CHAPTER I
TABLE 1-1
TABLE 1-2
FIGURE 1-1
CHAPTER II
TABLE II-l
TABLE I1-2
TABLE I1-3
TABLE II-4
Comparison of Benefits and Costs of
Lead Regulation Options in 1988
Environmental Effects in 1988
of Reduced Gasoline Lead Use
Lead Used in Gasoline Production and
Average NHANES II Blood Lead Levels
Projected Gasoline Demands
Leaded Gasoline Demand
Due to Misfueling
Metric Tons of Lead Removed
Cost of Reducing or Banning
Leaded Gasoline
I. 9
1.11
1.13
II. 3
II. 4
II. 8
11.11
EXHIBIT 1 Flow Diagram of Topping Refinery
Processing Low Sulfur Crude Oil
EXHIBIT 2 Flow Diagram of Hydroskimming Refinery
Processing Low Sulfur Crude Oil
EXHIBIT 3 Flow Diagram of Fuels Refinery
Processing High Sulfur Crude Oil
EXHIBIT 4 Flow Diagram of High Conversion Refinery
Processing High Sulfur Crude Oil
EXHIBIT 5 Functional Characterization of
Petroleum Refinery Processes
EXHIBIT 6 Yields and Operating Costs Coefficients
Crude Distillation Unit
EXHIBIT 7 Yields and Operating Costs Coefficients
Catalytic Reforming Unit
EXHIBIT 8 Estimated U.S. Refinery Processing
Unit Capabilities for 1988
11.18
11.19
11.20
11.21
11.22
11.25
11.26
11.27
-------
TABLES, EXHIBITS, AND FIGURES (Continued)
CHAPTER III
TABLE III-l Summary of Studies on Maintenance Differences
Between Leaded and Unleaded Vehicles
TABLE III-2 Summary of Findings: Valve Recession
at Varied Lead Concentrations
TABLE II1-3 Number of Engines at Risk of Severe
Valve Recession Without Leaded Gasoline
TABLE III-4 Summary of Maintenance Savings
III.6
111.35
111.43
III.45
CHAPTER TV
TABLE IV-1
TABLE IV-2
TABLE IV-3
TABLE IV-4
TABLE IV-5
TABLE IV-6
TABLE IV-7
Increase in Emissions Due to Misfueling
1982 Misfueling Rates by Age
of Vehicle and by I/M Status
Discounted Future Emissions Avoided
by Eliminating Misfueling in 1988
Benefits Valued by "Next-Step" EPA
Regulations
Benefits of Reducing Asthmatic Attacks
1988 Benefits of Reducing HC, NO*
and CO Emissions by Direct Estimation
Summary of Benefits in 1988 of Reducing
HC, CO, and NOX Emissions
TECHNICAL APPENDIX FOR CHAPTER IV
APPENDIX TABLE IV-1 Light-Duty Vehicle Projections
APPENDIX TABLE IV-2 Light-Duty Truck Projections
APPENDIX TABLE IV-3 Summary of Fleet Model Parameters
APPENDIX TABLE IV-4 General Fleet Assumptions
APPENDIX TABLE IV-5 Misfueling Rates By Age
IV. 2
W. 3
W. 5
IV. 7
W.27
IV.41
IV.43
PV.45
IV.46
W.47
PV.49
IV.52
-------
TABLES, EXHIBITS, AND FIGURES (Continued)
CHAPTER V
TABLE V-l
TABLE V-2
TABLE V-3
TABLE V-4
TABLE V-5
TABLE V-6
TABLE V-7
Reduction in Number of Children
Above 30 ug/dl in 1988
Medical Cost Savings in 1988
Benefits in 1988 of Reduced Cognitive Losses
Estimated Lead Used for Gasoline in 1988
Changes in Mean Blood Lead in 1988 for
Black and White Children Aged 5 and Under
1988 Population Projections
Monetized Benefits of Reduced Numbers of
Children Above 30 ug/dl Blood Lead Level in 1988
V. 3
V.10
V.14
V.23
V.23
V.24
V.31
FIGURE V-l Children's Blood Lead Levels Vary Directly
With Levels of Lead in Gasoline
FIGURE V-2 Lead Used in Gasoline Production and
Average NHANES II Blood Lead Levels
FIGURE V-3 Average NHANES II Blood Lead Levels
vs. Lead Used in Gasoline Production
FIGURE V-4 Average Blood Lead Levels for Black Children
in Chicago and Gasoline Lead in Chicago
V. 4
V.17
V.18
V.19
CHAPTER VI
TABLE VI-1
TABLE VI-2
Blood Lead Levels of Persons Aged Six Months
- 74 Years in United States 1976-1980
Computation of Joint P-Value from
Epidemiological Studies of Cognitive
Effects from Low Level Lead Exposures
in Children
TABLE VI-3 Decrease in Number of Children in 1988
Above Thresholds for Cognitive Effects
VI. 2
VI.2 7
VI.3 6
-------
TABLES, EXHIBITS, AND FIGURES (Continued)
TABLE VI-4 Possible Change in Person-IQ Points in 1988
as a Function of Threshold Levels for
Children 6 Months to 7 Years
TABLE VI-5 Decreased Number of Children
(13 years old and under) above
Apparent Threshold Levels in 1988
TABLE VI-6 Estimated Distribution of Blood
Lead Levels in 1988
TABLE VI-7 Summary of the 1988 Health Benefits of
Reducing Low Level Lead Exposure
VI.3 7
VI. 39
VI.40
VI.42
FIGURE VI-1 Percent of Children with MCV below 74 f1
FIGURE VI-2 Percent of Children with Anemia
FIGURE VI-3 Mean IQ Difference Between High Lead Groups
and Controls, Adjusted for Socioeconomic
Factors
VI. 12
VI.17
VI.29
-------
EXECUTIVE SUMMARY
COSTS AND BENEFITS OF REDUCING LEAD IN GASOLINE
COSTS
Lead is a relatively inexpensive way to boost gasoline octane, but
eliminating or severely limiting lead would increase the manufacturing cost of
gasoline by less than 1%. Eliminating lead altogether may result in excessive
valve wear in some trucks and older cars; a low-lead fuel of 0.10 grams/gallon
would prevent this problem.
BENEFITS
Maintenance Savings; Lead forms corrosive conpounds that increase automobile
maintenance costs. Cars that use leaded gasoline need more frequent tune-ups,
exhaust system replacements, and oil changes.
Misfueling; Recent EPA surveys indicate that over 12% of all cars equipped with
catalytic converters to control auto emissions are currently being "misfueled"
with leaded gasoline because consumers want either to save money or to obtain
higher octane. Misfueling poisons catalysts and substantially increases the
conventional auto pollutants: hydrocarbons, carbon monoxide, and nitrogen
oxides. Given current misfueling rates, misfueled vehicles will account for
one-third of leaded gasoline demand in 1988, significantly increasing our
estimates of future lead and conventional pollutant emissions. The impact of
these emissions on public health and welfare is substantial.
Health; Our analysis and other major studies both in the U.S. and abroad
indicated that the amount of lead in blood is directly related to the amount
of lead in gasoline. Lead has long been known to cause pathophysiological
changes, including the inhibition of major enzymatic processes, adverse effects
on the central nervous system, and decreases in cognitive ability. Children
are especially vulnerable to lead, and black children are more severely affected
than others. Children with elevated blood lead levels require medical monitoring
and/or treatment.
Adverse effects of lead in the blood are now found at levels that were
previously thought safe, and additional effects are suspected. The Centers
for Disease Control is currently investigating lowering its definition of lead
toxicity.
SUMMARY
Our examination of the costs and benefits of two options for further
restricting the use of lead in gasoline, summarized in Table 1, on the next
page, indicated that the benefits exceed the costs. Although we were able to
place dollar values on reduced medical costs and cognitive damage for children
with high lead levels we did not monetize other factors affecting this group,
such as behavioral and other problems, nor the pain of medical treatment and
parents' lost work time. No monetary values at all were estimated for children
with lower lead levels, although they suffer some negative effects. These non-
monetized benefits are represented by "H" in the table. Table 2 provides a
sunmary of the environmental effects of reducing lead in gasoline.
-------
SUMMARY TABLE 1
Comparison of Benefits and Costs of
Lead Reduction Options in 1988
(millions of 1983 dollars)
Low-Lead Option* All Unleaded**
COSTS
Manufacturing Costs
Non-monetized Valve Damage
to Engines that Need Lead
TOTAL COSTS
$503
$503
BENEFITS
Maintenance Benefits
Environmental and Health Benefits
Conventional pollutants
Reduced damage by eliminating misfueling
Non-monetized health benefitst
Lead
Reduced medical care costs
Reduced cognitive damage
Non-monetized health benefits''
TOTAL BENEFITS
NET BENEFITS
$660
$404
$41
$184
$691
D
$691+D
$755
$404
$786+H1+H2
$43
$193
H3
$!1395+Hi+H3
$704+H1+H3-D
* This option would make a low lead gasoline (0.10 grams of lead per gallon) avail-
able only for those few vehicles that require some lead. It assumes no misfueling.
** All lead in gasoline would be banned by 1988.
1" These include chronic health effects of ozone and CO, and any effects of reduced
sulfate particulates.
1"!" since medical costs and cognitive damage were only monetized for children with
high blood lead (>30 ug/dl), H2 and H3 represent other benefits for this group
(pain, lost work time to parents, etc.) as well as all the benefits (medical,
cognitive, behavior, etc.) for the lower lead group (<30 ug/dl). H2 and H3 differ
because the numbers of children at risk under the two options differ.
-------
SUMMARY TABLE 2
ENVIRONMENTAL EFFECTS IN 1988 OF REDUCED
GASOLINE LEAD USE
REDUCTIONS IN EMISSIONS
(thousands of metric tons)
Lead
HC
CO
NOX
Ozone (resulting from reduced
HC and NOx emissions)
REDUCTIONS IN THE NUMBER OF CHILDREN
AT RISK OF ADVERSE HEALTH EFFECTS
Reduction in number of children
at risk of:
- Inhibition of enzyme
activity (PY-5-N and AIA-D)
Reduction in number of children
at risk of:
- Changes in EEC patterns
- Impairment of heme synthesis
- Elevated levels of ALA and
possible interference with
neurotransmission processes
- Impairment of vitamin D activity
- Possible adverse cognitive
effects
Reduction in number of children
at risk of impaired globin synthesis
Reduction in number of children
at risk of:
- Potentially requiring
active medical care
- Probable adverse cognitive
effects
LOW-LEAD
33.4
314
2,202
130
1.5% reduction
4,257,000
1,475,000
476,000
43,000
ALL UNLEADED
35.6
314
2,202
130
1.5% reduction
4,486,000
1,553,000
500,000
45,000
-------
CHAPTER I
INTRODUCTION, FINDINGS, AND QUALIFICATIONS
I.A. Background
Since 1973, the U.S. Environmental Protection Agency (EPA)
has regulated the use of'lead as an additive to gasoline. Section
211 of the Clean Air Act gives the EPA Administrator authority
to control or prohibit any fuel or fuel additive that:
0 causes, or contributes to, air pollution which may
reasonably be anticipated to endanger the public health
or welfare, or
0 will impair to a significant degree the performance of any
emission control device or system which is in general use...
To avoid the adverse effects of lead in the environment and
to protect emission control equipment which is rendered ineffec-
tive or "poisoned" by lead additives, EPA required that cars,
beginning with model year 1975, meet tighter emissions limits.
To do this automobile manufacturers installed catalytic converters
requiring unleaded gasoline. In several stages during the period
1976-1982, EPA mandated that the lead content of leaded gasoline
be reduced from over 2.0 grams per gallon to 1.1 g/gal.
During this period, studies on blood lead levels showed
that reducing the lead content of gasoline would also reduce
blood lead levels in all major population groups in the United
States. It was anticipated that the combination of these two
actions would restrict and eventually eliminate the exposure of
the general population, especially young children, to airborne
lead from mobile sources, as well as reduce health and welfare
damage from conventional pollutants.
-------
1.2
While EPA's rules have virtually eliminated leaded premium
gasoline, consumers of regular gasoline must choose between
relatively inexpensive gasoline containing lead additives or
more expensive unleaded gasoline. In addition to savings, some
consumers want the slightly higher octane of leaded regular.
(However, few recognize lead's corrosive effects on their engines
or the increased maintenance cost of using leaded gasoline.)
Recently, several EPA and private studies have found wide-
spread "misfueling" (i.e., the use of leaded gasoline in vehicles
designed for unleaded gasoline). The studies showed that con-
stant misfueling rates of over 12% have slowed the decline in
lead emissions significantly, and challenged the assumption that
leaded gasoline would soon be eliminated because of lack of
demand. These findings have occurred at the same time that the
public health community's long-standing concern about lead has
produced a substantial literature about the adverse effects of
lower lead exposures. Specifically, recent studies have strength-
ened the identification of gasoline lead as a major source of
blood lead and new information on the effects of lead on physio-
logical functions has become public.
Some of this information began to surface during the hearings
and subsequent comment period related to EPA's proposed lead phase-
down rule making in 1982. At that time, the Agency had proposed
several regulatory alternatives. As a result of the information
gained from the public response during the proceedings, EPA
tightened the restrictions on the amount of lead permitted in
-------
1.3
leaded gasoline. The restriction also set a uniform limit for
both small and large refiners.
The growing problem of the misuse of leaded fuel in cars
with catalytic converters, the increasing recognition of serious
health effects from even low lead levels, and the fact that gaso-
line has been identified as the major source of environmental
exposure to lead all indicated that a simple continuation of
current policies needed reexamination. EPA presently has two
review processes underway for assessing the effects of lead.
The first is the Agency's formal Criteria Document process,
which is managed by the Office of Research and Development.
The Lead Criteria Document will evaluate all of the environmental
effects of lead. A Draft Lead Criteria Document was circulated
for comment in October 1983; a final document is expected by
August 1984. Nothing in this paper is intended to prejudge or
supercede the outcome of that process.
Concurrently, and on a somewhat faster timetable, EPA's
Offices of Policy, Planning and Evaluation, and Air and Radiation
have been reviewing data from the 1982 phase-down effort to
evaluate the costs and benefits of additional restrictions on
the amount of lead in leaded gasoline. This paper is primarily
an analysis of the monetized costs and health and welfare benefits
of reducing the lead content of gasoline.
-------
1.4
I.E. Approach
For this paper, we have contrasted the costs and benefits
of two hypothetical options against a base case which continues
existing regulations and compliance practices. Both the options
and the base case are projections for 1988. The first option
is a low-lead fuel (0.10 grams of lead per gallon) for the few
classes of vehicles, such as trucks and older cars, that may
require the valve lubrication that lead provides. Contrary to
the current situation, such a low lead fuel would cost more to
manufacture than regular gasoline. We assume this cost inversion,
coupled with availability restrictions on this fuel, would elimi-
nate misfueling as a practical problem. The second option is
the banning of all leaded gasoline.
I.B.I. Base Case
At present, EPA regulations restrict the use of lead in
gasoline in two ways. First, beginning with the 1975 model year,
almost all light-duty vehicles have been equipped with catalytic
converters and require unleaded gasoline. By 1981, virtually all
new light-duty vehicles should have been using unleaded gasoline.
Second, EPA limits the lead content of leaded gasoline to 1.1
grams per gallon. Because lead is a relatively inexpensive octane
enhancer, this is about half of what refiners would use if not
statutorily constrained.
The lead standard must be met on a quarterly average, how-
ever, not for each gallon produced. In addition, refiners may
average their own production or sell off-sets to each other.
-------
1.5
That is, two refiners may agree that one of them will produce
gasoline with 1.0 grams of lead per gallon and the other will
use 1.2 grams of lead (and, presumably, pay the first one an
agreed amount). This allows the refinery industry as a whole to
optimize its use of octane manufacturing capacity and to minimize
the cost of meeting the restrictions on lead use.
I.B.2. Hypothetical Options
To address the problems of misfueling and airborne lead
pollution, the first alternative we considered was an outright
ban on the use of lead in gasoline. Such a regulation would
meet both the public health and misfueling concerns, and we
examined this option carefully. However, some vehicles could
experience severe valve damage if no leaded fuel were available.
We therefore added a second "low-lead" option, which assumed
that marketing restrictions would be designed so as to eliminate
misfueling. The amount of lead in leaded gasoline would be
restricted to 0.10 grams per gallon, which is sufficient to
protect valves from undue wear, but which minimizes environmental
contamination.
I.C. Summary of Analysis
Our analysis evaluates and compares the costs and benefits
in 1988 of reducing or eliminating lead in gasoline. To
calculate the costs of restricting lead as an octane-enhancer,
we used a linear programming model of the refinery industry.
-------
1.6
In the benefits area, we calculated vehicle maintenance
savings that would be realized by eliminating the corrosion and
engine fouling problems associated with lead in gasoline. We
also monetized the benefits of reducing the emissions of conven-
tional pollutants that result from misfueled vehicles, and
analyzed the number of children at risk of various health
effects from lead exposure.
We have valued the benefits associated with reducing the
number of children suffering from "undue lead absorption,"
currently defined by the Centers for Disease Control as blood
lead levels above 30 micrograms per deciliter (ug/dl) and free
erythrocyte protoporphyrin (FEP) levels over 50 ug/dl. For
children with blood lead levels below 30 ug/dl, we calculated
the change in the number of children with blood lead levels
above the lowest observed effects level for pathophysiological
changes but we did not ascribe any dollar values to reducing
their lead exposures. We also estimated the change in the number
of children who might suffer small decreases in cognitive ability,
but again we attached no monetary value to this.
Chapters II (on costs) and V (on the health effects of blood
lead levels over 30 ug/dl) contain detailed sections describing
the methods we used in our analysis.
I.C.I. The Costs of Reducing Lead in Gasoline
Lead is added to gasoline because it is the least expensive
way for petroleum refiners to boost the octane of fuel. Reducing
-------
1.7
or eliminating the lead content of gasoline will require extra
energy use (and potentially more equipment) and, consequently,
greater resource costs. We estimated the increased costs of raw
materials and refining would be less than 1%. As a result, many
consumers would pay slightly more for gasoline.
Chapter II contains a description of consumer demand for
gasoline, the leaded/unleaded split, and current needs for octane.
Based upon our models and projections by the Energy Information
Administration and Data Resources, Inc. (DRI), we have projected
gasoline demand and the leaded/unleaded split under existing
policies and misfueling rates, and under the two hypothetical
options: low-lead and all unleaded. The refinery cost figure
is an estimate of the extra manufacturing costs incurred by
refineries if they must use other octane-producing processes to
meet U.S. demand for gasoline. These costs were derived from
the same linear program of the refining industry that was used
in EPA's economic analysis of the 1982 lead phase-down regulations.
We projected that, meeting current consumer requirements for
octane, the 1988 cost to refiners of reducing lead in the low-
lead option would be $503 million and the cost of the all unleaded
option would be $691 million. Because we could not predict how
changes in production costs might affect the marketing strategies
of retailers under our two options, we did not attempt to estimate
the change in gasoline prices to consumers.
-------
1.8
I.C.2. The Benefits of Reducing Lead in Gasoline
Chapters III-VI describe the monetized benefits of reducing
the amount of lead in gasoline and some unmonetized health
benefits of reducing overall exposure to lead.
Chapter III (Maintenance Savings) describes the vehicle
operation and maintenance savings that would result, from restrict-
ing lead in gasoline. Lead compounds and their associated
scavengers foul and corrode the engines and exhaust systems of
all vehicles using leaded gasoline, whether designed for it or
not.* Operation and maintenance savings come from three primary
areas: less frequent tune-ups, less frequent exhaust system
replacements, and less frequent oil changes. We estimated that
vehicle owners who switch from leaded to unleaded gasoline could
save 3-4 cents per gallon of gasoline. The total benefits were
computed by multiplying the savings per gallon times the total
number of gallons consumed. The estimates of maintenance savings
we have included in Table 1-1 (on the next page) were at the low
end of our range. We also discussed the possibility of valve
damage to leaded vehicles, which could occur in our all unleaded
option, but not in our low-lead option. We were unable to estimate
a monetary value for this because we did not have information on
how many vehicles are driven under the conditions where it could
occur.
* Scavengers are necessary to remove lead from the engine after
combustion. Without these scavengers, engine performance
would rapidly deteriorate to complete inoperability.
-------
1.9
TABLE I-l
Comparison of Benefits and Costs of
Lead Reduction Options in 1988
(millions of 1983 dollars)
COSTS
Manufacturing Costs
Non-monetized Valve Damage
to Engines that Need Lead
TOTAL COSTS
Low-Lead Option* All Unleaded**
$503
$503
$691
D
$691+D
BENEFITS
Maintenance Benefits
Environmental and Health Benefits
Conventional pollutants
Reduced damage by eliminating misfueling
Non-monetized health benefitst
Lead
Reduced medical care costs
Reduced cognitive damage
Non-monetized health benefits
TOTAL BENEFITS
$660
$404
Hl
$41
$184
H
$755
$404
$43
$193
H
NET BENEFITS
$786+Hi+H2
* This option would make a low lead gasoline (0.10 grams of lead per gallon) avail-
able only for those few vehicles that require some lead. It assumes no misfueling.
** All lead in gasoline would be banned by 1988.
"*" These include chronic health effects of ozone and CO, and any effects of reduced
sulfate particulates.
1"T Since medical costs and cognitive damage were only monetized for children with
high blood lead (>30 ug/dl), H2 and H3 represent other benefits for this group
(pain, lost work time to parents, etc.) as well as all the benefits (medical,
cognitive, behavior, etc.) for the lower lead group (<30 ug/dl). H2 and H3 differ
because the numbers of children at risk under the two options differ.
-------
1.10
We estimated that total savings from reduced maintenance and
operation expenses in 1988 would be $660 million for the low-lead
option and $755 million for the all unleaded option.
Chapter IV (Benefits of Avoiding Excess HC, CO, and NOX
Emissions) examines misfueling practices and their consequences
for emissions of the conventional auto pollutants: hydrocarbons,
carbon monoxide, and nitrogen oxides. As we have noted, using
leaded gasoline in vehicles designed to run on unleaded gasoline
poisons their catalytic converters, which causes a substantial
increase in HC, CO, and NOX. While all vehicles equipped with
catalytic converters are required to use unleaded gasoline, over
12% of all vehicles equipped with catalysts are currently being
misfueled with leaded gasoline.
We estimated the excess emissions in grams per mile and
computed the increases in total emissions due to poisoned
catalysts. Because HC and NOX combine to form ozone, we also
estimated the increase in ozone which formed as a result of more
conventional pollution. Our estimates of the size of these
changes appear in Table 1-2. We used existing literature and
data on the negative health and welfare effects of these
conventional pollutants to value these changes in emissions.
We used three methods to value the benefits of avoiding these
excess emissions in 1988: 1) an estimate valuing the avoided
emissions at the average cost per ton of the most cost effective
alternative for controlling these pollutants, 2) an estimate
valuing the avoided emissions at the average cost per ton of the
-------
1.11
TABLE 1-2
ENVIRONMENTAL EFFECTS IN 1988 OF REDUCED
GASOLINE LEAD USE
REDUCTIONS IN EMISSIONS
(thousands of metric tons)
Lead
HC
CO
NOX
Ozone (resulting from reduced
HC and NOX emissions)
REDUCTIONS IN THE NUMBER OF CHILDREN
AT RISK OF ADVERSE HEALTH EFFECTS
Reduction in number of children
at risk of:
- Inhibition of enzyme
activity (PY-5-N and ALA-D)
Reduction in number of children
at risk of:
- Changes in EEC patterns
- Impairment of heme synthesis
- Elevated levels of ALA and
possible interference with
neurotransmission processes
- Impairment of vitamin D activity
- Possible adverse cognitive
effects
Reduction in number of children
at risk of impaired globin synthesis
Reduction in number of children
at risk of:
- Potentially requiring
active medical care
- Probable adverse cognitive
effects
LOW-LEAD
33.4
314
2,202
130
1.5% reduction
4,257,000
1,475,000
476,000
43,000
ALL UNLEADED
35.6
314
2,202
130
1.5% reduction
4,486,000
1,553,000
500,000
45,000
-------
1.12
program requiring catalytic converters on cars, and 3) an
estimate using econometric damage functions to value the avoided
emissions. We used the average of the last two alternative
methods, $404 million, as a point estimate in Table 1-1.
Chapter V discusses the health benefits of reducing lead in
gasoline by valuing the damage resulting from blood lead levels
over 30 ug/dl (which is, in combination with elevated FEP levels,
currently the Centers for Disease Control's definition of lead
toxicity).
Lead emissions from cars increase the blood lead levels of
children. We analyzed blood lead data from the Second National
Health and Nutrition Examination Survey (NHANES II) and found a
strong, statistically significant relationship between blood lead
levels and the amount of lead in gasoline, after controlling for
age, race, sex, income, degree of urbanization, education, region
of the country, and alcohol consumption. The strong relationship
is shown graphically in Figure 1-1. Inner city black children
have the highest rates of elevated blood lead levels, but a
substantial number of white children also are affected.
Lead is known to damage the kidney, the liver, the
reproductive system, blood creation, basic cellular processes,
and brain functions. Using the projected lead reductions from
Chapter II, we estimated how many fewer children would be likely
to be at risk of undue lead exposure in 1988. From these we
estimated the benefits of avoiding the pathophysiological and
cognitive and behavioral effects of elevated blood levels.
-------
110
8 100
90-
w
o
W
I
i-3
a
rt
O
3
(0
80-
70-
w
50
H
O
0 60 H
o
o
o
50-
LEAD USED IN GASOLINE PRODUCTION AND
AVERAGE NHANES II BLOOD LEAD LEVELS
(FEB. 1976 - FEB, 1980)
LEAD USED IN
GASOLINE
AVERAGE
BLOOD
LEAD LEVELS
1976
1977
1978
YEAR
1979
1980
•16
-15
-13
•12
-11
-10
I
w
o
o
o
§
3
W
f
to
3
H-
O
CO
(D
O
I-1-
I-1
H-
ft
ro
H
(0
V
-------
1.14
When children's blood lead levels are over 30 ug/dl, they
require follow-up and/or medical treatment. The estimate in
Table 1-1 was based on a regression that projected the number of
additional children who would require medical treatment for
elevated blood lead levels as a result of gasoline lead use.
It did not include children who would need treatment for lead
poisoning because of lead-based paint or other sources of
exposure. We included the costs of medical treatment, even for
the children whom public health officials do not find and treat,
because we assumed that the social cost of elevated lead levels
for an untreated child was at least as great as what we spend on
treatment for those who are identified.
Some of these children have blood lead levels high enough
to reduce cognitive performance, including the loss of several
IQ points. Researchers have found that these cognitive deficits
remain three years later, even after medical attention. Table
1-1 also includes the costs of compensatory education to overcome
the additional learning difficulties that children with high
lead levels incur. As in the case of medical costs, we included
costs even for those children who do not receive compensatory
education. Again, we assumed that the costs to society of a
learning disability were at least as great as the cost of a
program to partially compensate for the damage.
We estimated, for the all unleaded case, that the benefits
of avoiding medical and associated costs for children with
blood lead levels over 30 ug/dl in 1988 were $43 million, and
-------
1.15
that the value of avoiding the cognitive damage likely to occur
at those levels was $193 million. We estimated that the benefits
in the low-lead case are $41 million for medical savings and
$184 million for avoided cognitive damage.
Chapter VI discusses the health effects of blood lead levels
below 30 ug/dl. As measurement tools have improved, research
has detected pathophysiological effects at blood lead levels
that were previously thought to be safe, and additional effects
are suspected. These results warrant concern about even small
changes in the total body lead burden of children, especially
those children who are subject to sources of lead exposure in
addition to lead from gasoline.
While the full clinical significance of the effects of blood
lead levels below those requiring medical management under current
practice is not yet clear, the Centers for Disease Control is now
considering lowering its current (30 ug/dl of blood lead and FEP
levels of 50 ug/dl) criteria for lead toxicity.
Among the recent data on these pathophysiological changes
are inhibition of the enzymes Pyrimidine-5'-nucleotidase (PY-5-N)
and aminolevulinic acid dehydrase (ALA-D), which begins to be
detectable at about 10 ug/dl of blood lead; changes in EEC
patterns, detectable at about 15 ug/dl; elevated ZPP or FEP in
red blood cells at about 15 ug/dl; inhibition of globin synthesis
at about 20 ug/dl; increased risks of abnormally small red blood
cells at 20-25 ug/dl; and other disruptions of aminolevulinic
acid (ALA) and vitamin D homeostasis at about 15 ug/dl. In
addition, our analysis of the combined evidence from all the
-------
1.16
relevant studies indicated that mild cognitive effects also
occurred at low lead exposure levels.
Our estimates of the reduced number of children at risk of
health effects in 1988 are presented in Table 1-2. We have not
valued these changes monetarily, but crude valuation procedures
suggest the benefits are likely to be large.
I.D. Limitations of the Analysis
This paper is a cost-benefit analysis of reducing the lead
content of gasoline. To do this, we have proposed two hypothe-
tical options: a low-lead and an all unleaded scenario. Our
analysis measured the effects in one year, 1988. With such a
far-reaching issue, the limitations of our findings should be
clarified.
We have forecast circumstances and events that will occur
four years in the future, and the future is, at best, uncertain.
One problem is shifts in underlying behavior such as a change in
consumer preferences back to large cars or changes in external
events (e.g., another big war in the Mideast). In addition,
because we are extrapolating from our perceptions and experience
to date1, any misapprehension of what is will tend to be magnified
as we project several years ahead. (An example of this may be
the misfueling problem.)
Although we believe that our model of the refinery industry
is as accurate as possible, we can not predict marketing
behavior. We believe we have estimated real resource costs
-------
1.17
fairly accurately, but we can not predict with confidence what
would happen to consumer prices.
In the benefits area, we are still learning about the health
effects of lead and other criteria pollutants. The body of know-
ledge is neither well-defined nor unequivocal. While the trend
in new findings seems to be uncovering more effects at lower
levels, the clinical significance of these findings is not always
clear. Also, the distributions of effects that we are predicting,
especially at 30 ug/dl of blood lead, are near the tails of the
distributions, and therefore, more susceptible to changes and
uncertainties. However, we have no indications that our
estimation procedure is biased, so the effects are as likely to
be larger as smaller. In addition, it is difficult to measure
IQ loss, and even more difficult to put a dollar value on lost
IQ points.
While we have used accepted state-of-the-art methods for
valuing health and environmental effects, there are uncertainties
about the health and welfare effects of hydrocarbons, nitrogen
oxides, and carbon monoxide; and about the transformation of
hydrocarbons and nitrogen oxides into ozone. Finally, there are
some uncertainties inherent in the monetary valuation of these
effects.
I.E. Quantifying Effects
We have, in the course of this analysis, explored many
alternative assumptions and methods for valuing effects. Through-
out, our overall results have proven to be very robust to changes
in details; that is, small changes did not alter results.
-------
1.18
The effects for which we have presented monetary values in
Table 1-1 have a solid basis. Where the data could not support
a point estimate or even a range, we did not provide a monetary
value. All significant effects, however, whether monetized or
not, are included in Table 1-1 to allow the reader to gain a
full perception of the problem.
The clear conclusion from the data summarized in Table 1-1
is that the benefits of the low-lead option substantially exceed
the costs. For the all unleaded case, the issue is less clear
because of the unresolved nature of the cost of valve damage.
However, as engines which need leaded gasoline retire from the
fleet, the issue of valve damage becomes less important. Thus,
in the long run, the option of eliminating lead in gasoline
appears very attractive.
-------
CHAPTER II
COSTS OF REDUCING LEAD IN GASOLINE
Petroleum refiners add lead to gasoline as the least
expensive way to boost octane. There are alternative additives
that also help boost octane, but they generally are more
expensive and, like lead, can also be toxic.
The most attractive alternative refiners have for raising
the octane of unleaded gasoline is additional processing of the
gasoline in either catalytic reformers or in isomerization units.
Increasing the use of reformers and isomerizers requires more
energy consumption, and thus raises the cost of manufacturing
gasoline. (This may also increase the density of gasoline which
raises slightly the energy content per gallon of gasoline.)
If refiners need to produce more unleaded gasoline but are
limited by isomerization or reforming capacity, they can construct
more capacity, incurring a capital charge. Alternatively,
refiners can purchase either a better grade of crude oil or add
other octane boosters, incurring higher operating costs. The sum
of all these costs, along with miscellaneous energy costs, etc.,
is the additional cost of making gasoline with less or no lead.
In this chapter we discuss some of the basic input assump-
tions we used to estimate the refinery costs of producing
gasoline under our two options.
Estimates of the reduction in lead emissions under our
two policy options are presented. We then show the costs
derived from applying our assumptions to the Department of
Energy (DOE) refinery model. A description of the DOE model
-------
II.2
and a brief explanation of refinery processes are presented in
the last section of this chapter.
II.A. Input Assumptions
The cost to refiners of manufacturing unleaded or low lead
gasoline depends principally on three factors:
o the total gasoline volume produced,
o the portions of gasoline production that are leaded and
unleaded, and
o the level of octane specified for the gasoline pool.
II.A.I. Gasoline Volume
Table II-l presents the gasoline demand assumptions that we
used to estimate the costs of manufacturing unleaded gasoline for
our baseline and two policy options. Gasoline demand estimates
are obviously subject to uncertainty. Demand in 1983 was approxi-
mately 6.6 million barrels per day. We assumed that demand in
1988 would fall to 6.5 million barrels per day, because newer
cars are more fuel efficient (i.e., get more miles to the gallon)
than the older vehicles they replace. We believe fuel efficiency
effects will slightly outweigh the effects of the growing number
of vehicles, even though vehicle miles travelled is also expected
to increase. (For comparison with our estimate, a recent Data
Resources Inc. [DRI, 1983] model forecast for 1988 is for 3%
fewer gallons of gasoline than we assumed; last year's Energy
Information Administration [EIA, 1983] estimate was much lower
but has been revised upward to about our level.)
-------
Leaded
3.13
2.80
2.45
2.11
1.80
Unleaded
3.41
3.78
4.09
4.39
4.66
Total
6.54
6.58
6.54
6.50
6.46
Leaded
N.A.
2.32
1.88
1.48
1.07
Unleaded
N.A.
4.26
4.66
5.04
5.39
Total
N.A.
6.58
6.54
6.50
6.46
II.3
TABLE II-l
Projected Gasoline Demands
(millions of barrels per day)
With Misfueling Without Misfueling
Year
1982 (actual)
1984
1986
1988
1990
*To convert to billions of gallons per year, multiply by
42 gallons/barrel times 365 days/year.
II.A.2. Leaded-Unleaded Split
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% (plus or minus 0.6% at
a 95% confidence level) of the gasoline market in 1988. Our
vehicle fleet model (described in Chapter III), using historic
scrappage rates from DRI, predicted essentially the same unleaded
share. For our cost analysis, therefore, we assumed a 67.5%
share.
The first set of projections, "with misfueling," is our
reference or baseline of current regulations and current
misfueling rates. The projections for leaded gasoline "without
-------
II.4
misfueling" are used in our first policy option, whereby the
lead content of gasoline is reduced to the level necessary to
protect older vehicles' valves, but misfueling is eliminated.
We believe this low-lead option will eliminate misfueling
because leaded gasoline with only 0.1 grams per gallon will cost
more to manufacture than unleaded regular gasoline, so it will
no longer be the lowest cost product. Also, restrictions on
availability will reduce the incentive to misfuel, for instance,
if this grade were limited to full service stations.
II.A.3. Misfueling
While the population of vehicles that legally may use
leaded gasoline is shrinking, misfueled vehicles are slowing
the decline in demand for leaded gasoline. By 1990, misfueling
will account for over one-third of leaded gasoline demand.
Table I1-2 shows these percentages. (The model that forecast
them is discussed in Chapter III.)
TABLE II-2
Leaded Gasoline Demand Due to Misfueling
Billions Percent of Percent of
Year of Gallons Total Gasoline Leaded Gasoline
1984 7.36 7.3% 17%
1986 8.72 8.7% 23%
1988 9.97 10.0% 31%
1990 11.19 11.3% 41%
-------
II.5
EPA surveys indicated that over 12% of vehicles designed to
use unleaded gasoline in fact use leaded (EPA, 1983). Because
surveyed motorists may refuse to allow their cars to be inspected/
however, these survey results probably greatly underestimate the
misfueling rate. Misfueling has significantly increased the cur-
rent demand for leaded gasoline. Only 52% of gasoline demand in
1982 was unleaded, as opposed to Dupont's 1979 projections of 62%.
The two most common reasons given by motorists for misfueling are
the price differentials between leaded and unleaded fuel, and
driver dissatisfaction with performance resulting from the lower
octane generally found in regular unleaded gasoline.
A problem of octane-related performance occurs because
regular leaded gasoline generally has 89 octane* while regular
unleaded has 87 octane. Some cars designed to use leaded gaso-
line do not function as well with the lower octane in unleaded
regular. In our analysis we addressed this issue by projecting
actual octane need and requiring our refinery model to meet that
demand. This increased the cost of manufacturing unleaded gaso-
line. We included the octane-related performance issue as a
cost of manufacturing, not as performance degradation. The
specific assumptions we made about the distribution of octane
requirements for misfuelers and leaded gasoline vehicles are
discussed below.
* In this paper, we define octane to be the average of research
and motor octane, commonly expressed as (R + M)/2.
-------
II.6
II.A.4. Octane Requirements
If the no lead or low-lead options eliminate misfueling,
we must then identify what octane fuel the former misfuelers
will choose. If all misfueling resulted from the current seven
cent per gallon average price differential, misfuelers would
revert to the lowest priced alternative. In our two hypothetical
options, regular unleaded would be the least expensive. If about
half of misfueling were due to price and half to performance
considerations, half of the misfuelers would drop to 87 octane
regular unleaded* but the other half would still require 89
octane.** If all misfueling were due to octane needs, all
previously misfueled cars would still require 89 octane after
reverting to unleaded.
We believe the intermediate case (half misfueling due to
price and half to octane) is the most reasonable assumption,
but we have calculated the range of costs for the various
assumptions.
* Currently, the average octane of "87 octane" unleaded is
really above 87 octane and the average octane of leaded
gasoline and premium unleaded is also above the number
specified. We have used the real average octanes, not
their numerical specifications, for the three gasoline
grades in our model, but we refer to them as "87 octane,"
etc., for convenience.
** An 89 octane unleaded grade need not be specifically
manufactured for retail outlets. Most gasoline stations
now have three pumps. If they ceased selling leaded,
they could attach the third pump to a blend of regular
unleaded (87 octane) and premium unleaded (91 octane),
thereby producing a mid-grade fuel.
-------
II.7
The second policy option we analyzed was eliminating all
leaded gasoline. To estimate the costs for this case we used
the projected total 1988 demand for gasoline.* Here, again, we
had to allocate octane demand. We used the same assumptions about
misfuelers as before. We assumed people who owned leaded gasoline
vehicles would continue to require an average of 89 octane.
II.B. Reduction in Lead Emissions
Our analysis of reduced lead emissions assumed that every
gram of lead entering a car's gas tank came out its tailpipe.
In fact, some lead ends up in the oil (and may end up as waste
oil recycled for home heating) and some adheres to the exhaust
system and tailpipe, eventually flaking off. Ultimately, however,
all lead in gasoline ends up in the environment as a potential
source of lead contamination.
To estimate the reduction in lead emissions, we first
computed the number of tons of lead that would be removed in
1988 under our base case. We used gasoline demands from Table
II-l and assumed 1.1 grams of lead per gallon of gasoline (the
amount allowed under current regulations).
To calculate the tons of lead removed under the all unleaded
option, we took the volume of leaded gasoline that would be used
in 1988 assuming no changes in current rules or practices (i.e.,
* We may have overestimated costs by assuming that unleaded
demand would equal total demand in the all unleaded case.
We assumed demand would not change as a result of changing
prices, i.e., we assumed no elasticity of demand.
-------
II.8
1.1 grams of lead per gallon and continued misfueling). Multi-
plying that volume (32.4 billion gallons) by the lead content
(1.1 g/gal) gave us the total amount of lead reduced in the all
unleaded option. The result, shown in Table II-3, was 35,600
metric tons of lead removed.
For the low-lead option we needed to calculate the lead
emissions resulting from that reduction option (i.e., demand
[22.4 billion gallons] times 0.10 grams per gallon). Subtracting
the lead emissions under the low-lead option from 1988 emissions
based on no changes in rules gave us emission reductions of
33,400 metric tons, shown in Table II-3.
Table I1-3
Metric Tons of Lead Removed in 1988*
Low-Lead All Unleaded
33,400 35,600
*Computed by assuming 1.1 grams of lead per gallon and using
gasoline demands from Table II-l.
II.C. Cost Estimate
To estimate the costs of lowering the lead content of
gasoline, we used the Department of Energy's linear program-
ming model of the petroleum refining industry. The model and
oil refinery processes are described in greater detail in
Section II.F.
-------
II.9
Using the DOE refinery model and the assumptions described
above, we have estimated the cost differences for our two cases.
The costs, and their sensitivities to octane assumptions, are
discussed below. These costs have been estimated for several
different scenarios that indicate sensitivities to the basic
assumptions.
Our cost analysis indicated that reducing the amount of
lead in gasoline would involve relatively little capital cost.
This is because refiners overbuilt catalytic reforming capacity
before the 1978 Iranian revolution and were left with a surplus
as oil prices rose and gasoline demand fell. The capital costs
of this excess capacity are already sunk.
II.C.I. Incremental Cost of the All Unleaded Case
We computed costs for three different categories of octane
demand: a high octane scenario, a low octane scenario, and an
intermediate octane scenario. We also looked at how sensitive
our cost numbers were to changes in projected demand for gasoline,
We examined one additional factor that influenced costs.
There are several octane boosting additives besides lead on
the market. One of them, ethanol, receives large government
subsidies. If we allowed ethanol demand to vary among our cases,
and the model "saved money" by replacing lead with subsidized
ethanol rather than using a more expensive alternative, we would
be underestimating the cost of removing lead. We avoided this by
holding the quantity of ethanol used constant as lead was removed,
-------
11.10
Because other additives frequently contain fewer BTUs per barrel
than gasoline, whenever additive use increased we readjusted
total demand to keep BTUs, rather than volume, constant.
Case 1: High Octane Demand. If we assumed all misfueling
was for octane, not price, the annual cost of removing lead from
gasoline would be $759 million, of which $104 million was the
cost of moving misfuelers back to unleaded and $655 million was
the cost of eliminating leaded gasoline.
Case 2: Low Octane Demand. At the other extreme, if we
assumed that 50% of the people using 91 octane premium unleaded
would be satisfied by an 89 octane mid-grade unleaded, and that
50% of misfueling was due to price, the annual cost would decrease
to $538 million, of which $66 million was the cost of moving
misfuelers back to unleaded and $482 was the cost of eliminating
leaded gasoline. (The petroleum industry's Coordinating Research
Council studies of octane satisfaction suggested that about half
of the people using 91 octane premium unleaded would be satisfied
by 89 octane unleaded.)
Case 3: Intermediate Scenario. If we left all the premium
unleaded demand at 91 octane and assumed that half of misfueling
was due to price, the annual cost would be $691 million (of
which $104 million, as in case 1, was the cost of moving mis-
fuelers) . We have used this number in Summary Table 1 because
we believe that at least half of misfueling was due to price.
Also, we cannot be sure that premium unleaded users will switch
-------
11.11
to a lower grade, although we believe that some will. This point
estimate represents caution, not expectation.
Case 4: Volume Sensitivity. This was measured against
demand in the high octane case (6.5 million barrels), the most
expensive case. If gasoline demand were 6.75 million barrels
per day, our estimate for the high octane scenario would be
$761 million. If gasoline demand were 6.25 million barrels
per day, it would be $759 million.
II.C.2. Low-Lead Case
The 0.10 gram of lead per gallon of gasoline case resulted
in annual costs of $550 million in 1988, assuming all misfueling
were due to octane, and $503 million, if half were due to price.
In the low octane demand case, costs would be reduced to $410
million.
TABLE II-4
Cost of Reducing or Banning Leaded Gasoline
cents*/
Point Misfueling leaded
Range Estimate Portion gallon
(millions of 1983 dollars)
Low-Lead $410-550 $503 $104 1.66$zf
(0.10 g/gal)
All Unleaded $538-759 $691 $100 2.13jz?
* This is the increased cost of making gasoline under the two
options divided by the number of gallons of leaded gasoline
in the base case.
-------
11.12
As a check on the plausibility of the model, we examined
the spot price* differential between leaded regular and unleaded
regular for barge load quantities of fuel. This differential has
been between one and four cents/gallon for the last few years.
While spot prices can differ from manufacturing costs, they will
not differ for long periods unless there are supply constraints.
As the last column in Table II-4 indicates, when we allocated the
cost of removing all lead from gasoline to our projected leaded
gasoline demand, we obtained a cost per gallon well within the
range of market price differences between leaded and unleaded
regular gasoline. This confirms that our cost estimates are
reasonable.
II.C.3. Cost of Lead Reduction
The low-lead and the all unleaded options would reduce lead
emissions by about 33,400 and 35,600 metric tons, respectively,
in 1988. The cost per metric ton of avoided lead emissions,
therefore, would be about $15,100 for the low-lead option and
$19,400 for the all unleaded option. (These figures are not net
of vehicle maintenance savings, which we discuss in Chapter III.)
II.D. Price Differentials
Our estimates assessed incremental changes in manufacturing
costs; they do not indicate what changes might occur in consumer
* "Spot price" refers to the price of large quantity purchases
on the open market, as compared to long-term supply contracts.
-------
11.13
prices. Consumer price differentials between leaded, unleaded
regular, and unleaded premium gasoline currently are considerably
larger than the differences in manufacturing costs of the three
grades and considerably larger than the refiners' price differen-
tial to intermediate and bulk purchasers. For example, average
spot price differentials between leaded regular and unleaded
regular for barge load quantities in New York harbor were 1.29
cents per gallon in December 1983. The differential at the
Gulf termini of the pipelines bringing gasoline from the Gulf
to the Northeast was 1.1 cents per gallon. Contract price dif-
ferentials in the Gulf were 2.75 cents per gallon in Houston.
(The source of these price differentials is Platts Oilgram.) On
the other hand, retail price differentials are usually seven
cents per gallon. This indicated that most of the price differen-
tial was added at the retail level, and may be part of the retail-
ers' marketing strategy of cross-subsidization, where leaded
gasoline serves as a "loss leader" product.
Apparently, price differentials depend on market conditions
and oil company marketing strategies as well as costs. For
example, most gasoline marketers presently seem to be selling
regular leaded gasoline as a very low margin product, and are
making their profit on unleaded grades. This situation has
occurred in the past with regular or subregular leaded grades
vs. premium leaded gasolines. The two most common explanations
are that consumers shop on the basis of the lowest cost gasoline
offered regardless of whether they purchase that gasoline, and
-------
11.14
that the price elasticity of demand for gasoline is higher for
users of leaded gasoline, perhaps because they own older cars.
It is difficult and beyond the scope of this analysis to
predict what marketing strategies might be adopted if either
the low-lead or all unleaded policy options were implemented.
Under either of our hypothetical options, however, regular
unleaded gasoline would be the lowest cost product. In fact,
the model showed that in the 0.10 gram case the marginal cost of
making unleaded gasoline would decrease slightly from its
cost in the base case, while the costs of leaded gasoline and
premium unleaded gasoline would both increase by about one cent
per gallon. If marketers continue to make the lowest cost product
the "fighting" grade, then the current situation will invert,
with regular unleaded gasoline prices falling and leaded and
intermediate-grade unleaded becoming the high profit products.
The differences in prices that individual consumers pay will
depend upon changes in retail marketing strategies.
In this analysis, however, we used the real resource costs
of manufacturing to measure economic costs. We expect these to
reflect the differences in prices that consumers pay on average.
That is, we believe that all manufacturing costs will be passed
on to consumers, and that average retail margins will not
increase, although their distribution among grades may change.
II.E. Longer Term Projections
The costs for both the 0.10 grams per gallon and the all
unleaded cases will decline over time because the total demand
-------
11.15
for leaded gasoline will shrink as the fleet of vehicles designed
for leaded gasoline retires. Thus, these restrictions will
affect fewer gallons of gasoline in later years.
II.F. Refinery Model
Our estimates of the costs of lowering the lead content of
gasoline, given these various projections, were calculated using
the DOE linear programming model of the petroleum refining indus-
try. The model simulates current and projected U.S. refining
capacity, using available crude oil supplies, and projected
imports to meet expected U.S. petroleum product demands. The
objective function is to minimize costs, subject to constraints
on lead usage. The model recently has been subjected to two
verification checks by the Department of Energy (DOE), described
in Attachment 1.
II.F.I. General Description of DOE Petroleum Refinery Yield Model
The DOE Refinery Yield Model estimates optimal refining
industry operations under a range of assumptions and operating
conditions. The solution provides "optimum" petroleum flows,
prices, investments, etc., for the petroleum refining industry.
In addition to the optimal answer, the model provides valuable
economic information on important aspects of the refining indus-
try's operations, such as the rate at which costs change (the
marginal costs and values of specific refinery processes) as
refinery operations are altered to change the yield of products
or to accommodate different inputs.
-------
11.16
The model contains approximately 350 equations to simulate
the process by which crude oil and other inputs are turned
into various products and the costs that are thereby incurred.
The model can show which products can be made at varying costs
in the many different refineries that exist throughout the world.
It allows investment in new equipment in later years at a real
(constant dollar) capital charge of 15%.
The DOE model is based on many fairly similar models
developed and used widely by the petroleum refining industry for
years. The refinery industry model was one of the earliest
industrial applications of linear programming.
The basic model has been used by EPA in its analyses of the
impacts of regulations on the petroleum industry and on petroleum
product purchasers, and served DOE in many ways, including:
0 evaluating Strategic Petroleum Reserve crude mixes for
selections of storage sites,
0 assessing the impacts of petroleum disruptions on
product supplies, and
0 evaluating the industry's capability to respond to
changes in feedstock quality or product demands.
To understand the model, it is useful to describe briefly
how refineries work. Exhibit II-l is a schematic of a very
simple refinery, often called a topping plant, which processes
low sulphur crude oils. A complex refinery contains distil-
lation units and other types of processing units. Exhibits
II-2, II-3, and II-4 (provided by Sobotka and Company, Inc.)
illustrate schematics of such refineries. (The model presents
-------
11.17
considerably more detail than even these exhibits indicate.)
In all refineries, there is a selection of a combination of
different process "units" that can be assembled into final
structures that accomplish different but related purposes,
and that look similar. 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 takes in a raw material
(crude oil or an intermediate product) and makes one or more
intermediate or final products (and often some pollutants).
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., electri-
city 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.
Exhibit II-5 is a summary of the basic types of refinery pro-
cesses. Attachment 2 to this chapter 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 refinery units, each of which has an
output (or a series of products), the quantity of which is a
-------
11.18
EXHIBIT II-l
FLOW DIAGRAM OF TOPPING REFINERY
PROCESSING LOW SULFUR CRUDE OIL
GAS
GASOLINE HOW OCTANE)
8
L
E
N
D
N
G
LIGHT GAS OIL
HEAVY GAS OIL
RESIDUE
e
L
N
0
I
N
G
REFINERY
GAS FUEL
(CONSUMED
INTERNALLY)
PETCMEM FEED
a MILITARY
JET FUEL
KEROSENE,
DISTILLATE
'FUEL OIL 8
DIESEL FUEL
YIELD, VOLUME
PERCENT OF
RAW MATERIALS
3.1
33.1
RESIDUAL
> FUEL OIL
REFINERY
LIQUID FUEL
(CONSUMED
INTERNALLY)
'''Included with gas fuel
25.5
37.2
*
-------
11.19
EXHIBIT II-2
FLOW DIAGRAM OF HYDROSKIMMING REFINERY
PROCESSING LOW SULFUR CRUDE OIL
GAS
GASOLINE (LOW OCTANE)
GASOLINE (HIGH OCTANE)
CRUDE
OIL
RESIDUE
REFINERY GAS
FUEL (CONSUMED
INTERNALLY)
*LPG
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
25.5
FUEL (CONSUMED
INTERNALLY)
38.3
*
* Included with gas fuel
-------
11.20
EXHIBIT II-3
FLOW DIAGRAM OF FUELS REFINERY
PROCESSING HIGH SULFUR CRUDE OIL
r*
HYDROGEN
RECOVERY 8
UAMI iCArTllBE
IHYC
SULFUR 1
RECOVERY]
)ROGEN
NATURAL GAS LIQUIDS
GASOLINE (LOW OCTANE)
GASOLINE (HIGH OCTANE)
SULFUR
FOR HYDROGEN
TREATING
REFINERY GAS
FUEL (CONSUMED
INTERNALLY)
LPG
GASOLINES
YIELD. VOLUME
PERCENT OF
RAW MATERIALS
1.5*
11.0
2.3
53.8
KEROSENE,
JET FUEL,
DISTILLATE
FUEL OIL 8
DIESEL FUEL
(ASPHALT)
RESIDUAL
FUEL OIL
REFINERY
LIQUID FUEL
(CONSUMED
INTERNALLY)
27.4
8.1
**
* Percent by weight
** Included with gas fuel
-------
11.21
EXHIBIT I1-4
FLOW DIAGRAM OF HIGH CONVERSION REFINERY
PROCESSING HIGH SULFUR CRUDE OIL
SULFUR
RECOVERY
NATURAL GAS LIQUIDS
HYDROGEN
RECOVERY a
MANUFACTURE
HYDROGEN
(HIGH OCTANE)
GASOLINE (LOW OCTANE
GASOLINE (HIGH OCTANE)
LIGHT GAS OIL
HEAVY GAS
OIL
GASOLINE (MEDIUM OCTANE)
GASOLINE (LOW OCTANE)
I NAPHTHA
—I GAS OIL
SULFUR
___ FOR HYDROGEN
«-J TREATING
"T1 REFINERY
^ GAS FUEL
(CONSUMED
INTERNALLY)
GASOLINES
YIELD. VOLUME
PERCENT OF
RAW MATERIALS
1.4*
13.0
KEROSENE,
JET FUEL.
DISTILLATE
FUEL OIL a
DIESEL FUEL
REFINERY
LIQUID FUEL
(CONSUMED
INTERNALLY)
, HIGH SULFUR
COKE
77.5
10.8
**
4.6*
* Percent by weight
** Included with gas fuel
-------
11.22
EXHIBIT II-5
FUNCTIONAL CHARACTERIZATION OF PETROLEUM REFINERY PROCESSES
SEPARATION
A. Separation on the Basis of
Molecular Weight
Distillation (atmospheric and
vacuum fractionation of crude
oil, naphtha splitting,
depropanizing, stabilization)
Absorption (recovery of olefins
from 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)
B. 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
-------
11.23
function of the material that the unit is "fed." Each unit
incurs some costs that vary with how hard it is run -- called
"severity".
The model can be operated in either of two modes -- minimum
cost or maximum profit. It can constrain product quantities and
compute a minimum cost solution. (This is useful for analyzing
large refining regions in which aggregate demands can be forecast.)
Alternatively, the simulation can vary product quantities at
preselected prices.
The principal purpose of using computer models to simulate
petroleum operations is to measure differences between alter-
native scenarios in order to estimate the changes in petroleum
activities when some conditions change. Simulations of petroleum
activities are complex. The models are more reliable for deter-
mining differences in costs between scenarios than they are for
predicting the total costs of manufacturing all petroleum
products in the United States. So the major focus of analyses
should be differences between alternative model solutions.
These practical considerations should be kept in mind in the
interpretation of model results.
Exhibits II-6 and II-7 illustrate the basic structure of the
linear programming refinery model. All processes consist of a
series of linear relationships that describe the process output
and operating cost, given a specific input and a set of operating
conditions. The relationships are stored in the model in the
form of a process data table. Each column in this process table
represents the processing of a specific type of crude oil and
-------
11.24
each row represents a specific input or output stream, fuel,
utility consumption, etc. For example, the first column in
Exhibit II-6 specifies that as one barrel of Saudi Light crude
is processed, a mix of fifteen intermediate streams is created.
The operation consumes fuel, power, steam, and capacity, and
incurs variable operating costs of 9.2 cents per barrel.
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 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 combined by the model such that the qualities of the
blended mixes meet the minimum requirements of product
specifications.
Exhibit II-8 presents projected capacity in 1988 for
various processing units in the model.
-------
11.25
EXHIBIT II-6
YIELDS AND OPERATING COST COEFFICIENTS
CRUDE DISTILLATION UNIT
SAUDI ARAB LIGHT
SAUDI ARAB HEAVY
MEXICAN MAYA
CAPACITY FACTOR
Crude Oil Type
Saudi Arab Mexican
Light Heavy Maya
-1
1.0
-1
1.0
-1
1.0
Yields (Fraction of Intake)
STILL GAS
PROPANE
ISOBUTANE
NORMAL BUTANE
LT ST RUN (C5-175) LO 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 SULFER
HY GAS OIL (800-BTMS) (2.0% S)
ASPH VERY HI SUL (4.3% S)
.001
.003
.002
.013
.051
.070
.050
.020
.020
.020
.115
.015
.130
.180
.143
.001
.003
.002
.015
.040
.060
.044
.011
.020
.014
.090
.005
.090
.180
.300
.001
.003
.002
.009
.035
.025
.025
.010
.050
.005
.030
.070
.040
.100
.105
.350
Operating Cost Coefficients
(Per Barrel of Throughput)
FUEL, FUEL OIL EQUIVALENT
POWER, KWH
STEAM, LB
OTH VAR OP COST, $
CAPITAL CHARGE
-.021
-0.6
-60.7
-.092
-.022
-0.6
-63.4
-.093
varies
-.020
-0.6
-57.9
-.092
Note: The negative signs (-) indicate consumption of crude oil,
fuel oil, power, steam, etc.
-------
11.26
EXHIBIT II-7
YIELDS AND OPERATING COST COEFFICIENTS
CATALYTIC REFORMING UNIT
(200 PSIG Operating Pressure)
Paraffinic
Feedstocks
Naphthenic
Feedstocks
90 RON 100 RON 90 RON 100 RON
REF FEED (250-325) PARF
REF FEED (250-325) NAPH
CAPACITY FACTOR
-1
.95
_i
1.05
-1
.95
-1
1.05
Yields (Fraction of Intake)
H2 (100 PCT FOE)
STILL GAS
PROPANE
ISOBUTANE
NORMAL BUTANE
REFORMATS (90 RON)
REFORMATE (100 RON)
LOSS
.034
.036
.031
.020
.037
.852
.010
.041
.069
.076
.029
.052
.739
-.006
.,047
.025
.015
.003
.005
.930
-.025
.056
.036
.030
.007
.012
.886
-.027
FUEL, FUEL OIL EQUIVALENT
ELECTRICITY, KWH
STEAM, LB
OTH VAR OP COST, $
CAPITAL CHARGE
Operating Cost Coefficients
(Per Barrel of Thoughput)
-.042
-2.6
-75.
-.099
-.045
-2.6
-75.
-.108
-.042
-2.6
-75.
-.099
-.045
-2.6
-75.
-.108
varies
Note: The negative signs (-) indicate consumption of crude oil,
fuel oil, power, steam, etc.
-------
11.27
EXHIBIT II-8
ESTIMATED U.S. REFINERY PROCESSING UNIT CAPACITIES FOR 1988
(thousands of barrels per day)
PROCESSING UNIT CAPACITY
CRUDE DISTILLATION 15,900
COKER-DELAYED 1,175
COKER FLUID 170
VISCBREAKER 170
NAPHTHA HYDROTREATER 3,770
DISTILLATE HDS 2,670
FCC FEED HYDROFINER 1,085
RESID DESULFURIZER 580
CAT REFORMER 450 PSI 760
CAT REFORMER 200 PSI 3,145
FLUID CAT CRACKER 5,325
HYDROCRACKER - 2 STAGE 980
ALKYLATION PLANT 960
CAT POLYMERIZATION 77
HYDROGEN PLT, MBPD FOE 115
AROMATICS RECOVERY PL 300
PEN/HEX ISOMERIZATION 140
BUTANE ISOMERIZATION 55
LUBE + WAX PLANTS 240
-------
11.28
Attachment 1 to Chapter II
Evolution of DOE Refinery Model and Current Status
In late 1983 Decision Analysis Corporation and Sobotka & Company, Inc.,
jointly updated the Department of Energy's Refinery Yield Model (RYM) and
performed model verification tests for the Energy Information Administration.
The recent update involved revisions to the model's raw material availability,
product demands, and product specifications to reflect a 1982 environment.
Processing capacities were revised to represent operable capacity on January 1,
1983, as reported by DOE. In addition, the model's technical representations
were altered to reflect changes or improvements in processing technology that
have taken place since the original model development, to update major crude
assays, and to expand processing flexibility in the residual fuel portion of
the crude oil barrel.
The verification tests of the updated model were conducted to determine
how closely the RYM could simulate refinery activities in 1982. The verifica-
tion test runs on the updated model were designed to verify material balance
closure in the model solution and to assess the capability of the models to
simulate actual regional refining activities. Each regional model was run with
most crude and products specified at actual 1982 levels. The model then simu-
lated the 1982 operations with some flexibility to vary marginal feedstocks and
products. After the initial check for overall material balance closure, the
model results were compared with actual 1982 refining balances, process utiliza-
tions, and economic relationships. The verification tests and results are
discussed in more detail below.
Verification Methodology
The Refinery Yield Model (RYM) verification tests consisted of two simula-
tions for each model region, Verification A and Verification B, specified as
follows:
-------
11.29
Ve r ificatlon A; All crudes except for a marginal high and a marginal
low sulfur crude were fixed at the actual 1982 level^/ as were natural gasoline,
plant condensate, outside fuel and utility purchases, and unfinished oils.^/
The marginal crudes were permitted to vary within a range equal to about 2 to 3
percent of actual crude input. Butane purchases were also allowed to vary but
were not allowed to exceed actual. Product output was fixed at the 1982 level
except for liquefied petroleum gas (LPG) , coke, and low and high sulfur residual
fuel.
Verification B; All input was specified at the 1982 level. Gasoline,
distillate fuel, LPG, coke, and low and high sulfur residual fuels were allowed
to vary while all other output was fixed at 1982 volumes.
The primary purpose of the first simulation test, Verification A, was to
check the model for face validity. This included first a check for material
balance closure in the overall refining operations as well as in each process-
ing and blending operation. The results were then compared against actual
operations to check the ability to meet end product demand with available
feedstocks and to check the model's calculation of fuel consumption. Finally,
the initial simulations were checked to ascertain if model economics and pro-
cessing operations were within acceptable limits.
The second verification simulation runs, Verification B, allowed for an
additional check of model face validity. The refinery material balances and
projected economics were again checked against actuals. In this case, the
models were allowed more flexibility to optimize and would be expected to
operate major conversion processing at maximum. The product prices provided
2/ Actual crude types were estimated by the contractors, based on avail-
able DOE data.
£/ Actual natural gas input was assumed to be equivalent to reported
natural gas consumed for fuel. Actually, refiners may use additional natural
gas as hydrogen plant feed. In the district 13 model, a large volume of natural
gas was routed to the hydrogen plant, and therefore, natural gas purchases were
increased about 25 percent.
-------
11.30
are those which would result if all these facilities were in short supply
(which was the actual situation during 1982).
Verification B runs also provided an assessment of model overoptimization.
The volumes of light and heavy products produced from the 1982 volume of feed-
stocks were compared for each region run to evaluate the impact of overoptimiza-
tion of product yield capabilities. In this comparison, the. sum of gasoline
and distillate production was compared to actual rather than production of
individual products. The actual gasoline-distillate mix is a function of
regional weighted average price differentials for 1982. The available price
data are not sufficient to accurately test the model's projection of gasoline-
distillate production costs.
Verification A Results; The results of the initial verification showed
that the model was able to balance all material and account for all processing
streams. The model provides a summary for each processing and blending opera-
tion which includes a balance row indicating any stream not accounted for in
the model. The balance rows for all regional models were zero.
The model was able to produce a product slate close to actual operations
with available raw materials. The flows calculated by the model were very
close to actual figures. The model used about two percent less feedstock and
produced about two percent less output. The model calculated a four percent
loss of petroleum (products excluding refinery fuel) which is exactly the
actual 1982 loss. Refinery fuel was about five percent higher in the model,
indicating that the process efficiencies within the model may be slightly
under-estimated.
Crude and product prices varied from region to region, but in general were
reasonable. Gasoline and distillate prices were close, with regular gasoline
typically less than one dollar per barrel above middle distillate. Low sulfur
residual fuel was $6-8 per barrel below distillates and high sulfur residual
around $15 below. These results compare well with 1982 actual price differen-
tials.
-------
11.31
Verification B Results: The aggregate U.S. refining balance from Verifi-
cation B was close to actual 1982 operations. The models overstated the cap-
ability to produce light products by about 276,000 barrels per day (i.e., the
yield of gasoline plus distillate per barrel of crude was overstated by 2.3%).
The Verification B results indicated a large reduction in high sulfur residual
versus actual. The high sulfur residual reduction was also due in part to the
nature of the test. Refining regions were not required to produce a very low
sulfur fuel grade that is typical of some regions, and were thus able to blend
a greater volume of high sulfur components to low sulfur residual to meet
product demands. Fuel consumption calculated by the model was about 11% higher
than actual, but as a percent of total crude input there is less than a 1%
difference.
The combination of Verifications A and B provide substantial confidence in
the model's ability to predict the changes in costs and in operations that would
take place in the domestic petroleum refining industry if gasoline specifica-
tions, such as limits on the use of lead additives, were changed. And the model
also provides adequate flexibility in combining refinery process units so that
the same analytical question can be answered for subsections of the petroleum
industry categorized by size of plant or firm, or by the processing complexities
of plants, or geographically.
-------
11.32
ATTACHMENT 2 to CHAPTER II
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. Exhibit II-5 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 complet-
ed 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 C02« The regenerator flue gases are
passed through cyclones and, sometimes, electrostatic precipita-
tors, to remove entrained catalyst. They are then vented to the
-------
11.33
atmosphere or sent to a CO boiler where carbon monoxide is burned
to produce C02« 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 H2S and
ammonia are separated from the desulfurized product; the H2S
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-
-------
11.34
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°F 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, H2S 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
-------
11.35
converted to H2S. The Beavon process converts H2S to sulfur
through a series of absorption and oxidation steps. The SCOT
process concentrates the H2S 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.
-------
11.36
References
Dupont Petroleum Chemicals, Tech Brief 17909, December 1979.
Energy Information Administration, Annual Report to Congress,
1983.
Platts Oilgram Price Report, Price Average Supplement,
December 1983 Monthly Averages; Thursday, January 26, 1984
U.S. EPA, 1983 Motor Vehicle Tampering Survey 1982, National
Enforcement Investigations Center, April 1983.
U.S. EPA, 1982 "Regulation of Fuel and Fuel Additives,"
47 Fed. Reg. 49382 (October 28, 1982).
-------
CHAPTER III
BENEFITS FROM REDUCED VEHICLE MAINTENANCE REQUIREMENTS
Lead in gasoline has served both beneficial and destructive
functions. Refiners add lead because it is the least expensive
way to boost the octane of motor gasoline. Thus, for gasolines
of equivalent octane/ leaded gasoline would be less expensive to
make than unleaded gasoline. However, when vehicles burn leaded
gasoline, deposits are formed in the engine and exhaust system.
To reduce combustion chamber deposits, organic halogens —
primarily ethylene dibromide (EDB) and ethylene dichloride (EDC)
— are added to scavenge the lead.* These compounds react with
most of the lead to form compounds more volatile than those formed
with lead alone, and are discharged in exhaust gases. While this
effectively reduces combustion chamber deposits, a significant
portion still deposits on internal engine and exhaust system
surfaces. Such deposits (e.g., halogen acids and lead salts)
become very corrosive in the humid and warm environments within
engines and exhaust systems. For these and other reasons, the use
of unleaded gasoline reduces maintenance costs.
The deposits from leaded gasoline form a coating on exhaust
valve seats. On pre-1971 and some other vehicles, this thin layer
protects against the abrasive and adhesive wear that can occur
between the exhaust valve face and valve seat during certain engine
*Restricting or removing lead from gasoline would restrict or
remove EDB and EDC. This would have compounding environmental
benefits as EDB, EDC, and lead are substances of concern in
leaks from underground storage tanks and in tailpipe and
evaporative emissions. We have not included EDB or EDC benefits
in our analysis.
-------
III.2
operating modes. By 1971, however, several major engine manufac-
turers were building vehicles with valve seat metallurgy that had
minimized or eliminated valve recession with unleaded gasoline.
Because leaded gasoline combustion products form engine
deposits and corrode exhaust systems, studies have found four
main categories of savings in operating and maintenance costs
from switching to unleaded gasoline:
0 less corrosion of the exhaust train, requiring fewer
muffler and exhaust pipe replacements;
0 better engine performance due to less fouling and
corrosion of spark plugs;
0 less corrosion and rusting in the engine, decreasing
engine wear and allowing longer periods between oil
changes; and
0 better fuel economy, relating partly to better engine
performance (from the effects listed above) and
partly to the fact that unleaded gasoline contains
more energy content per gallon than leaded gasoline.
Quantitative estimates of this last benefit, however,
are less reliable than the others.
We discuss each of these in the sections under Maintenance Savings
(Sections III.A. and II.B. below).
As the remainder of this chapter indicates, switching
misfueled and other vehicles currently using leaded gasoline to
unleaded would likely produce millions of dollars in vehicle
operation and maintenance savings. The fuel economy benefits
are less certain and are not included in our summation of mone-
tized benefits. Most cars using regular leaded gasoline would
run as well1, or better, on unleaded gasoline of the same octane.
However, some older engines and non-diesel trucks require the valve
lubrication that lead in gasoline provides. Valve recession
-------
III.3
can occur in these engines from the inadequate lubrication of
exhaust valves, potentially resulting in premature valve failure.
The major constraint on an all unleaded policy in this decade is
a technical one: some vehicles need the valve lubrication currently
provided by lead in gasoline. The cost and practicality of
other solutions to this problem (e.g., other protective additives
or retrofits with improved valve seats) seem to pose significant
obstacles, but more information is needed to evaluate these
options.
As an alternative, we examined the maintenance benefits
and technical feasibility of a low-lead option (0.10 grams of
lead per gallon of gasoline) to lower sustantially the current
concentrations of lead in gasoline (1.1 grams per gallon), but
still allow sufficient lead to protect valves. The question of
the linearity of the maintenance benefits, and an estimate of
the dollar savings at 0.1 grams of lead per gallon of gasoline
follow the discussion of the all unleaded case.
We have not included three additional adverse effects due
to misfueling: plugging of catalytic converters, clogging of
exhaust gas recirculation (EGR) valves, and reduced performance
from exhaust gas oxygen sensors. In a recent communication
with the Motor Vehicle Manufacturers Association (MVMA, 1984),
these effects were raised, and subsequent contacts with automotive
engineers substantiated the engineering rationales for these
effects. The mechanisms for coating engine systems and exhaust
systems would also plug catalysts and EGR valves and coat oxygen
sensors. Extensive testing on another metal-based fuel additive,
-------
III.4
methylcyclopentadienyl manganese tricarbonyl (MMT), clearly
demonstrated the existence of catalyst and EGR valve plugging
and interference with oxygen sensors.
Catalyst plugging may result in back pressure problems.
Interference with oxygen sensors and closed-loop systems will
affect fuel metering systems regulating the air-fuel ratio.
Both affect driveability and fuel economy. Finally, plugging
the EGR valve can also adversely affect fuel economy,, knock,
and driveability. Some of these effects were observed with MMT
use and there is a plausible case for their occurrence with lead
use in cars designed for unleaded gasoline.*
MVMA (1984) valued the cost to a single consumer for catalyst
plugging at $300, for EGR plugging at $60, and for oxygen sensor
disruption at $75, all after 50,000 miles of driving. Given the
large number of misfuelers, over 12% of light-duty vehicles
(EPA, 1983), the aggregate costs could be large. However,
because we do not have the "dose-response" function for these
effects, we could not evaluate them under the regulatory options
examined here. Therefore, we have not monetized them. Excluding
these effects obviously underestimates the maintenance benefits
of reducing lead in gasoline.
* DuPont (1982) has observed severe catalyst plugging due to
lead in gasoline (0.5 grams/gallons); the implications of
this are still under study.
-------
III.5
III.A. Maintenance Savings in the All Unleaded Case
III.A.I. Sources of Data
In assessing the effects of lead on vehicle maintenance
requirements and the potential savings of switching to unleaded
gasoline, we evaluated nine studies and some independent ancillary
data. Most of these studies were conducted in the late 1960s
and early mid-1970s (Cordera, 1965; Pahnke and Conte, 1969; Pahnke
and Bettoney, 1971; Gallopoulos, 1971; Gray and Azhari, 1972;
Wintringham et al., 1972; Pless, 1974; Gergel and Sheahan, 1976;
Hickling Partners, 1981).
Concerning exhaust systems and spark plugs, we examined four
on-road vehicle studies involving nine samples of light-duty
vehicles in both commercial and personal use. One oil company
has provided a theoretical calculation based on its experiences
with the effects of gasoline quality on vehicles (reported in
Pahnke and Bettoney, 1971). For this analysis, we also examined
changes in automobile manufacturers' recommendations for vehicle
maintenance periods and the reasons for the changes, and we
quantified the portion attributable strictly to a switch from
leaded to unleaded gasoline. We have summarized the findings of
these studies for spark plug and exhaust systems in Table III-l;
we scaled the reported rates to reflect equivalent mileages to
facilitate the comparison of results.
There are fewer data concerning the effects of lead in
gasoline on oil change intervals, and some discussion is only
qualitative. In addition to drawing upon manufacturers' recommen-
dations, we used four sources of information. One was research
-------
III.6
TABLE III-l
SUMMARY OF STUDIES ON MAINTENANCE DIFFERENCES BETWEEN LEADED AND UNLEADED VEHICLES
STUDY
REPLACEMENT RATES
PER 11,000 MILES (OR PER 1 YEAR)
Pahnke & Bettoney
(DuPont, 1969)
Humble
(rpted. in Pahnke &
Bettoney, 1971)
Gray & Azhari (1972)
(Amoco)
MY 1967:
MY 1968:
Gray & Azhari (1972)
(Amoco)
Wintringham et.al. ,
(Ethyl, 1972)
Detroit:
Baton Rouge:
Hickling Partners
(Environment Canada)
(1981)
Municipal Fleets
Changes in
Manufacturers
Recommendations
SPARK PLUGS
UNLEADED
.534
.330
.373
.307
.247
weight.
.440
.347
2.9 times
LEADED
.726
.550
.840
1.085
.295
3d avg.
.677
.519
as many
w/leaded vehicles
2.2 more replace-
ments for
vehicles,
leaded
net of
other technology
changes
EXHAUST SYSTEMS
UNLEADED
LEADED
.0033 .187
.220 ' .275
.149
0
.004
weigh tec
.535
.217
.071
1 avg.
.155 .289
.004 .358
2.4 times as
many for leaded
vehicles (they
exclude Toronto
fleet)
(Not Applicable)
AVG. MPV/YR
11,400
10,000
7,500
7,500
Not reported
14,575
16,850
(Unknown)
(Not Applicable)
ACCUMULATED
AVG. MVP
65,000
—
24,000
17,000
1 to 6 yrs.
72,883
84,260
23,810 leaded
24,990 unleaded
(Not Applicable)
TYPE OF SERVICE
Personal Use
Theoretical
Commuting and
business use
Personal Use
(Consumer
Panel)
Employee Fleet
(Business and
Personal Use)
Municipal Service
(Not Applicable)
# 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
(Not Applicable)
LOCATION
South New Jersey
and Wilmington,
Delaware
—
Chicago and
suburbs
Eastern states
concentrated
in Mid-Atlantic
Detroit
Baton Rouge
Montreal
Ednonton
Toronto
(Not Applicable)
-------
III.7
conducted on four fleets of commercial vehicles under conditions
that strain oil performance (Pless, 1974). Another study used
engine tests on a road simulator to compare the use of leaded
gasoline at standard oil change intervals with unleaded gasoline
at extended intervals (Gergel and Sheahan, 1976). The third
source was a detailed analysis of the potential lengthening
of periods between oil changes by switching to unleaded gasoline
(Gallopoulos, 1971). Finally, Cordera et al. (1965) related
engine rust build-up to lead concentration in gasoline.
In addition, some studies found other categories in which
unleaded vehicles experienced lower maintenance expenses -- notably
fewer carburetor adjustments (Gary and Azhari, 1971; Wintringham
et al., 1972) and fewer engine overhauls. We did not include them,
however, for several reasons: some of these effects may not be
related exclusively to differences between leaded and unleaded
gasoline, several studies used data bases that were too small to
support meaningful conclusions, and some were not considered
reasonable to extrapolate to vehicles operating in 1988.
III.A.2. General Comments on the Method
In quantifying the consumer benefits of switching from leaded
to unleaded gasoline, we considered changes in the observed main-
tenance behavior of vehicle owners. For matched pairs of vehicles
and drivers, changes in observed maintenance reflect, and are used
as a proxy for, underlying effects of gasoline quality on vehicle
performance and durability. If most people maintain their vehicles
at manufacturers' recommended schedules, and would continue to do
so with a switch to unleaded gasoline, our method could overestimate
-------
III.8
maintenance benefits. This would also be true if manufacturers'
recommended schedules were based on the performance and durability
of the worst cars, rather than average cars. in this case,
scheduled maintenance may provide a large safety factor relative
to the average car, and we may have overestimated maintenance
savings.
On the other hand, manufacturers may develop maintenance
schedules by balancing the extra maintenance expenses of the
average or better vehicles against the expected avoided costs of
the more problem-prone vehicles. In this case, our evaluation
of changes in maintenance behavior probably does not overstate
benefits.
In any case, the evidence — from the fleet studies we cite
here, consumer surveys, and conversations with auto specialists
— indicates that, in general, people substantially under-maintain
their vehicles relative to recommendations. (See, for example,
the 1984 AAA Potomac Division survey that found most of 2,600
cars suffering from maintenance problems.) In sum, we expect
that observing owners' behavior correctly reflects the intervals
at which they begin to notice performance degradation. (An
exception to this is exhaust systems, which comprise half the
estimated savings, because people repair these only when they
fail.)
III.A.3. Fewer Replacements of Exhaust Systems
All of the studies found demonstrable differences in
expected lifetimes (measured in miles) of exhaust systems between
matched pairs of unleaded and leaded vehicles. The range of
-------
III.9
estimated differences between leaded and unleaded replacement
rates, however, was very broad, from only 20% fewer muffler
changes (at equivalent mileage) for unleaded vehicles (but based
only on a theoretical calculation) to, more commonly, virtually
no replacements for unleaded vehicles in four of the nine distin-
guishable fleets. 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.
Unfortunately, however, 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 travelled another 10,000 to 20,000 miles. The
reported findings, thus, may have overestimated the differences
between unleaded and leaded vehicles.
It is useful to look most closely at the Ethyl Corporation
(Wintringham et al., 1972) findings, since their vehicles had the
greatest mileage, and there is a clear geographic distinction
between the fleets. Their Baton Rouge fleet, after over 84,000
miles of travel per car (compared to a projected lifetime of
100,000 miles), had essentially zero exhaust system repairs for
unleaded vehicles, but rates 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
-------
III.10
unleaded vehicles of one exhaust system repair per 46,000 miles,
but rates of one 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. This was consistent with
the Environment Canada findings (Hickling Partners, 1982) for two
municipal fleets, which had 42% fewer exhaust system replacements
(at equivalent mileage) for cars using unleaded fuel in cold
climates.
On the other hand, the DuPont (Pahnke and Bettoney, 1971)
and Amoco (Gray and Azhari, 1971) findings, conducted in the mid-
Atlantic region, 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. (The average muffler replacement
rates for leaded cars among the different studies also varied
greatly, ranging from 1 per 20,500 miles to 1 per 154,900 miles.)
Weighting Ethyl's findings for Detroit and Baton Rouge
according to the portion of registered cars in Sunbelt versus
Snowbelt states in 1982 (43% and 57%, respectively, according
to MVMA, 1983), mufflers nationally would last an average of
three times longer on unleaded vehicles than on leaded ones.
However, because of our concern that these limited duration
studies may have underestimated muffler replacements over the
lives of vehicles using unleaded fuel, we conservatively concluded
that mufflers on vehicles using unleaded fuel would last twice
as long (in miles) as those on vehicles using leaded fuel.
Given the projected lifetime of a car (100,000 miles), this
-------
III.11
meant about two exhaust system changes per leaded vehicle versus
one per vehicle using unleaded.
We assumed mufflers on vehicles using leaded gasoline would
last about 50,000 miles. In the studies we reviewed, the leaded
fleets averaged about 40,000 to 60,000 miles between exhaust
system replacements. Several automotive specialists independently
confirmed the reasonableness of this assumption.*
We calculated exhaust system replacement savings as follows:
for leaded exhaust systems replaced once every 50,000 miles,**
each mile therefore accounts for .00002 of the system replacement;
for unleaded vehicles replaced once every 100,000 miles (doubled
exhaust lifetime), the system replacement figure is .00001/mile.
The difference is .00001/mile. At $120 per repair (muffler,
tailpipe, and exhaust pipe), this was 0.12 cents/mile, or 1.68
* Passing references in literature and several reviewers of
this document 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, this design
improvement might affect performance so that our estimates
of benefits might be substantially overstated. 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 made of rolled
steel. These are the parts that we estimated would corrode
from leaded gasoline. Thus, this technology change should
have no effect on our estimates of savings.
** It can be argued that effects due to fuel use are best deter-
mined in terms of total gallons consumed rather than miles
traveled. For a majority of the studies examined in this
paper, fuel consumption data were unavailable. Thus, we used an
assumed value of 14 miles per gallon. To the extent that this
value is higher or lower than the actual fuel economy of the
vehicles in the studies used here, our estimates will vary
accordingly.
-------
III.12
cents/gallon (at the average 14 miles per gallon achieved by
cars in the late 1960s).* A savings of 1.68^ per gallon times
light-duty vehicle demand (22.8 billion gallons of leaded
gasoline in 1988) yielded exhaust system savings of $383
million in 1988 for the all unleaded case (1983 dollars).
III.A.4. Better Performance or Less Frequent Spark Plug Changes
The second category of maintenance savings is better vehicle
performance by avoiding the fouling and corrosion of spark plugs
by lead deposits. The fleet studies results were more consistent
in establishing spark plug effects than exhaust system effects.
Eight fleets in four studies (Pahnke and Bettoney, 1971; Gray
and Azhari, 1971; Wintringham et al., 1971; and Hickling Partners,
1982) clearly showed that owners of vehicles using unleaded fuel
increased mileage intervals between spark plug changes by 35% to
* Changing to savings per gallon, then extrapolating to 1988 via
changes in leaded gasoline demand, automatically adjusts for
changes in fuel economy and changes in miles per year among
different cohorts of vehicles. Vehicles traveling fewer miles
would burn fewer gallons and, hence, acquire fewer savings.
Likewise, vehicles with better fuel economy would achieve lower
savings than average. It should be noted that our benefits
estimation assumes that these savings are a function of fuel
use. Given the current trend towards more fuel efficient cars,
such an assumption may considerably underestimate actual bene-
fits, as the age of the vehicle becomes an important variable
in determining the life of a muffler. Implicit in our model
is the assumption that an automobile that gets 28 rnpg will need
a new muffler every 200,000 miles, or at 42 mpg, 300,000 miles.
To the extent that these muffler lifetimes are overestimated,
benefits are underestimated. Unfortunately, we were constrained
by the lack of data concerning the effects of time on muffler
life.
-------
III.13
300% over the intervals for leaded vehicles. The average of the
studies was about a 60% increase in the distance traveled between
spark plug changes on unleaded versus leaded vehicles. We used
this change in replacement to compute the savings of lowering
lead concentrations from the pre-phasedown level of 2.3 g/gal to
zero lead.
By comparison, the Environment Canada/Hickling Partners
study (1982) found over a 50% increase in spark plug life for
unleaded cars in the municipal fleet studies. They also found
almost a doubling of the intervals recommended by auto manufac-
turers for spark plug changes on unleaded vehicles compared to
leaded ones, a function of several technological improvements
(e.g., the addition of high energy ignition systems).
However, some evidence suggested that the 1982 lead phase-
down rule making, which lowered average lead concentrations to
1.1 grams per gallon, has already achieved a portion of this 60%
increase in spark plug life. (Other data suggested that the same
is not true for exhaust systems and oil changes — lowering lead
to 1.1 g/gal may not have provided savings in these categories.
These are discussed in greater detail in section III.B. of this
chapter.) Therefore, we pro-rated the 60% savings according to
the portion left to be gained by further restrictions of lead.
There are scant data on spark plug fouling at very low lead
levels. In 1971, Toyota (Champion, 1971) reported finding that
fouling of spark plugs occurred at equivalent rates with unleaded
and low-lead gasoline of 0.20 g/gal (both maintaining ignition
-------
III.14
performance for 30,000 miles). At the 1972 Champion Spark Plug
Conference, Union Oil also reported that spark plug performance
was similar for unleaded and low-lead gasoline of 0.5 g/gal.
Both outlasted by over four times the spark plugs operating with
leaded fuel of 3.0 g/gal. These findings suggested that there
was some threshold for gasoline lead content above which the
lead in gasoline degraded spark plug performance. For lack of
other information, we assumed that this threshold was 0.5 g/gal,*
and that the relationship between lead and spark plug fouling
was linear from this threshold to higher lead levels.
Earlier, we noted that intervals between spark plug changes
could increase 60% by reducing lead from 2.3 g/gal to zero
(0.0 g/gal.) If the threshold by which all benefits have been
achieved is 0.5 g/gal, then the 1982 lead phase-down rule (which
lowered the lead content of gasoline to 1.1 g/gal) would have
resulted in (2.3-1.1)/(2.3-0.5) times 60% = 40% (estimated
increase in change intervals). Thus, we have achieved 40% of
these benefits already. The remaining 20% (i.e., 60% - 40%) gain
would be achieved by phasing down from 1.1 g/gal to 0.5 g/gal
(or to 0.0 g/gal). We used this 20% rate of savings to calculate
the benefits of fewer spark plug changes.
We assumed that drivers of vehicles using leaded gasoline
began to experience significant performance degradation by about
* The Toyota and Union Oil results could have been averaged for
a threshold of 0.35 g/gal. Use of a lower threshold will
result in benefits of 23% increased spark plug life in phasing
down from 1.1 to 0.35 g/gal (Champion, 1971, 1972).
-------
III.15
12,000 miles of spark plug life. This was a little longer than
automobile manufacturers' recommendations would imply, but some-
what less than the actual change intervals for leaded vehicles
observed in most of the fleet studies. The observed intervals
averaged about every 15,000 to 16,000 miles (but ranged from
10,000 to 37,000 miles for leaded vehicles).
A 20% increase in spark plug life accompanying a switch to
unleaded gasoline would provide savings of about 0.35 cents/
gallon of gasoline.* This, multiplied by the projected 22.8
billion gallons of demand for leaded gasoline in 1988 in the
base case, translated to savings from fewer spark plug changes
in 1988 of about $80 million under an all unleaded policy.
Interestingly, the effects appeared smaller than the
researchers had hypothesized. Apparently, owners tuned up their
vehicles and changed the spark plugs more as a function of
mileage (and habit) than performance. Using the difference in
observed behavior between paired drivers of leaded and unleaded
vehicles, as these studies did, may underestimate the performance
degradation of leaded gasoline on spark plugs and engine timing.
* Calculation: A 20% increase in the 12,000 miles between
spark plug change experienced in cars using leaded gasoline
translates to 14,400 miles between changes. The difference
in the number of changes per mile is therefore 1/14,400 -
1/12,000 or .000014/ mile. Given a price of $18 per spark
plug change1, this becomes .025^/mile, or .35jd/gallon (at
14 miles per gallon).
-------
III.16
In fact, using the data from MVMA (1984) reveals an estimate of
$328 million* — over four times the value derived In our
analysis.
III.A.5. 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% of the
lead in gasoline ends up in the used oil, comprising up to 50%
of the weight of engine oil sludge (Gallopoulos, 1971).
The particles that accumulate in the used oil cause sub-
stantial abrasive wear in the engine, while the internal engine
rust may cause hydraulic valve lifter sticking (Cordera et al.,
1965). Besides the long-term engine wear that reduces the dura-
bility of the engine, the vehicle driver may also experience
excessive valve noise and other performance degradation due to
this premature contamination of oil. While rusting can occur
even in the absence of the halogen acids derived from lead salts,
engine oil tends to be the major cause of internal rusting under
normal driving conditions.
* MVMA estimated spark plug changes to occur every 30,000 miles
for vehicles using unleaded gasoline, compared with every
15,000 miles for those using leaded gasoline. This yields
1/15,000 - 1/30,000 = .000033 spark plug changes/mile
difference. At $18 per plug change this is ,06^/mile or
1.44jzf/gallon using the MVMA's figure of 24 miles per gallon.
Given an estimate of 22.8 billion gallons of leaded fuel in
1988, total benefits are just over $328 million.
-------
III.17
The fleet studies investigating differences in maintenance
costs between unleaded and leaded vehicles tended either not to
consider effects on engine oil, or found very small savings. In
general, these studies were not conducted in a manner to deter-
mine easily the effects on oil change intervals or engine wear
from using leaded or unleaded gasoline. Possibly consumers were
not aware of the potential decrease in oil change requirements
when using unleaded gasoline, and/or did not tend to change their
habitual maintenance behavior. Therefore, this analysis relies
more heavily on experimental studies of engine wear with unleaded
and leaded gasolines at varying oil change intervals than the
fleet studies. The exception is Pless (1974) which was a fleet
study specifically designed to examine oil change effects. Even
if consumers did not realize the possible short-term cost savings
of fewer oil changes, they would have benefited from better
engine durability with unleaded gasoline. Since most evidence
indicates that vehicle owners do not change oil often enough,
this would be especially true.
Many of the experimental studies in the early 1970s on oil
change requirements did not provide conclusive evidence on oil
quality after extended intervals between changes. The results
consistently did show that unleaded gasoline decreased rusting,
corrosion, and sludge; low temperature piston varnish tended to
increase, however. No significant difference was found for oil
thickening, high temperature varnish, or adhesive wear or
scuffing. It was not clear whether, overall, unleaded fuel
would allow substantially longer intervals between oil changes.
-------
III.18
In any case, manufacturers have changed their specifications for
oil changes from every 3-5,000 miles to every 7-10,000 miles.
A study by Gallopoulos, of the General Motors Corporation
(GM), was one of the earliest works that we examined. He
concluded in 1971 that with unleaded gasoline it might be feasible
to extend requirements from 2-3 yearly changes to only one annual
oil change, but added that more investigation was needed.
Pless, also of GM, reported more conclusive results in 1974
from experiments on taxicabs in conditions that take an unusually
severe toll on oil quality. In a group of twenty 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. However, the unleaded vehicles experi-
enced more oil filter plugging and higher used-oil viscosity.
On a fleet of 1971 taxis, he found that doubling oil change
intervals with unleaded gasoline (from 8,000 miles to 16,000
miles) significantly increased oil filter plugging and used-oil
insolubles.
On a fleet of 1972 taxis, Pless 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 some-
what greater plugging of oil filters, Pless concluded that this
-------
III.19
was not a significant finding. Finally, another fleet traveling
predominantly short trips (closer to "typical" consumer driving
patterns) led Pless to conclude:
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 'standard1
oil-change interval.
He qualified this conclusion by stating that only unleaded
gasoline and SE or better quality oils be used. (Currently, SF
oils, which are better than SE, are the most widely used.)
Subsequent to these findings, both GM and Chrysler recommended
lengthened periods between oil changes. Both companies now only
manufacture cars built to run on unleaded fuel.
In 1976, Gergel and Sheahan (Lubrizol Corp.) found results
similar to those of Pless, but found no significant plugging of
oil filters. Importantly, they concluded that engine wear was
the limiting factor in extending oil change intervals suggesting
a maximum of 12,000 miles between changes for leaded 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 (either through reduction in
abrasive lead particles or rust), oil degradation, and general
engine and engine part cleanliness (e.g., lack of deposits and
sludge). For analytical purposes, we need to determine the
-------
III.20
functional relationship between lead in gasoline and oil change
intervals. The available direct evidence is from Pless who tested
engine oil change intervals on unleaded gasoline and 2.70 grams/
gallon leaded gasoline. However, the existing lead phase-down
regulation limits the lead content to 1.1 grams/gal. Given this
data, the issue is whether some of the benefits of reduced oil
change intervals already occurred in going to 1.1 grams/gal, and
how much remains to be obtained in decreasing from 1.1 grams/gal
to 0.1 or 0 grams/gal.
Gallapoulous, in discussing future engine oil requirements
for unleaded vehicles, noted a number of studies which examined
lead or lead scavenger use in relation to engine or engine part
rusting. He concluded that the use of unleaded gasoline would
result in less internal rusting. The inference was that with
less sludge, oil degradation, and deposit build-up, the overall
task of engine oils is reduced. As a result, a switch to unleaded
gasoline would produce a net increase in engine oil lifetime.
With this engineering data in mind, we examined the studies
relating lead additives or lead scavengers to engine rust. While
it may be argued that most of these studies were designed to
identify lead scavenger effects, it is also true that such scav-
engers would not be used in the absence of lead in gasoline.
Furthermore, a substantial portion of the corrosive elements in
the engine are acids derived from the lead halide salts, a product
of both lead and its scavengers. In fact, all of the studies
looked at various lead concentrations as well as lead scavenger
concentrations.
-------
III.21
One notable study (Cordera et al., 1965) examined the
relationship betweeen engine rust and lead scavenger concentra-
tions, while varying lead content. Cordera et al. showed that
in addition to a relationship between lead scavenger concentra-
tion and degree of internal rust, there also was a relationship
between lead concentration and rust. These authors evaluated
valve lifter rusting at 0, 0.53, and 3.2 grains of lead per gallon
of gasoline. The level of rust decreased non-linearly with
decreasing lead content. An examination of the data indicated
that in going from roughly 2.3 grams (pre-phasedown) to 1.1
grams there was a 12.7% improvement. From 1.1 to 0.1 g/gal
there was an additional 58.3% less rust, and from 0.1 to 0 there
was an additional 29% improvement. Thus, going from a current
gasoline lead level of 1.1 grams to 0.1 grams would yield 58.3%
of the benefits of eliminating lead and its scavengers, whereas
going from 1.1 to 0 gives 87.3%.
The preponderance of evidence indicated that using unleaded
gasoline decreases oil contamination, engine wear, and rust, even
with a doubled oil change interval. We believe a decrease from
2.3 grams/gallon to zero would yield dollar savings at least as
great as those which would accrue by the doubling of oil change
periods. (Vehicle manufacturers recommended such a doubling for
their vehicles concurrently with the switch to unleaded gasoline.)
It is important to note that even if owners did not change main-
tenance behavior, i.e., if they continued with their prior oil
-------
III.22
change intervals, they would still achieve longer engine dura-
bility from the greatly preserved oil quality when using only
unleaded gasoline, and therefore achieve long-term savings.
Manufacturers' specifications have changed from one oil
change roughly every 3,000-5,000 miles, to about one every
7,000-10,000 miles. This translated to about one or two —
instead of two or three -- oil changes per year. We assumed an
oil change required 4 quarts of oil at $1.50 each, that oil
filters ($4 each) would be replaced every other oil change (so
$2/change) and we assumed 15 minutes of labor. (We valued that
labor at an hourly wage rate of $10.00, the average for manufac-
turing.) This calculation was for a "typical" owner changing
his/her own vehicle's oil and would be substantially less than
the prices people generally pay at service stations. This
yielded $10.50 per avoidable oil change, or savings of about
1.47 cents per gallon of gasoline.* In 1988, this produced
additional savings of $332 million for the all unleaded case.
But note, in the study by Cordera et al. we found that only
87.3% of these benefits were achieved in going from 1.1 to 0
grams of lead per gallon. Thus, we have lowered this value by
* For vehicles using leaded gas, one oil change every 5,000
miles was assumed versus one every 10,000 miles for the
vehicles using unleaded gas. Therefore, 1 change/5,000
miles minus 1 change/10,000 = .001/ mile. At $10.50 per
oil change and an average fuel economy of 14 miles per
gallon, 1.47 cents/ gallon is the average savings. With
22.8 billion gallons projected consumption in 1988, the
value is $332 million.
-------
III.23
12.7% for a savings of $292 million. Again, this is less than
the value of $547 million predicted from using the MVMA analysis.*
III.A.6. Improved Fuel Economy
There are three reasons why drivers could expect to get
better fuel economy by switching from leaded to unleaded
gasoline:
0 Unleaded gasoline has more energy content per
gallon. These small per-gallon savings would
accrue to any consumer of unleaded, rather than
leaded, gasoline (Exxon, 1978).
0 Lead fouls spark plugs, which hurts fuel economy.
The benefits of avoiding this would be counted
mostly by our spark plug estimate.
0 For vehicles built after 1980, misfueling with
leaded gas affects oxygen sensors, which can
adversely affect fuel metering.
Energy Content
An analysis by Exxon (1978) on the energy content of
different kinds of gasoline showed that vehicles using unleaded
gasoline should get better mileage because unleaded gasoline
contains more aromatic compounds and is "denser" (i.e., has higher
energy content per unit volume) than leaded gasoline. Also,
engines that run on unleaded gasoline build up more deposits in
* The MVMA assumed unleaded gasoline vehicles require an oil
change every 7,500 miles versus 5,000 miles for leaded gas
users. The difference is, therefore, .00007 change/mile.
Using MVMA estimates of $15/oil change and an average of 24
miles per gallon, this produced a value of 2.4 cents/gallon,
or $547 million in 1988.
-------
III.24
the combustion chambers which tend to increase the compression
ratio, thereby improving engine efficiency slightly. For these
reasons, vehicles burning unleaded gasoline should be slightly
more fuel efficient. The Exxon memo calculated that improved
fuel economy might be about 1 to 1.5%. Using a Society of
Automotive Engineers formula (SAE #J1082b) to adjust miles per
gallon for differences in gasoline density, and using the Exxon
density estimates, we computed about 0.8% fuel economy improve-
ments from using regular unleaded gasoline. This produced savings
of about $199 million in 1988.*
We are not sure what the difference in density may be between
future grades of leaded and unleaded gasoline. Because of this
uncertainty we did not include these savings in our tabulation
of benefits.
Spark Plug Fouling
Spark plug fouling caused by leaded gasoline also reduces
fuel economy. This loss necessarily occurs in the interim
between spark plug cleanings and changes, not just if maintenance
does not occur. High energy ignitions, used in most vehicles by
the mid-1970s, help extend spark plug life by maintaining reli-
ability and may have some impact on delaying fouling and adverse
fuel economy impacts. We probably have included much of this fuel
economy increase in the previous section on spark plug savings, so
* $199 million = 22.8 billion gallons of light-duty vehicle
demand for leaded gasoline in 1988 times $1.10 per gallon
times 0.8% fuel economy improvement.
-------
III.25
we did not include it again here. But as a check on our previous
estimate of spark plug savings, we calculated the fuel economy
loss if spark plugs were not changed frequently enough. The
fuel economy penalty of extra spark plug fouling would have to
be only 0.32%* to be comparable to the estimated spark plug
savings of $79 million in Section III.A.4.
Oxygen Sensor Fouling
For some misfueled vehicles, lead deposits can also affect
oxygen sensors causing engines to run richer and thereby reducing
fuel economy. How much this occurs depends on the types of
feedback and failure modes of specific electronic controls, as
well as how particular oxygen sensors react to the introduction
of lead.
EPA's Office of Mobile Sources has estimated that the
impact of these factors on gasoline consumption could be from
0-15% of misfuelers1 consumption. Arbitrarily taking a low
estimate from their range — a 3% loss** — we estimated roughly
what preserving fuel economy might be worth for would-be
misfuelers in 1988. The 3% loss, times 10.3 billion gallons of
* Given retail gasoline prices of $1.10, our estimate of
$79 million from spark plug savings is equivalent
to 71.8 million gallons of gasoline. This, divided by
the 22.8 billion gallons of light-duty vehicle demand
for leaded gasoline, is equivalent to a .32% increase
in fuel consumption.
** Four Canadian studies have estimated that fuel economy
may be up to 4% greater for vehicles using unleaded
gasoline. However, the applicability of these findings
to the U.S. situations is questionable, so we did not
use them in this analysis.
-------
III.26
misfuelers1 demand, times $1.10 per gallon, equals $339 million
in 1988. Currently, we have insufficient data to estimate this
more precisely, or to include it in our tabulation of benefits.
III.B. Maintenance Savings for the Low-Lead Case
The previous sections estimated the maintenance benefits
likely to result from an all unleaded policy. If leaded gasoline
were unavailable, however, some vehicles might experience excess
valve wear. This risk suggested that we evaluate an option lower-
ing the concentration of lead in gasoline to a level that still
would be sufficient to protect against valve recession. (Valve
recession and alternatives to prevent it are discussed in the
next section of this chapter.)
This section examines the relationship between lead
concentrations and maintenance benefits at high, low, and no lead
levels. We then discuss the savings likely from a low-lead case
allowing 0.10 grams of lead per gallon of gasoline. To estimate
these benefits we had to assess the shape of the effects function
in order to interpolate between the relatively high Lead levels
at which most research has been conducted (about 2.3 to 3.0 g/gal)
and zero lead. With that information, we calculated the portion
of the "all unleaded option" savings that would be achieved by
the low-lead option.
Data with which to interpolate the relationships, and thus,
to estimate savings, were scant. We are confident that current
lead concentrations of 1.1 g/gal are above the threshold where
-------
III.27
lowering lead levels would result in savings related to exhaust
systems or oil changes. However, it is likely that the 1982
lead phase-down regulations may have already achieved some of
the potential spark plug savings of going from 2.3 g/gal to 0
lead.
III.B.I. Exhaust System Savings
Most of the studies we evaluated to estimate maintenance
savings involved fleets of vehicles, half of which used commer-
cially available leaded gasoline. In the late 1960s, when
these studies were conducted, the weighted average lead content
of gasoline (weighted by the portions of premium and regular
grades) was about 2.3 grams of lead per gallon. Unfortunately,
because the discussion at the time focussed on relatively high
lead levels versus "zero" lead, there are extremely few data
with which to define the relationship between low lead concen-
trations and exhaust system corrosion between 2.3 and 0 g/gal.
Gray and Azhari (1971) was the only study that examined
exhaust system corrosion rates at lead levels as low as 0.5 g/gal,
They found no difference between corrosion rates at 2.3 g/gal and
0.5 g/gal, with corrosion rates at both lead levels 10-20 times
higher than those of vehicles using unleaded gasoline. This
suggested that there was some threshold at or below 0.5 g/gal,
below which lead levels must fall before any savings may be
achieved by fewer muffler replacements. It also suggested that
no savings were achieved from previous "lead phase-downs", since
-------
III.28
the current lead concentration is 1.1 g/gal — well over Gray and
Azhari's upper bound threshold of 0.5 g/gal. With no information
to the contrary, we assumed that the relationship between lead
levels and exhaust corrosion was linear below this threshold.
To calculate the exhaust system savings at 0.1 g/gal, we
distinguished between two categories of would-be users of leaded
gasoline: misfuelers and those vehicles designed to use leaded
gasoline. Because of the likely changes in price differentials,
marketing strategies, and possible administrative controls on
the distribution of leaded gasoline, we assumed that there would
be no misfueling under the 0.1 g/gal low-lead case. For consumers
who previously had misfueled, the savings would be the same under
both the low-lead and no lead cases: 1.68 cents/gallon savings,
or $173 million in 1988.*
For light-duty vehicles designed to use leaded gasoline,
savings in the low-lead case would be (.5-.I)/.5 or 80% of the
all unleaded savings for fewer exhaust system replacements.
This would equal 1.34 cents/gallon, or $168 million in 1988.**
Adding this to the savings for misfuelers, we estimated the
exhaust system replacement savings under the low-lead option
would be $341 million in 1988.
* $173 million = 1.68 jzJ/gal times 10.3 billion gallons of
misfuelers1 demand in 1988.
** $168 million = 1.34 jz?/gal times 12.5 billion gallons of
legal (non-misfueling) light-duty vehicle demand for
leaded gasoline.
-------
III.29
III.B.2. Spark Plug Savings
As with exhaust system corrosion, we had little information
about the form of the relationship between low lead concentrations
and spark plug fouling. As we discussed in section III.A.4 of
this chapter, two citations indicated that all spark plug savings
are likely to be achieved by lowering lead concentrations to 0.5
g/gal (from the studies' beginning point of 2.3 g/gal). For the
purposes of this analysis, further savings would be gained from
less spark plug foulings by going from current levels of 1.1 g/gal
only to 0.5 g/gal; no further savings would be achieved by reduc-
ing from 0.5 g/gal to zero lead.* Thus, given the state of our
current knowledge, total savings would be identical for misfuelers
and other leaded light-duty vehicles in both options under con-
sideration. (But the uncertainty surrounding the correct threshold
for additional engine fouling is substantial, and affected our
estimates for both the low-lead and all unleaded cases.) As
earlier, we estimated that spark plugs would last 20% longer
under either policy, resulting in .35 cents/gallon savings, or
about $79 million for both the low-lead and all unleaded cases
in 1988.
* As previously noted in section III.A.4 the threshold could
be an average of the two available studies (Toyota and
Union Oil), in which case the threshold would be 0.35
g/gal with 23% of the savings available in going from
1.1 to .35 g/gal versus the 20% value used in going from
1.1 to 0.5 g/gal. We have thus understated benefits by
using a higher threshold.
-------
III.30
III.B.3. Oil Change Savings
Our discussion of the savings to be achieved from fewer
required oil changes was made in Section III.A.5. Principally,
we relied on the work of Pless, who found decreased engine wear
with unleaded gasoline and doubled oil change intervals compared
with engines using gasoline containing 2.70 grams of lead per
gallon and a standard oil change interval. We also relied on
Gallopoulos and Cordera, who described the relationship between
engine rust and lead additive variations. To interpolate oil
change savings to our low-lead case of 0.1 g/gal, we used the
same methodology as in the no lead case. In going from 1.1 to
0.1 grams of lead per gallon, 58.3% of the benefits are achieved.
For legal leaded gasoline users this becomes a savings of
$107 million. Since we assume that no misfueling would occur
under the low-lead option, the savings achieved by misfuelers
under this option are the same as the all unleaded option, or
$132 million. Total savings are the savings for the legal leaded
gasoline user plus the misfuelers1 savings: $239 million.*
111.B . 4. Sum of Maintenance Savings for the Low-Lead Case
As calculated in the previous three sections, we estimated
$339 million savings from decreased exhaust system replacements,
$79 million savings for less spark plug fouling, and $200 million
savings with fewer required oil changes. In total, lowering lead
* Calculation: Legal leaded users: 1.47^/gallon times .583 =
.857jzf/gal savings. Thus, ,857^/gal times 12.49 billion
gallons of legal leaded use equals $107 million. Misfuelers:
1.47£/gal times .873 = 1.283jZ
-------
III.31
concentrations to 0.1 g/gal would yield about $618 million in
vehicle maintenance benefits.
III.C. Risk of Valve Recession
Balancing these savings is the fact that reducing the amount
of lead in gasoline may increase wear on some engines requiring
the lubrication that lead compounds now provide. In particular,
some studies argued that severe exhaust valve recession could
occur, resulting in leaking valves, loss of compression pressure,
greatly increased hydrocarbon emissions, degraded vehicle perfor-
mance, and reduced fuel economy. Reviewing the available data,
it appeared that eliminating leaded gasoline could mean exhaust
valve recession in some light-duty vehicles and other engines
that were originally designed to run on leaded gasoline.
The following paragraphs describe:
0 the process of valve recession,
0 the conditions under which it is likely to occur,
0 the concentration of lead in gasoline needed to
prevent damage, and
0 alternative mechanisms that might provide
sufficient protection.
We also estimated the types and numbers of engines that might
be at risk without leaded gasoline.
Exhaust valve recession appeared to result from both
abrasion and adhesion on the valve seat when engines operated
under high temperatures, loads, or engine speeds. (For detailed
discussions of the mechanisms of valve wear, see Godfrey and
Courtney, 1971; Giles, 1971; or Kent and Finnegan, 1971.)
-------
III.32
Several researchers have examined rates of valve recession
as a function of engine operating variables and the amount of lead
in the fuel. Giles, and Godfrey and Courtney were consistent in
finding that recession rates appeared to be mostly a function of
engine speeds. Giles, for example, found that valve recession
increased almost linearly with higher engine speeds to a point
(on the engine he tested, about 3700 rpm), and then rose as an
exponential function of engine speed. The shape of this function
apparently varied significantly by vehicle models and years.
We reviewed two types of research about the causes and rates
of valve recession. The first type of study was engine tests on
dynamometers, done either using unusually high engine loads to
test valve durability, or using 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. An advantage of
engine tests was their greater measurement precision and control
over test conditions. The vehicle studies, on the other hand,
may have been more likely to reflect "real world" effects.
Table III-2 summarizes the available studies of valve
recession as a function of lead concentrations. It should be
noted 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 two described
later, led them to conclude that:
-------
III.33
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 a vehicle traveling at
60 mph than vehicles traveling at 70 mph, despite going 22% to
280% more miles. The following conditions were used in engine
studies finding serious valve wear with unleaded gasoline:
0 Giles (1971) conducted tests with passenger car engines
under varied conditions from steady-state wide-open
throttle (WOT) to simulated road-load cycles. He found
that the valve recession rates were about ten times
higher without lead than with 2 to 3 grams of lead per
gallon of gasoline. But since he does not report the
magnitudes of recession, it was impossible to tell how
serious the recession was under the various conditions.
(Valve recession occurred at a slight rate even with
lead additives.)
0 Giles and Updike (1971) ran one engine for 50 hours at
a steady-rate of 3500 rpm WOT (speed selected to maxi-
mize valve recession rates while minimizing the other
engine durability problems of other components at higher
engine speeds); another engine ran for 50 hours at 2600
rpm WOT. Finding: about three times the rate of reces-
sion with unleaded gasoline vs. leaded.
0 Kent and Finnegan (1971) also found severe valve reces-
sion in tests simulating a 1970 V-8 pick-up truck
hauling a camper at freeway speeds of approximately
65-70 mph, with some engine cycling, for a total
running time of 80 hours. In contrast, they found very
low exhaust valve wear when running engines at 2300 and
2400 rpm.
-------
III.34
0 Godfrey and Courtney (1971) found somewhat lower rates
of recession than did other high load engine studies
when they tested an engine at 4400 rpm WOT for 10 1/4
hours. They also found recessions of unreported magni-
tudes on six other engines running 9,000 to 11,000 miles
at 70 mph under conditions designed to generate artifi-
cially high temperatures.
0 Felt and Kerley (1971) used both dynamometer and road
tests, mostly testing at 70 mph freeway schedules, and
some on a cycled route of combined city and freeway
driving. They found 2/3 less valve wear at 60 mph than
at 70 mph.
0 Pahnke and Bettoney (1971) found serious valve reces-
sion in three unleaded dynamometer tests after the
equivalent of 8,000 miles at a steady speed of 70 mph.
All these studies were designed either to investigate the mech-
anisms 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 valve
recession under normal driving conditions.
Overall, it seemed that using unleaded gasoline exclusively
in vehicles requiring lead's lubrication may risk premature valve
failure under severe engine loads. These studies indicated that
such severe recession is most likely to occur in vehicles travel-
ing at high loads or speeds well above the legal speed limit of
55 mph for extended periods of time (tens or hundreds of hours,
or for thousands of miles). Several related studies using fleets
of drivers under more typical conditions found little or no
incidence of valve problems with unleaded gasoline (Pahnke and
Conte, 1969; Orrin et al., 1972; Gray and Azhari, 1971). Two
studies cited more valve problems for unleaded than for leaded
vehicles (Wintringham et al., 1971; Felt and Kerley, 1971).
-------
III.35
TABLE III-2
SUMMARY OF FINDINGS: VALVE RECESSION AT VARIED LEAD CONCENTRATIONS
Paper
Pahnke & Conte, 1969
Gray & Azhari, 1971
Pless, 1974
Orrin et al., 1972
Giles, 1971
Giles & Updike, 1971
Duelling, 1971
Kent & Finnegan, 1971
Pahnke & Bettoney, 1971
Fuchs, 1971
Felt & Kerley, 1971
Godfrey & Courtney, 1971
Grouse et al., 1971
Test Type
Employee fleet,
Personal use
a. Employee fleet
b. Consumer survey
Taxi fleets
Taxi fleets
Varied engine loads
Varied high loads
Engine tests
High load
a. Consumer survey
g Pb/gal
2.8,0.1,0
2.3,0
2.3,0
0
2.8,0
2.5,<0.03
0
0.14,0.07,
0.04,0
3.0,0.5,0.2,0
2.3,0
Findings
No extra valve problems with unleaded
a.
b.
No severe valve problems, but some valve
stem wear in one fleet with unleaded
No extra valve problems with unleaded
Need no re than 0.03 g/gal
Recession rate inc. above 70 nph.
Avg. driving should not pose problem.
Between 0.04 and 0.07g/gal is adequate
0.20 g/gal is adequate
a. No clear difference, but somewhat
b. High load, engine 0.5,0
tests
Engine tests 0.5,0
a. Employee fleet 0.5,0
b. High load & cycled 0
High load engine tests 0
a. Patrol fleet, very 3.1,0
severe service
b. 50K mile road test 2.6,0
(.008 g/gal)
more valve problems with unleaded
b. Severe recession after 8000 miles
on unleaded; none at 0.5 g/gal.
0.5 g/gal virtually eliminates recession
a. More valve problems with unleaded
b. Recession rates accelerate with
high speeds; 0.5 g/gal is adequate
High loads and speeds are
major causes of recession
a. Recession after 10-15K miles or
more in severe service
b. In matched pairs directionally
less tip wear on unleaded; severe
recession in one unmatched engine
Wintringham et al., 1972
Employee fleets
2.3,0
More expensive valve problems with unleaded
-------
III.36
Other fleet studies were inconclusive concerning the relative
incidence of valve problems for unleaded vehicles (Pahnke and
Bettoney, 1971; Grouse et al., 1971; Pless, 1974). Finally,
reported incidents of valve problems were rare among users of
unleaded gasoline in the late 1960s (Wintringham et al., 1972).
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 (from 0.001 inch per hour (iph) initially,
stabilizing at less than 0.0001 iph after 25 hours). Giles then
showed that, after switching the engine to unleaded gasoline,
recession rates continued to be low until the lead deposits wore
away (after about 10 hours). Recession then rose again to high
rates. In sum, it may take 10-25 hours for lead deposits to
build up sufficiently to mitigate valve wear; if leaded fuel is
not used, these deposits will wear off in several hours (about
10), leaving the exhaust valve vulnerable again to wear.
III.C.1. How Much Lead is Required to Protect valves
A critical question is: "How much lead or similar additive
in gasoline is necessary to protect against severe valve reces-
sion?" Most studies were performed with the high lead concen-
-------
III.37
trations in gasoline that were common in the late 1960s — about
2.3 grains of lead per gallon of gasoline. Also present in that
gasoline are traces of sulfur, which occurs naturally in petroleum,
and small quantities of phosphorus, which is added with lead to
modify the deposits in the cylinder.
Several studies concluded that 0.5 grams of lead per gallon
of gasoline was a sufficient concentration to protect against
valve recession (Kent and Finnegan, 1971; Pahnke and Bettoney,
1971; Felt and Kerley, 1971; Fuchs, 1971). Kent and Finnegan
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
concentrations to discover how little lead was necessary to
eliminate valve recession. Doelling (1971) conducted tests at
2650 rpm at lead levels of 0.04, 0.07, and 0.14 g/gal, for 100
hours each. Focusing on maximum recession of any of the valves,
Doelling found no recession at 0.07 or 0.14 g/gal, but found
excess wear at 0.04 g/gal. He thus concluded that leaded gaso-
line would protect exhaust valves beginning at levels between
0.04 and 0.07 grams of lead per gallon.
III.C.2. Alternatives to Lead to Avoid Valve Recession
Other mechanisms besides lead in gasoline can mitigate
significant valve recession. Among these, improved metallurgy of
the valves and phosphorus additives to gasoline are of greatest
interest for this analysis.
-------
III.38
Since 1971, the automobile industry has used induction
hardened valve seats, seat inserts, or chrome or nickel plating
in light-duty vehicles to mitigate valve recession without lead.
Essentially, these technologies either stop the oxidation of the
iron valve seat or harden the valve seat (or face) to protect
against the adhesive and abrasive processes. By 1971, General
Motors Corporation was using this improved metallurgy on all its
light-duty vehicles to compensate for the use of unleaded gaso-
line. Ford made these improvements on most of its light-duty
vehicles by 1971 as well. After that date, other manufacturers
also implemented these changes. By the 1975 model year, virtually
all cars were "clear fuel tolerant," although the changes in light
trucks may have been slower. The widespread use of these improve-
ments since the early 1970s has greatly limited the number of
vehicles that might be at risk of valve damage due to the unavail-
ability of leaded gasoline.
In addition, other substances could feasibly provide vehicles
with the lubrication they now receive from lead in gasoline. Most
plausible among the alternatives are phosphorous compounds which
are already added to gasoline along with lead (i.e., the alterna-
tive technology is already in use and has been found to be effec-
tive in reducing valve wear).
Several experiments suggested that phosphorus in unleaded
gasoline could reduce or eliminate the threat of valve recession
at high speeds. Specifically, at about 0.06 or 0.07 g phosphorous/
gallon, valve wear proceeds at one-half to one-third the rate
-------
III.39
occurring with no additives (Giles and Updike, 1971; Kent and
Finnegan, 1971; Felt and Kerley, 1971; and Wagner, 1971). 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 reces-
sion for nearly 90% of the cars tested. The phosphorus more than
halved the rates of recession for the cars that, without lead or
phosphorus, 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) showed that neither altering the scavenger
mix nor eliminating scavengers from the fuel curtailed exhaust
valve life in the engines. They found the presence of phosphorus
in the gasoline was critical to exhaust valve life durability.
All of these results indicated that the addition of phosphorus to
unleaded gasoline would substantially reduce the risk of valve
recession for those vehicles at risk.
In addition to gasoline engines in light-duty vehicles,
three other categories of engines might require leaded gasoline.
* Giles (1971) said that the limit of tolerable recession
was about 0.125 to 0.150 inches of change, at which point
the hydraulic valve lifters had problems operating.
-------
III.40
The first of these is a variety of classes of small engines:
lawnmowers, snowmobiles, snowblowers, garden tillers, and others
small equipment. We asked representatives of the three major
manufacturers of small engines in the United States (Briggs,
Tecumseh, and Kohler) what kind of gasoline they recommend for
these engines. They said that their engines almost always could
use either leaded or unleaded gasoline, or they should use unleaded
gasoline. These representatives also believed that this had been
true for their engines for at least a decade (and knew of no
changes that would make this untrue for earlier engines). Impor-
tantly, the reason they cited for preferring unleaded gasoline for
this equipment was that leaded gasoline caused harmful deposits
and corrosion in the engines.
Second, we investigated the possibility that marine engines
required leaded gasoline, but the responses from manufacturers
of boat engines were not clear-cut. Some boat engines may require
the lubrication they now get from lead, while others are supposed
to use clear fuel. Two-stroke engines should not be affected by
using unleaded gasoline. One complicating issue was the relative
octanes of leaded and unleaded gasolines. With the limited
information we had, it appeared that some, but not all, marine
engines would require leaded gasoline.
Third, gasoline-powered heavy-duty trucks are designed to
use leaded gasoline. (Such trucks account for roughly 4% of all
gasoline demand.) Because heavy-duty trucks are more likely than
passenger cars to travel under heavy loads for long durations,
-------
III.41
this category may carry the highest risk of premature valve
failure if fueled with only unleaded gasoline. However, the
extra magnitude of risk was difficult to assess because we did
not know what fraction of these trucks used a high portion of
their potential power. Giles (1971) wrote that
Heavy duty engines, however, usually have valve seat
inserts, rotators, and heavy duty valves already
included in the design package.... Valve face reces-
sion and seat wear both are observed with heavy.duty
engines running on leaded fuels today. Wear rates are
low, however, and recession is noticed only because of
the extended operating life of these engines. Some
increase in wear rates might occur when these engines
are switched to lead-free gasoline, but catastrophic
wear is not expected. Limited dynamometer testing
does indicate that wear is not increased significantly
but each engine design and application should be
weighed separately. (p.1483)
Nonetheless, manufacturers reportedly would recommend against
allowing heavy-duty trucks to operate solely on unleaded gasoline.
As a result, heavy trucks may be the single category significantly
affected if leaded gasoline were not available.
We have estimated approximately how many engines might be
"at risk" of severe valve recession if leaded gasoline were not
available.* Using the method and assumptions described in detail
in the Technical Appendix to Chapter IV, we estimated that about
2.2 million light-duty vehicles without improved valves would exist
* For most of these vehicles at risk, the probability of severe
valve recession due to lack of lubrication from lead appears
to be well below 10% in any year. Because of limited data,
we were unable to quantify this probability with any greater
precision. The probability for any individual vehicle will
depend very much on its particular design and the ways in
which it is operated.
-------
III.42
in 1988. An additional 12.2 million cars (model years 1971-80)
were designed for leaded gasoline and some of these may be at
risk. However, GM has indicated that since 1971 all of its cars
used the improved valves. Because roughly 50% of the market
was GM vehicles, we have reduced this number by half, to 6.1
million cars. Ford also used improved valves and we have reduced
the value by Ford's market share of roughly 20%. This final
value (3.7 million cars) is most likely too high also because
other manufacturers probably used improved valves as well.
Table III-3 shows a more disaggregated forecast and presents
projections for the full range of engines. These 25.5 million
"at risk" vehicles represent about 13.5% of the 188 million
projected total fleet of highway vehicles.*
III.D. Summary
This chapter has presented national estimates of the vehicle
maintenance benefits of a reduction of lead in gasoline. Two
scenarios have been examined: a reduction of lead in gasoline
from the current 1.10 grams/gallon to 0.10 grams/gallon, as well
as the total elimination of leaded gasoline. These estimates
are based on projections of gasoline demand and vehicle fleet
characteristics in 1988, and are valued in 1983 dollars.
Three sources of vehicle maintenance benefits have been
tabulated and are presented in Table III-4. Both policy options
* Equal to 1.17 times the projected light-duty truck and light-
duty vehicle fleet of 160 million, using Bureau of Census
1977 proportions.
-------
III.43
Table III-3
NUMBER OF ENGINES AT RISK OF SEVERE VALVE
RECESSION WITHOUT LEADED GASOLINE
Thousands of Vehicles in 1988
Type of Vehicle
Cars Pre-'Vl
LDTt Pre-171
Cars 1971-1980
LDTt 1971-1975
Heavy-Duty Trucks
Boats
TOTAL "AT RISK" GROUP:
High Risk*
2,240
1,146
10,865**
Not calculated
14,251
Low Risk*
3,700
7,505
Not calculated
11,205
TOTAL ALL VEHICLES: 25,456,000
* The "high risk" group represents heavy-duty and those
light-duty vehicles manufactured before 1971. While
many vehicles manufactured between 1971-1980 were
built to use leaded gasoline, most of these have
newer more durable valve and valve seat materials and
and thus form a "low risk" group.
** This is equal to 30% of our projection of light trucks
in 1988 — the same proportion as was found in the Bureau
of the Census publication 1977 Census of Transportation
(1980).
t Light-duty trucks.
-------
III.44
are expected to result in savings in decreased exhaust sys.tem
replacements, longer life for spark plugs, and increased time
intervals between oil changes. The point estimate of: vehicle
maintenance benefits for the low-lead option is $618 million,
while the no lead scenario yields an estimate of $741 million
in benefits. Several other sources of potential vehicle
maintenance benefits have also been discussed in this chapter,
but no monetary estimate of their magnitude has been attempted.
Chapter III also presents estimates of the number and types
of vehicles expected to be at increased risk of severe valve
recession if leaded gasoline is completely eliminated. As shown
in Table III-3, at most 25 million vehicles are potentially at
increased risk of damage in 1988. However, monetary values
for these damages have not been computed due to considerable
uncertainty as to the number of vehicles likely to experience
damage.
-------
III.45
TABLE II1-4
SUMMARY OF MAINTENANCE SAVINGS
(1983 dollars, in millions)
Billions of gallons of
leaded gasoline demand in 1988
ALL UNLEADED SAVINGS;
Exhaust Systems @ $1.68£/gal
Spark Plugs @ 0.35/d/gal
Oil Changes @ 1.47jz*/gal
LDVs* Designed
Misfuelers to Use Leaded Total
10.29 12.49 22.78
$173
$ 36
$132
$210
$ 44
$160
$383
$ 80
$292
LOW - LEAD SAVINGS;
Exhaust Systems
Spark Plugs
Oil Changes
$173!
$ 36l
$132
TOTAL:
$1682
$ 443
$107
$755 Million
$341
$ 80
$239
TOTAL;
$660 Million
1. Assumes that changes in price differentials, marketing strategies,
and possible administrative controls on distribution of leaded
gasoline would effectively eliminate misfueling in the 0.1 g/gal
case.
2. Assumed threshold to achieve any savings is at 0.5 g/gal.
3. All savings achieved by 0.5 g/gal.
Light-duty vehicles.
-------
III.46
REFERENCES
CHAPTER in
Adams, W. E. (Ethyl Corporation) "Discussion of SAE Paper
#720084," Detroit, January, 1972.
American Automobile Association (Potomac Division) "What's
Wrong with the Average Washington Car," (News Release).
Falls Church, VA., January 3, 1984.
Bettoney, W. E. (DuPont de Nemours and Co.) "Discussion
of SAE Paper #720084," Detroit, January, 1972.
Canadian Energy and Emissions Committee, Comments on "Control
Options For Lead Phase Down in Motor Gasoline," Environment
Canada Report, International Lead ZMC Research Organization,
May 1983.
Champion Spark Plug Co., Champion Ignition and Engine
Performance Conferences, volumes 1971-1976. ~~
Committee on Motor Vehicle Emissions, Report on the Development
of a Long-Term National Motor Vehicle Emission Strategy,
Australian Transport Advisory Council, 1972.
Coordinating Research Council, 1982 CRC Octane Numbers
Requirement Survey, July 1983.
Cordera, F. J., et al. (Shell Oil Company) "TEL Scavengers in
Fuel Affect Engine Performance and Durability," SAE Paper
#877A, June 1964.
Doelling, R. P. (Cities Service Oil Company) "An Engine's
Definition of Unleaded Gasoline," SAE Paper #710841.
DuPont, Petroleum Laboratory, "Exhaust Catalysts for Leaded
Fuel - A Progress Summary," PLMR-42-81, submitted to EPA
May 1982.
Exxon Memo, "Re: Gulf/East Coast Gasoline," January, 1978.
Felt, A. E., and Kerley, R. V. (Ethyl Corporation) "Engines
and Effects of Lead-Free Gasoline," SAE Paper #710367,
October 1970.
Gallopoulos, N. E. (General Motors Corporation) "Projected
Lubricant Requirements of Engines Operating with Lead-Free
Gasoline," SAE Paper #710585.
-------
III.47
Gergel, W. C., and Sheahan, T. J. (Lubrizol Corporation)
"Maximizing Petroleum Utilization Through Extension of
Passenger Car Oil Drain Periods - What's Required?," SAE
Paper #760560, 1976.
Giles, W. S. (TRW Incorporated) "Valve Problems with Lead Free
Gasoline," SAE Paper #710368.
Godfrey, D., and Courtney, R. L. (Chevron Research Company)
"Investigation of the Mechanism of Exhaust Valve Seat Wear in
Engines Run on Unleaded Gasoline," SAE Paper #710356, 1971.
Gray, D. S., and Azhari, A. G. (American Oil Company) "Saving
Maintenance Dollars with Lead-Free Gasoline," SAE Paper #720084,
January 1972.
Hickling, J. F., Analysis of Lead Phase-Down Control Options,
Management Consultants Ltd., October 1983.
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,
March 1981.
Kent, W. R., and Cook, W. A. (Union Oil Company) "The Effects of
Some Fuel and Operating Parameters on Exhaust Valve Seat Wear,"
SAE Paper #710673, 1971.
Motor Vehicle Manufacturing Association, Motor Vehicle Facts and
Figures "83". Detroit, 1983.
Motor Vehicle Manufacturing Association, "Incentives for Proper
Usage of Unleaded Fuel", memo to EPA, January, 1984.
Pahnke, A. J., and Bettoney, W. E. (DuPont de Nemours and Co.)
"Role of Lead Antiknocks in Modern Gasoline," SAE Paper #710842,
1971.
Pahnke, A. J., and Conte, J. E. (DuPont de Nemours and Co.)
"Effects of Combustion Chamber Deposits and Driving Conditions
on Vehicles' Exhaust Emissions," SAE Paper #690017, 1969.
Pless, L. G. (General Motors Corporation) "Interactions Among
Oil Parameters Affecting Engine Deposits and Wear," SAE Paper
#720686, 1972.
Pless, L. G. (General Motors Corporation) "A Study of Lengthened
Engine Oil-Change Intervals," SAE Paper #740139, 1974.
-------
III.48
Schwochert, H. W. (General Motors Corporation) "Performance of
a Catalytic Converter on Non-leaded Fuel," SAE Paper #690503,
May 1969.
U.S. Environmental Protection Agency, Motor Vehicle Tampering
Survey - 1982. National Enforcement Investigations Center,
Office of Enforcement, April 1983.
Wintringham et al. (Ethyl Corporation) "Car Maintenance Expense
in Owner Service with Leaded and Non-leaded Gasolines," SAE
Paper #720499, May 1972.
-------
CHAPTER IV
BENEFITS OF AVOIDING EXCESS HC, NOX AND CO EMISSIONS
This chapter discusses the effects of increased emissions
from poisoned catalysts of vehicles misfueled with leaded gaso-
line. We estimated the excess emissions of hydrocarbons (HC),
nitrogen oxides (NOX), and carbon monoxide (CO) that we could
avoid by eliminating misfueling by 1988. We then examined three
alternative methods to value avoiding this air pollution. We
synthesized this information to generate a best estimate of the
economic benefits of reducing misfueling.
Both the all unleaded and low-lead policy options are
assumed to eliminate misfueling and its consequent excess
emissions. "Misfueling" or "fuel switching" refers to the use
of leaded gasoline in a vehicle originally designed and certi-
fied to use unleaded gasoline. Because leaded regular gasoline
is cheaper and higher in octane than regular unleaded gasoline,
some drivers deliberately misfuel their vehicles in an attempt
to reduce expenses or to improve vehicle performance. Our
low-lead option assumes a low-lead fuel (0.10 grams of lead per
gallon of gasoline) for the few classes of vehicles that may
require the valve lubrication lead provides, but with avail-
ability restrictions designed to eliminate misfueling as a
practical problem.
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
-------
IV.2
sell gasoline that is mislabeled or contaminated (U.S. EPA,
1983c), but this accounts for less than 1% of misfueling.
It is illegal for service stations or commercial fleet
owners to misfuel or to allow misfueling of vehicles originally
equipped with catalytic converters. However, federal law does
not apply to individuals who misfuel their own vehicles. This
limitation hurts EPA's ability to curb this harmful practice.
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 IV-1
shows the emissions increases. The excess HC and NOX emissions
also increase ozone concentrations.
TABLE IV-1
Increase in Emissions Due to Misfueling (grams/mile)
Light-Duty Vehicle Model Years HC CO NOX
1975 - 80 2.67 17.85
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.
According to a 1982 survey by EPA (U.S. EPA, 1983c) the
current misfueling rate of light-duty vehicles designed to use
unleaded gasoline is about 12%.* We assumed for our analysis that
*The unweighted average the survey found was 10.5%. Weighting
the results by the portions of the light-duty fleet in areas
with and without Inspection and Maintenance (I/M) programs, by
the fractions of light-duty vehicles of each age (in 1982), and
by the number of vehicle miles traveled by each model year, the
weighted average is 12.2%. (About 17%-18% of the light-duty
fleet was in I/M areas.) We believe this estimate may be an
underestimate of actual misfueling rates for several reasons
discussed later in this and other chapters.
-------
IV.3
this rate would stay constant to 1988. If this rate rises over
time, as preliminary data from the 1983 survey imply, our emission
estimates may be too low.
Misfueling rates apparently vary by the age of vehicle, by
whether the vehicles are in localities that have Inspection and
Maintenance (I/M) mobile source enforcement programs, by whether
they are part of a commercial fleet, and other factors. Table IV-2
provides 1982 misfueling rates by model year of vehicle and by I/M
status.
TABLE IV-2
1982 Misfueling Rates by Age of Vehicle and by I/M Status
Model
Year
1982
1981
1980
1979
1978
1977
1976
1975
Overall
Misfueling Rates
5.2%
7.4%
8.1%
12.1%
12.2%
12.4%
14.5%
17.7%
I/M Areas
4.4%
4.3%
5.7%
4.9%
5.9%
9.9%
9.6%
6.3%
Non-I/M Areas
6.3%
9.6%
10.1%
20.3%
19.5%
16.5%
20.2%
30.9%
Weighted Average:* 13.5% 6.2% 15.1%
*This weighted average does not take into account the number of
miles driven by each model year.
The EPA surveys probably underestimated real misfueling rates
by a significant margin. One of the main reasons for this was that
vehicle inspections for misfueling were voluntary, which would bias
the results downward. In some areas, the rates of drivers refusing
inspections were very high. The refusal rates ranged from less
than 1% to 8% in I/M areas, and from 3% to 44% in non-I/M areas.
-------
IV.4
Imperfect indicators of misfueling also provided a possible
downward bias in these rates. EPA used three tests to check for
misfueling: whether the fuel inlet restrictor was removed or
damaged, whether the gasoline in the tank had more than 0.05 grams
of lead per gallon, and whether traces of lead could be found
inside the tailpipe (the plumbtesmo test). Each of these three
tests is likely to miss a substantial portion of misfuelers, and
the plumbtesmo test had a high rate of false negative findings
when administered hastily in field tests. By using the three
indicators together, EPA tried to minimize the likelihood of
missing catalysts damaged by misfueling. The results suggested
that excess emissions from misfueling in 1983 were significant.
As explained in detail later in this chapter, misfueling accounted
for roughly 2.48% of all HC emissions, 5.18% of all CO emissions,
and 0.78% of all NOX emissions, nationally.
In economic analysis, because 1988 dollars are not equal to
1983 dollars, future costs are discounted to arrive at a "present
value." To make the benefits of avoiding excess emissions compar-
able to the estimate of costs that we presented in Chapter II, we
discounted our emission figures by 3% (the standard rate used for
long-term analysis). Table IV-3 shows the 1988 estimates of the
discounted stream of future emissions avoided by implementing
either of the policy options this paper is examining. The
emissions estimates are from all cohorts of vehicles that would
misfuel in 1988 in the absence of any change in policies or
practices.
-------
IV. 5
TABLE IV-3
Discounted Future Emissions Avoided
by Eliminating Misfueling in 1988
(thousands of metric tons)
EC CO NOX TOTAL
314 2,202 130 2,646
We estimated the tons of excess emissions that would be avoided in
1988 if EPA were to eliminate misfueling for all light-duty vehicles
in 1988 (under either the low-lead or all unleaded option). We did
not consider emissions that would occur in 1988 from misfuelings in
previous years. Since the costs of eliminating misfueling are
calculated for one year (1988), the benefits should include only the
avoided emissions attributable to eliminating misfueling in that
year, and none from other years. The technical appendix to this
chapter provides a description of the fleet model and the discount-
ing procedure for avoided emissions beyond 1988, and a discussion
of uncertainties that may have biased our results.
There is no consensus on a good, simple way to value the bene-
fits of eliminating habitual misfueling and its consequent excess
emissions. As a result we have used three different approaches:
A. the value by the costs of alternative regulations;
B. the value of preserving catalytic converters; and
C. the value of avoiding damage to health, vegetation,
and materials.
Table IV-8, which appears at the end of this chapter, summarizes
the values we derived by each of these three methods. In the next
three sections, we present our calculations by each method in more
detail.
-------
IV.6
IV.A. Value by the Costs of "Next-Step" Regulations
Our first method computes the value of avoiding misfueling
emissions by using the cost per ton of other HC, CO, and NOX
regulations that the Agency is considering promulgating (Table
IV-4). The rationale for this method is that these "next-step"
regulations reveal the low end of the range of values that EPA
or Congress imputes for controlling further increments of these
pollutants. (This does not imply that EPA would not promulgate
such regulations if we were to adopt either the low-lead or
all unleaded option.)
To value these emissions, we chose future regulatory options
from among the least costly alternatives that could potentially
control a similar amount of each pollutant. However, EPA has
imposed much more costly regulations for these pollutants in the
past. In many of the more expensive cases, the cost per ton of
pollutants abated was not a good measure of what the Agency or
Congress was valuing with that particular regulation. For
instance, Congress frequently required EPA to choose the best
technology available -- sometimes without regard to costs. In
other cases, controlling certain sources was considered more
valuable than abating an apparently similar quantity from other
sources. This might have been because potential exposures to
some sources were greater, or because the particular pollutants
may have different toxicities. Thus, "cost per ton" may be a
very crude measure of cost-effectiveness or the social value of
controlling pollution.
-------
IV.7
Table IV-4 shows our estimates of the present value of
emissions avoided by eliminating misfueling using the costs per
metric ton of alternative regulations. The total benefit from
avoiding these pollutants would be $121-452 million in 1988.
TABLE IV-4
Benefits Valued by "Next-Step" EPA Regulations
(1983 dollars)
HC CO NOX Total
Total Value
of Emissions
$/metric ton
$81-217M
$232-618*
$27-140M
$11-57**
$13-95M
$92-660t
$121-452M
N/A
* Cost of $232-$618/ton of HC removed by Stage II vapor recovery
from gasoline marketing. Estimates were from Pacific
Environmental Services, Inc., Update of the Gasoline Marketing
Emissions Data Base, September 1983 (4th Quarter 1982 dollars).
The "next-step" at petroleum refineries - another major source
of HC - would be secondary seals on gasoline storage tanks, at
$882 per ton of HC. This is from SCI, Impacts of Revising EPA
Regulations Relating to Petroleum Refining and Petrochemical
Production, June 1983. Using this cost per ton would value
avoiding excess HC emissions from misfueling at $217 million
in 1988 (1983 dollars).
** Cost of $ll-$57/ton of CO removed by engine modification,
catalysts, and inspection and maintenance on heavy trucks.
Estimates were from Chapter 3, Regulatory Impact Analysis of
the HC and CO Standards for Heavy Duty Trucks, U.S. EPA,
forthcoming.
t Cost of $92-660/ton of NOX removed by low excess air and
staged combustion at utility and industrial coal boilers.
Estimates for utility boilers were $84-251/short ton, from
ICF, Inc. , Analysis of a 10 Million Ton Reduction in Emissions
from Existing Utility Powerplants, June 1982; industrial
boiler estimates were $320-$600/short ton, calculated using
emission factors from AP-42 and from the draft U.S. EPA,
Background Information Document for the Industrial Boiler
NSPS (1979), while costs came from Costs of Sulfur Dioxide,
Particulate Matter, and Nitrogen Oxide Controls on Fossil
Fuel Fired Industrial Boilers, EPA-450/3-82-021, August 1982.
-------
IV.8
IV.B. The Value of Preserving Pollution Control Equipment
Our second method of valuation used the cost per ton of
emissions control by catalytic converters and other equipment
disabled by misfueling. To estimate the benefits of eliminating
misfueling, this method used EPA's implicit balancing of costs
and benefits in selecting catalytic converters as the method for
emissions control on mobile sources. We assumed that, this cost
per ton reflected the value that EPA and society placed on
reducing these pollutants. In addition, this method of valuation
most nearly approximated the loss of catalytic converters poisoned
by misfueling. Each year, consumers purchase roughly 9.7 million
catalytic converters on their new light-duty vehicles; over 12%
of these are subsequently disabled by misfueling with leaded
gasoline.
We estimated a cost of $283 per car for emission control
equipment (U.S. EPA, 1981).* (About half of the cost of oxygen
sensors and other equipment was allocated to fuel efficiency,
not pollution control.) We counted the emissions controlled by
that equipment over an average ten-year car life, beginning with
90% control efficiency in the first year, and leveling off by the
fifth year to 50% efficiency. (EPA regulations require that
manufacturers provide warranties on catalytic converters for only
* The costs are "retail price equivalents," which are 30% to 50%
of the manufacturers' suggested retail price of the components
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.
-------
IV.9
five years, but EPA data indicated that this equipment can be
effective for the life of the vehicle (Faucett, 1983).) We may
have understated the rate at which catalytic converter effi-
ciencies deteriorate at low mileage. If so, our estimate would
overstate the tons controlled and underestimate the value of
avoiding emissions by eliminating misfueling.
We projected the tons of HC, CO, and NOX controlled in each
year of a ten-year catalytic converter life. We then discounted
the future emissions at a 3% rate (as when calculating excess
emissions and costs) to the first year, when the equipment costs
would be incurred. This produced a cost per ton of $163 for
avoiding HC, CO, and NOX emissions.* We multiplied this by the
2.65 million tons of excess emissions avoided (from Table IV-3).
This gave us a benefit estimate of $432 million for eliminating
misfueling under either the low-lead or all unleaded policy
options.
* Our calculations used 1981 emissions standards of 0.41 grams
per mile for HC, 3.4 g/mi for CO, and 1.0 g/mi for NOX. This
totaled 4.81 g/mi for all pollutants in each future year.
We divided 4.81 g/mi by (1 - catalytic converter efficiency
in that year), multiplied by 10,000 mi/yr, and divided by
1000 b/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 1.733 tons controlled
by each car's catalytic converter over a ten-year life.
A cost of control equipment of $283, divided by 1.733 tons
controlled, equalled a cost per ton of $163 in 1983 dollars.
-------
IV.10
IV.C. Benefits Estimated Directly from Health and Welfare
Improvements
Research has suggested that HC, NOX, and CO emissions may
directly affect human health and welfare. Our third method of
valuing the adverse effects of misfueling provided direct
estimates of the health, vegetation, material, and ecosystem
benefits that would result from reducing these emissions. In
addition, since HC and NOX are precursors of ozone, we estimated
the benefits that a reduction in ozone would have on agricultural
yield, materials damage, and health.
Unfortunately, because of scientific uncertainty, lack of
data or quantitative estimates, or inability to value certain
effects monetarily, we have presented only a partial calculation
of the total benefits of reducing misfueling. For example, we
did not calculate the effects of ozone on ecosystems or the
effects of chronic exposure of ozone on health, nor did we
include a quantitative estimate of health benefits from reducing
CO emissions. When possible, however, we have described these
effects qualitatively.
A few of the studies used in this analysis are EPA contractor
reports in progress or in completed draft. As such, they have not
undergone full peer review and should be considered preliminary.
IV.C.I. Benefits of Reducing Ozone
To calculate the benefits of reducing ozone, we first had to
determine the relationship between reducing HC and NOX and the
subsequent decrease in ozone. Once we determined the reduction
-------
IV.11
in ozone, we used both dose-response (bottom-up) approaches and
proportionate share (top-down) approaches to determine the corre-
sponding amount of economic benefits. In the bottom-up approach,
we used disaggregated damage functions to estimate the impact of
a given change in ambient levels. In the top-down approach, we
interpolated aggregate damage numbers to obtain the impacts of a
single pollutant or of a given change in ambient levels. Regard-
less of the approach, the benefit estimates contain a good deal
of uncertainty and should be interpreted with caution.
We needed two general simplifying assumptions to use the
top-down technique. First, unless noted otherwise, we assumed a
constant elasticity between pollution reduction and the economic
effect of concern. Second, for certain estimates, we assumed
that the current base level for calculating changes in ambient
air quality was roughly equivalent to levels existing in the
mid-1970s. Given the overall uncertainty in the available
information on benefits, changing the base year is within the
"noise level" of the estimates.
The effects of ozone on human health, vegetation, materials,
and ecosystems were summarized in the EPA Criteria Document for
ozone (U.S. EPA, 1978), currently being revised. In addition,
considerable new research has become available since the
Criteria Document was published.
Ozone changes are influenced by the amount of solar
radiation and changes in the concentrations of NOX and HC. They
are, therefore, very dependent on local conditions. To estimate
the national change in ambient ozone, we assumed average U.S.
-------
IV.12
atmospheric and meteorologic conditions. This averaging will
introduce additional uncertainty into the estimate because local-
ized conditions are not fully represented. Using the estimates
of avoided emissions in Table IV-3 and projections of total
emissions in 1988, we calculated the reductions in the HC and NOX
emissions under either of the two policy options under considera-
tion.* The reductions in HC and NOX in 1988 would be approxi-
mately 2.48% and 0.8%, respectively.
Baseline projections for NOX, HC, and CO for 1988 were
calculated as follows:
For NOX we used an EPA emission factor generated in the draft
model of MOBILE III for on-road vehicles in 1988 of approxi-
mately 3.19 g/mi. We assumed 159.6 million on-road vehicles
traveling an average 11,436 miles (see Appendix). Multiplying;
3.19 g/mi x 11,436 mi/vehicle x 159.6 x 106 vehicles
1 x 106g/metricton
- 5.82 x 106 metric tons
Assuming motor vehicles account for 35% of the NOX emissions
from transportation, stationary source fuel combustion, and
industrial processes, total NOX emissions in 1988 will be
16.634 x 10° tons (U.S. EPA, 1982b). Therefore, the 130,000
avoided tons of NOX in 1988 (Table IV-3) is approximately a
0.78% reduction.
The emission factors for HC and CO generated by EPA's
MOBILE III were reduced by .75 to adjust for that model's
exclusion of localities with Inspection and Maintenance
programs and the state of California which has its own, more
stringent, emission controls.
For HC, a 2.5 g/mi emission factor was used, along with
the assumption that motor vehicles account for 36% of the
HC emission from transportation, stationary source fuel
combustion, and industrial processes. Therefore, the HC
reduction is 314,000/12.67 x 106, or roughly 2.48%.
For CO, a 20 g/mi emission factor was used (generating
36.5 x 106 tons of CO), with the assumption that motor
vehicles emit 86% of all CO from transportation and resi-
dential fuel combustion. The reduction of 2,202,000 metric
tons is 5.18% of the total (5.18 = 2.2 x .86/36.5) in 1988.
-------
IV.13
Converting these changes in HC and NOX to subsequent changes
in ozone involves considerable uncertainty. Disagreement exists
as to the ultimate magnitude of the effect. For example,
research by General Motors suggested that because of scavenging
effects of NOX on ozone, decreases in NOX (holding HC constant)
may actually increase ozone levels in the area near the NOX
source; further downwind from the source, ozone levels would
decrease (Glasson, 1981). To the extent that ozone is scavenged,
however, nitrogen dioxide levels would increase and potentially
contribute to health effects and materials damage. Other General
Motors research suggested that decreases in both NOX and H-C will
reduce subsequent ozone, but by less than that resulting from a
reduction in HC alone (Chock et al., 1981). Because of the
uncertainty in predicting changes in ozone, we considered three
separate estimates to determine a reasonable point estimate for
the change.
First, a preliminary report recently completed for EPA by
ETA Engineering (1983) employed a method for relating HC emissions
to ozone production using the Empirical Kinetic Modeling Approach
(EKMA) recommended by EPA. ETA Engineering also evaluated the
actual changes in HC and ozone in the Chicago metropolitan area.
The analysis suggested a one-to-one relationship as an upper
bound between the percent change in HC and the resulting percent
change in ozone. Using this method, the decrease in ozone
concentration would be 2.48%. Unfortunately, the ETA model did
not explicitly incorporate the impact of changes in NOX.
-------
IV.14
A second estimate of the change in ozone was provided by
the work of Kinosian (1982). Using EKMA curves derived from the
Los Angeles basin as data, he regressed ozone levels on HC and NOX
concentrations. He found the following functional form fit the
data well for a wide range of HC/NOX ratios:
0.5
Z = a + b(HxN)
where: a and b are empirical constants that vary across locations
( .04 <_ a _< .06; .6 <_ b _< .8 )
and where:
Z = Ambient ozone levels
H = Ambient hydrocarbons
N = Ambient oxides of nitrogen.
To approximate the percent change in ozone due to percent changes
in HC and NOX, we set a = 0 and b = bo. Taking the logarithm of
this equation and the total derivative, we obtained:
dlog Z = .5 (dlog H + dlog N)
Since the derivative of the log function is a percent change,
the equation yielded:
% change in Z = .5 (% change in H + % change in N)
= .5 (2.48 + .78)
= 1.63
However because "a" is actually greater than zero, the change
in Z according to this model would be less than 1.63%. With a
nonzero "a", we obtained an approximation for the percent change
in Z using a power series expansion. Specifically:
•
log Z = log(a + b (H x N)-5) = log T + 2a
2T + a
-------
IV.15
where T = b (H x N)-5
Taking the derivative and collecting terms:
dz.
Z = % change in ozone = 1.625[l-(2ab(HN)•5)/(2T + a)2]
To determine the change in ozone, we assumed a = .05, b = .7,
an HC to NOX ratio of 10, and an average daily maximum ambient
ozone level of .054 ppm. These point estimates were averaged
from currently available data (Council on Environmental Quality,
1980). We were then able to solve directly for H and N using
Kinosian's equation relating ozone levels to HC and NOX. Substi-
tuting these values into the above equation, we obtained:
% change in ozone = (1.625)(1 - .118) = 1.43
Therefore, this technique generates a point estimate of 1.43%
for the actual change in ozone.
A third estimate of the change in ozone was determined using
recent EPA data from 1982 ozone State Implementation Plans (SIPs)
(U.S. EPA, 1984b). These SIPs estimated the percent of HC control
that was required to obtain a given reduction in ozone. The data
indicated that, as a best estimate, a 1.5% change in HC would
change ozone, on average, by 1%. Extrapolating linearly, this
suggested that a 2.48% reduction in HC would generate a 1.63%
reduction in ozone.
These three techniques yielded potential changes in ozone of
2.48%, 1.43%, and 1.63%. Since the last two methods explicitly
incorporate the impact of changes in NOX on ambient levels, we
have given them greater weight and used a point estimate of 1.5%
as the predicted change in ozone.
-------
IV.16
Because of transport, oxidant pollution is a regional, rather
than local, problem. Oxidant transport can occur over a range of
several hundred miles or more. Given its regional nature and the
nationwide distribution of the sources of ozone, we assumed a 1.5%
reduction in ozone concentration for the nation. Since the bene-
fits from ozone reduction will occur in both urban and rural
areas, despite site-to-site variation,* this 1.5% change for a
national estimate appeared reasonable.
We have estimated four benefits of reducing ozone levels:
health, agriculture, non-agricultural vegetation, and materials.
This is followed by estimates of the direct benefits of reduced
HC, NOX, and CO.
IV.C.I.a. Ozone Health Benefits
Studies of the effects of ozone on human health have
investigated the relationships between changes in ozone concen-
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).
Regarding the "sub-clinical" effects, for example, Hammer
et al. (1974) found an association of increased oxidants with
symptom rates of eye discomfort, cough, headache, and chest
discomfort in young, healthy adults. He obtained the symptom
*The ultimate change in ozone levels for rural areas is least
certain.
-------
IV.17
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. Even low levels
of exposure to photochemical oxidants were shown to provoke
these respiratory symptoms for adults with predisposing factors,
such as smoking or respiratory illness (Zagraniski et al., 1979).
Evidence of decreased athletic performance and dysfunction of
pulmonary systems was provided by Lippmann et al. (1983) and
Lebowitz et al. (1974).
Unfortunately, it was not possible to estimate economic
benefits from these studies of "sub-clinical" effects. Most of
them focused on determining either a threshold level for health
effects or whether there was a particular effect relating to
ozone exposure. Thus, no exposure-response relationship was
available from this literature.
Recent work by Portney and Mullahy (1983) at Resources for
the Future (RFF)* and data reanalysis by Hasselblad and
Svendsgaard (1975) were exceptions. The former study considered
the effect of alternative levels of ozone on, among other mea-
sures, the number of minor restricted activity days (MRADs) over
a two-week survey period. This health measure indicated how
frequently a person curtailed normal activity without actually
missing work or being bedridden. The second study was a statis-
tical reanalysis of the Hammer et al. study cited above. It
*The RFF study will undergo formal EPA peer review in April, 1984,
-------
IV.18
used logistic estimation to relate sub-clinical health effects
to alternative levels of ozone. Our results from applying each
of these studies follow.
The RFF analysis consisted of regressing MRADs on a number
of independent variables, including socioeconomic and demographic
factors, chronic health status, urban variables, ozone, and other
pollutants. Because of the inherent uncertainty in the analysis,
we used the 1.5-2.5 range for the estimated regression coefficient
of ozone (measured as the average daily maximum 1-hour concentra-
tion during the two-week survey period) indicated in the RFF
study. This resulted in an elasticity of 0.17-0.29 MRADs to
ozone. Therefore, a 1.5% reduction in annual average ozone
levels would reduce MRADs by 0.255% to 0.435%.
To calculate the health benefit from the reduction in ozone,
we applied these elasticities to the entire U.S. population in
1988, projected to be 245 million. We used summary statistics
from the RFF report that indicated an annual average of 10.32
MRADs per person. Using the low estimate of elasticity, the
improvement in ozone would result in 6.4 million (245 x 10^ x
10.32 x .00255) fewer MRADs per year for the U.S. population.
The higher elasticity generated an estimate of 11.0 million MRADs,
assuming some risk at all levels of exposure.
To generate "low-low" and "high-high" estimates we placed,
somewhat arbitrarily, two alternative values on an MRAD. As a
lower bound, we assigned a value of $7 per episode, approximately
10% of the average daily wage, to indicate some minimum amount
-------
IV. 19
a person would pay to avoid the minor restriction in activity.
We then applied this value to the lower estimate of the total
MRADs to yield an estimate of $45.1 million.
For the "high-high" estimate, we used $20 as the value of
an MRAD which, applied to the 11.0 million MRADs, yielded an
estimate of $220.0 million. This still may be a conservative
estimate for several reasons. To obtain the health benefits of
reducing air pollution, Freeman (1982) used a value of $20 (1978
dollars) for a restricted activity day (non-minor) and also added
expected reductions in medical expenses. In addition, MRADs
may affect work productivity. Our "low-low" and "high-high"
estimates produced a range of $45 to $220 million for 1988.
An alternative estimate of the health benefits from reduced
ozone concentrations was derived from the preliminary results of
Gerking et al. (1983), which demonstrated that survey respondents
were willing to pay to avoid suffering an increase in ozone
concentrations. Specifically, the study estimated that a 10%
reduction in ambient ozone concentrations would generate a per
person "willingness to pay" of $1.55 to $1.92 per year. Assuming
linearity, a 1.5% reduction would result in a benefit of $0.23
to $0.29 per person, or $57 to $71 million nationally. Unfortu-
nately, potentially serious problems with data and methodology
render this study only suggestive of the benefits of reducing
ozone. Nonetheless, it lends credence to the monetary estimates
suggested by applying the RFF model.
-------
IV. 20
To check the plausibility of these results, separate
estimates of symptoms were obtained using the Hasselblad and
Svendsgaard (1975) results. The authors fit logistic curves to
estimate the relationship between ozone concentration (measured
as a daily maximum hourly concentration) and eye irritation,
headache, cough, 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)]
The following parameter values (A, B, and C) were determined:
For eye irritation : A = -4.96, B = .0907, C ~ .0407
headache : A = -4.88, B = .0470, C = .0976
cough : A = -2.98, B = .0092, C = .0450
chest discomfort: A = -3.53, B = .0023, C = -.0166
To calculate the change in this probability due to a change
in ozone, we differentiated the above with respect to X. The
change in the probability (dp) of a given symptom due to a change
in ozone (dx) was:
dp(X) = B(l - C)[exp (- A - BX)] dX/ [1 + exp (- A - BX)]2
The change in probability can be estimated given information on
the existing daily maximum ozone levels. To use pollution
measures commensurate with the original estimation, we obtained
EPA's data from the entire Storage and Retrieval of Aerometric
Data (SAROAD) network of ozone monitors. Because of the non-
linearity in the equation representing the probability of a
-------
IV. 21
health effect, the mean daily one-hour maximum of ozone was cal-
culated for two separate six-month periods. One period included
data from the second and third quarters, April through September
(when higher ozone levels generally occur) , while the other
period used the first and fourth quarters of the year. For the
two periods we obtained average ozone measures of 6.1 and 3.3
pphm, respectively, yielding an annual average of 4.7 pphm.
To reduce the chances of obtaining more than one symptom per
person and thus double-counting people affected, we used chest
discomfort to represent the symptoms reported as cough as well as
those reported as chest discomfort. We calculated separately
the number of persons with reduced eye irritation and headaches.
Substituting the appropriate values for A, B, C, X, and dX
into the above equation for each six-month period, we obtained
the number of reduced symptoms. For example, to calculate the
expected number of cases of chest discomfort in the second and
third quarters, we used the following values: A = -3.53,
B = .0023, C = -.0166, X = 6.1, dX = (.015)(6.1) = .0915.
Substituting, to obtain the probability of a chest discomfort
symptom per person per day, yielded:
dp(X) = (.0915) (.0023) (1.0166)H/(1 + H)2
where H = exp(3.53 - ( .0023 ) ( 6. 1) )
= 6.00 x 10~6 per person per day
Multiplying by 245 million people and 183 days, we obtained a six
month total projection of 269,000 cases of chest discomfort in
-------
IV.22
1988. Using this equation, the total change in the number of
symptom days expected in 1988 was:
Chest discomfort: 400,000
Eye irritation : 5,783,000
Headaches : 2,493,000
Total: 8,676,000
Although we were not able to determine the exact correspon-
dence between MRADs and these reported symptoms, the projection
for these symptoms fell within the 6.4 to 11.0 million range
estimated for MRADs and supported the estimate.
Another important health effect of ozone, reported by
Whittemore and Korn (1980) and Linn et al. (1981), was the
exacerbation of asthma and nonspecific obstructive respiratory
disease. To estimate the decrement in asthmatic attacks result-
ing from the reduction in misfueling, we used the analysis of
Whittemore and Korn. They used a logistic curve to estimate the
probability that an asthmatic would have one or more attacks on
a given day. This probability was hypothesized to depend on air
pollution levels, temperature and humidity, the day of the week,
and the presence of an attack on the preceding day.
The results suggested that the probability of an attack was
significantly related to exposure to ozone. Specifically, their
results suggested the following:
log (p/l-p)= 1.66 z + biXi
where p = the probability of an attack
z = ambient ozone (24-hour average concentration)
Xi= meteorologic and other control variables.
-------
IV.23
To determine the change in the probability of an attack (dp) due
to a change in ozone (dz), we partially differentiated both sides
of the above (i.e., dX = 0) and solved for dp:
dp = 1.66 (p) (1-p) dz
To estimate the actual change in probability, we had to
determine the ambient ozone level and the baseline probability of
an attack, represented as p. Because of the inherent uncertainty
in these numbers, we used the point estimate to determine economic
benefits and then conducted a sensitivity analysis using alter-
native values for p and dz. An ambient ozone level was
approximated using available data for 1979 through 1982 (U.S. EPA,
1982a; Evans et al., 1983). For this analysis, we used a point
estimate of 0.040 ppm, but considered 0.035 ppm in the sensitivity
analysis.
Data on the baseline probability of an asthmatic attack were
difficult to obtain. These attacks vary widely in frequency,
duration, and intensity.* For example, asthmatics with a condition
characterized as "mild and intermittent" (roughly 60% of all
asthmatics) may have two or three attacks a year. However,
they may be ill-prepared to respond to severe attacks and may
undertake significant medical expenses. "Moderate" asthmatics
(25% of all asthmatics) may have one attack a month, requiring
* Estimates of the frequency and severity of asthmatic attacks
were based on personal communications with Jeff Cohen, U.S.
EPA, Office of Air Quality Planning and Standards, based on
his survey of experts.
-------
IV.24
some medical expense, lost work, or restrictions in activity.
There may be some chronic respiratory impairment. "Severe"
asthmatics (up to 15% of all asthmatics) may have several attacks
a month. Evidence from daily diaries in Salt Lake City and New
York City suggested 30-40 attacks per year. This group may be
better prepared for the attack, but may be on continuous medica-
tion and/or be forced to undertake significant preventive
actions.
To estimate the baseline probability of an attack, we used
a weighted average of the expected number of attacks for each
group. Multiplying the proportion of asthmatics in each classi-
fication by their average number of attacks per year, we obtained:
(.6)(3) + (.25H12) + (.15)(36) = 10.2 attacks per year
This generated a probability of 2.8% per day (10.2/365=2.8%).
Other research suggested a daily probability of an attack
ranging from 1.4-2.5% per day with a point estimate of 1.8%.*
Therefore, as a point estimate we used 2.0% for the daily
probability of an attack, indicating an average of 7.3 attacks
per year. (This estimate obviously does not reflect the extreme
variability among asthmatics.)
The change in the probability of an attack was calculated
using a point estimate of .04 ppm for ozone exposure, a 1.5%
change in ozone, and a value of .02 for the baseline probability
* These estimates were also based on personal communications
with Jeff Cohen, U.S. EPA, Office of Air Quality Planning
and Standards.
-------
IV.25
of an attack. Substituting:
dp = (1.66)(.02)(.98)(.0006) = 1.95 x lQ-5/person/day
To estimate the population at risk, we used estimates of the
numbers of asthmatics and atopies (persons potentially sensitive
to ozone) in the entire U.S. population (245 million). Currently,
4% of the population is considered asthmatic, with an additional
9% considered atopic. Thus, the population at risk is 13% of
245 million, or 31.85 million. The total reduction in the annual
number of attacks would be:
227,000 = 1.95 x 10~5 x 31.85 x 1Q6 x 365
To determine the monetary benefit of these reduced attacks
we had to assign a value per attack avoided. Ideally, this would
equal the individual's (or society's) willingness to pay to
prevent the occurrence of an attack. Unfortunately, no data
exists on this willingness to pay. Likewise, we could not find
any published estimates of the average medical costs incurred
for an attack.
Based upon the existing evidence of the potential severity of
an attack, we arbitrarily valued each attack at $70, the average
daily wage, as a "ballpark" estimate. Because of their chronic
condition, 7% of all asthmatics are consistently forced to limit
their activities outside of their "major activity" such as working,
keeping house, or going to school (U.S. DHEW, 1973a). Another
7% are forced to limit their amount or kind of major activity.
Finally, an additional 1.5% are unable to pursue any major activity
-------
IV.26
Data also existed on the frequency and degree of annoyance
from an asthmatic condition (U.S. DHEW, 1973b). Of the asth-
matics sampled, 52% reported that they were bothered by asthma
"once in a while," 21% were bothered "often," and 14% were
bothered "all the time." Regarding the degree of bother, 11%
reported "very little," but 36% reported "some" and 43% reported
a "great deal" of bother. Evidence on visits to physicians
indicated that 40% of the asthmatics saw a doctor two or more
times a year, while 20% saw a doctor five or more times a year.
Finally, 51% of the asthmatics were taking medicine or following
treatment recommended by a doctor. Thus, given the degree of
bother and medical care involved, we used $70 per attack as a
point estimate. However, it may well be that many asthmatics
would be willing to pay more than $490 ($70/attack x 7 attacks
per year) to prevent any asthma attacks from occurring in the
year.
Multiplying the 227,000 expected attacks by $70, we
estimated benefits of $15.9 million in 1988. Table IV-5 displays
the sensitivity of this estimate to alternative values for the
baseline probability, the ozone level, and the value of an attack
A reasonable range for the benefits of reduced asthma attacks is
$10.5 to $28.2 million, with a point estimate of $15.9 million.
-------
IV.27
Table IV-5
Benefits of Reducing Asthmatic Attacks in 1988
High Estimate
Point Estimate
Low Estimate
Ozone Level
. 0 4 ppm
. 0 4 ppm
.035 ppm
Baseline
Attack Rate
Probability
2.5%
2%
1.5%
Value of
Attack
$100
$ 70
$ 70
Benefits
$28.2 million
$15.9 million
$10. 5 million
To avoid the double counting of asthma attacks with the MRADs
calculated above, we subtracted the estimated 227,000 attacks from
the number of MRADs that were estimated to occur. We then added
the point estimate of $15.9 million for reduced asthma attacks to
the adjusted "low-low" and "high-high" estimates of MRADs to obtain
the total acute health benefits. Consequently, these studies sug-
gested benefits from the reduction in acute health effects in 1988,
including both MRADs and asthma attacks, ranging from $59.3 million
to $233.4 million, with a point estimate of $146 million.
These ozone health benefits reflect the likely acute effects
generated by intense, short-term exposure to ozone. Long-term
exposure to ozone may also 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.
-------
IV.28
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 statis-
tically significantly increased chest illness, impairments of
respiratory function, and lower pulmonary function.*
In addition to the sparse epidemiological evidence of the
effects of long-term exposure to ozone, several animal experi-
ments have demonstrated 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, 1983a). While work is now under way to
extrapolate these animal data to human dose-response functions,
this is not presently possible. Therefore, we could not quantify
the chronic health effects attributable to ozone, but we believe
that some of these effects may be significant at current ambient
levels.
Using the studies cited above, the total health benefits in
1988 from the reduction in ozone due to reduced misfueling was
$146 million from reduced MRADs and asthmatic attacks, plus
potential reductions in chronic health conditions from decreased
ozone levels. If we symbolize these non-monetized health benefits
as OZC, the total health benefits are $146 million + OZC.
* 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 864 and there was some question about the adequacy of
the measurement of ozone exposure and about the subject selec-
tion and the test measures. Although it is both reasonable and
likely that long-term exposures are harmful to health, the
failure to correct for the effects of other pollutants raises
uncertainties about the specific findings.
-------
IV.29
IV.C.l.b. Ozone Agricultural Benefits
Research has shown that ozone alone, or in combination with
sulfur dioxide and nitrogen dioxide, is responsible for most of
the U.S. crop damage associated with air pollution. Ozone can
affect the foliage of plants by biochemical and cellular alter-
ation, thus inhibiting photosynthesis and reducing plant growth,
yield, and quality.
To generate a top-down estimate of agricultural benefits,
we used generalized relationships between ozone concentration,
yield, and economic loss. The aggregate estimates of Adams (1983)
and SRI (1981), as summarized by Freeman (1982), suggested that
the average annual benefits associated with a 10% reduction in
ozone concentrations were $200 to $500 million in 1983 dollars. To
use the top-down approach, we assumed that this relationship held
over a broad range of exposures. Thus, the 1.5% ozone reduction
could produce a benefit of $30 to $75 million from increased crop
yields (1.5%/10% times $200 to $500 million = $30 to $75 million).
As an alternate approach, we followed the bottom-up approach
of Kopp (1983). He estimated the effects of ozone changes on
soybeans, wheat, corn, peanuts, and cotton on a county-by-county
basis. Because this analysis directly incorporated estimates of
the demand and supply elasticities for these crops, it appeared
to be the most precise assessment of benefits available. His
estimates suggested that a 1.5% change in ozone would produce
approximately $120 million in lost economic value (1983 dollars).
These five crops accounted for roughly 76% of the total value of
-------
IV.30
commercial crop production in the United States. Therefore, we
scaled Kopp's estimate of crop damage by assuming that ozone
damages to all other crops occur in the same proportion as their
relative values. The benefits of the 1.5% change in ozone grew
to approximately $157.5 million (1983 dollars).
Another benefit estimate for reduced ozone, conducted on a
crop and region-specific basis, was provided directly from the
National Crop Loss Assessment Network (Heck et al. , 1983). Their
estimate of the effects on economic surplus (consumer and
producer well-being) included only crops in the corn belt --
corn, soybeans, and wheat -- which are less than half of all
expected crop losses from ozone. Their results suggested that a
3% reduction in ozone would increase economic surplus by $140 to
$230 million. Assuming linearity, a 1.5% change would generate
a surplus of $70 to $115 million. If these crops in the corn
belt constitute 50% of all ozone-related damages (probably a
high estimate), the total benefits of the 1.5% reduction would
be $140 to $230 million.
Together, the benefit estimates from these studies ranged
from $30 to $230 million per year. To determine a point estimate
of the damage to agricultural crops, we weighted the last two
analyses most heavily, because they contained the most precise
estimates of changes in economic welfare. This suggested a point
estimate for agricultural loss from ozone concentrations of
approximately $160 million.
-------
IV.31
IV.C.I.e. Nonagricultural Vegetation Benefits of Reduced Ozone
The estimates presented above addressed only agricultural
crops. They excluded damages to forests and ornamental plants,
which may be substantial. For example, in a very small contingent
valuation study, Crocker and Vaux (1983) found that the shift of
an acre of timberland in the San Bernardino National Forest from
either the severely or moderately harmed category into the
unharmed category would generate additional annual recreational
benefits of $21 to $68 per person. These findings are difficult
to generalize for the rest of the nation because the San Bernardino
area has very high ozone concentrations and because of other site
attributes and socioeconomic characteristics. Nevertheless, they
indicate that reduced damages to vegetation may produce signifi-
cant benefits.
The preliminary draft of the Ozone Criteria Document (1983a)
also provided additional qualitative evidence: "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 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/m^)." For commercial
timber purposes, however, damages are likely to be small, as most
-------
IV.32
commercial forests are in areas with low ozone concentrations.
In areas with relatively high concentrations, trees resistant to
ozone can be planted.
Finally, we present one quantitative estimate, noting that
it was based on very sparse data and generated by making some
significant abstractions from existing studies. Heighten et al.
(1983) have estimated that the benefits associated with non-
agricultural vegetation from a 10% reduction in ozone concentra-
tions are $0.0 to $100 million. Assuming linearity, the benefits
from a 1.5% reduction in ozone would be $0.0 to $15.0 million,
with a point estimate of $7.5 million for 1988. We stress,
however, that the existing evidence is uncertain.
IV.C.l.d. Ozone Materials Benefits
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-
chasing of specially resistant material) and aesthetic losses.
Only the direct costs were incorporated in this analysis, however,
In his survey of the literature, Freeman (1982) suggested
that annual material damages from oxidants and NOX amounted to
approximately $1.1 billion (1978 dollars). Using the Consumer
Price Index as well as census figures on projected population
increases to update the figure, produced an estimated $1.88
-------
IV.33
billion for 1983. An ozone reduction of 1.5%, assuming linear-
ity, suggested a benefit of $28.2 million annually.
We obtained an alternative estimate of the benefits from
reduced materials damage by using dose-response information from
the Ozone Criteria Document (U.S. EPA, 1978). The text supplied
per capita economic damages for elastomers, textiles, industrial
maintenance, and vinyl paint costs as a function of annual ozone
levels. Using an annual 24-hour mean for ozone of .040 ppm, as
reported above, and a population estimate of 245 million for 1988,
we calculated a benefit range of $16.4 to $22.6 million in 1983
dollars (point estimate of $19.5 million), for a 1.5% reduction
in ozone. Taking the arithmetic mean of the point estimate from
the two different approaches for materials benefits yielded a
point estimate of $24 million annually in 1988.
IV.C.2. Benefits of Reducing NOX Emissions
NOX emissions are believed to damage health and materials,
to contribute to reductions in visibility, and are associated
with acid deposition. In addition, damage to vegetation has been
demonstrated experimentally. Unfortunately, specific dose-
response information relating to NOX is sparse. As a consequence,
only broad aggregate estimates were presented to approximate the
effects of NOX emissions on health and welfare.
Materials damage from NOX are not included in this section
since it was contained in the ozone benefits section. While
there may be acid rain benefits as well, we have not included
-------
IV.34
them because of the uncertainties over the role of NOX in acidic
deposition. Therefore, we included only the benefits of reduced
health effects and improved visibility, as benefits for reducing
NOX emissions in 1988.
Regarding NOX health effects, EPA, is currently reviewing
its ambient air standard for nitrogen dioxide (NC>2).* The
Clean Air Scientific Advisory Committee (CASAC) recommended that
any NC>2 standard should protect against repeated short-term
"peak" exposures and against long-term "chronic" exposures
because of possible health effects.**
Repeated exposure to short-term peaks of NOX has been
associated with excess respiratory illness and symptoms in
children, and with small (but statistically significant) reduc-
tions in lung function (U.S. EPA, 1982c). Because repeated
episodes of respiratory tract irritation and illness in children
may carry into adult life in the form of reduced lung function
and chronic bronchitis, NOX reductions may also reduce subsequent
adult cases of chronic bronchitis. Long-term exposure to low
level NC>2 may contribute to emphysema. Thus, significant bene-
fits, although unguantified in this paper, may result from
controlling NOX.
Surveying several research efforts, including those linking
NOX to changes in property values (which may capture both health
* NC>2 is an indicator pollutant for all nitrogen oxides,
** CASAC closure letter on OAQPS Staff Paper for NOX,
July 6, 1982.
-------
IV.35
and welfare effects), the National Academy of Sciences (1974)
suggested a range of $1.0 to $8.0 billion, adjusted to 1983
dollars, for the annual effects other than materials damage.
Assuming proportionality between the predicted .78% reduction of
NOX and reduced damages, the benefits would be $7.8 to $62.4
million annually. We used the midpoint of this range, $35
million, as the point estimate.
IV.C.3 Reducing Emissions of Hydrocarbons
The various chemicals constituting hydrocarbons from automo-
bile emissions may affect health. Specifically, benzene, which
is believed to cause leukemia, constitutes 4% of total tailpipe
HC emissions (U.S. EPA, 1983b).
To estimate the number of benzene-linked leukemia deaths we
might avoid by eliminating misfueling, we used the EPA Carcinogen
Assessment Group (CAG) Risk Assessment for Benzene. This analysis
predicted that human exposure to automobile benzene emissions*
resulted in an estimated 50.89 leukemia deaths per year in 1976
(U.S. EPA, 1979). As displayed in Table IV-3, we estimated that
misfueling in 1988 would produce 314,000 metric tons of HC
emissions, or 4.99% of the 6.29 million tons of automobile HC
emissions in 1976, the year of CAG's analysis (U.S. EPA, 1982b).
This estimate, however, was based on a unit risk estimate for
* CAG assesses risks as the amount of exposure (in parts per
billion), times the population exposed, times duration of
exposure. Their benzene analysis yielded 150 million ppb-
person-years.
-------
IV.36
benzene (.024/ppm) which was revised by CAG in November 1981
(.022/ppm). Using this new unit risk estimate and CAG's analysis,
automobile tailpipe-benzene emissions were predicted to result in
an estimated 47.34 leukemia cases per year (U.S. EPA, 1974a).
Therefore, assuming linearity, 4.99% of the 47.34 leukemia deaths,
or 2.36 deaths, would be avoided by preventing misfueling in 1988.
This assumed that benzene would be the same fraction of the
reduced HC emissions as it was of total automotive HC emissions
in 1976.
Economic studies (Brown, 1978; Thaler and Rosen, 1976)
suggested that people are willing to pay $0.45 to $7.0 million to
save a "statistical" life. Under this assumption, the health
benefits of avoiding the HC emissions would be $1.06 to $16.52
million in 1988. We used the geometric mean of this range, $4.19
million, as our point estimate.
Hydrocarbons also 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 func-
tion of the amount of HC in the atmosphere). However, the ability
to quantify these complex relationships has just been developed,
and experts at Systems Applications Incorporated (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 hydrocarbon and NOX concentrations, and oxidant
and sulfur dioxide levels) for which we have only limited data.
A recent modeling analysis by SAI (Seigneur et al., 1982)
-------
IV.37
indicated that a 50% reduction in HC would reduce sulfates in
urban areas by 30% to 60%. However, because of the uncertainty
surrounding this estimate at this time, and the uncertainty in
interpolating this to a 2.4% change, we did not explicitly
consider the reduction in sulfates in this analysis. Because
the reduction in sulfates would generate significant economic
benefits from improved health and visibility and reduced soiling,
this omission may seriously underestimate the benefits.
IV.C.4. Reducing Emissions of Carbon Monoxide
Existing scientific knowledge concerning CO indicates that
health impacts are the primary concern at or near ambient levels.
Current information suggests that persons with cardiovascular
disease are most sensitive to low levels of CO. Additional
subgroups of the population also believed to be sensitive to CO
exposure are people with chronic respiratory diseases, pregnant
women, and the elderly. Unfortunately, clinical dose-response
functions relating low level CO exposure to particular health
effects, when estimated, have not been conclusive. Therefore,
we have not estimated quantitatively the impact that reduced CO
(through reduced misfueling) may have on health. However, we
have described the impact that misfueling may have on the
distribution of carboxyhemoglobin (COHb) levels for the U.S.
population.
Probably the greatest concern about CO exposure is its
effect on the cardiovascular system. The effect of CO thus far
-------
IV.38
measured at the lowest level of exposure is reduced exercise
time until the onset of angina pectoris. This clinical pheno-
menon is a result of insufficient oxygen supply to the heart
muscle and is characterized by spasmodic chest pain, usually
precipitated by increased activity or stress, and relieved with
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 is not believed to be associated with
permanent anatomical damage to the heart. Nonetheless, the
discomfort and pain of angina can be severe, and each episode of
angina may carry the risk of myocardial infarction (the death of
a portion of the heart muscle). However, epidemiological studies
as yet have provided inconclusive results on the association
between CO exposure and the incidence of myocardial infarction.
The health effects from exposure to CO are associated with
the percentage of total blood hemoglobin that is bound with CO,
producing carboxyhemoglobin (COHb), and thereby reducing the
oxygen-carrying capacity of the blood. The median concentrations
of COHb in blood are about 0.7% for nonsmokers and about 4.0% for
smokers. At 2.9% COHb, at least one clinical study (Anderson et
al., 1973) associated reduced exercise time until the onset of
pain in patients with angina pectoris. At 4.5% COHb, this same
study reported an increased duration of angina attacks.
The potential health improvements from reduced CO may be
great for two reasons. First, there are many people in the
-------
IV.39
population believed to be sensitive. EPA has estimated that
5.0% of the U.S. adult population — roughly 9.5 million people
— have definite or suspected coronary heart disease. Of this
group, 80% have suspected or definite angina pectoris (U.S. EPA,
1980). Second, the blood of many people shows concentrations of
COHb above 2.9%, the lowest level of COHb where adverse effects
are indicated. Data from the second National Health and Nutri-
tion Examination Survey (NHANES II) indicated that for the U.S.
population over twelve, 2% of those who have never smoked, 3% of
former smokers, and 66% of current smokers exceeded 2.9% 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 (3% to 7%) in work time
until exhaustion in healthy young men with COHb levels at 2.3%
to 4.3% (Horvath et al., 1975; Drinkwater et al., 1974; Raven
et al., 1974). At higher COHb levels (5% to 7.6% and above),
investigators have reported impaired visual perception, manual
dexterity, ability to learn, and performance of complex sensor-
imotor tasks in healthy subjects.
Finally, 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, that enhance oxygen need, or that elevate the
sensitivity of the tissues to any oxygen imbalance. Sensitive
groups may include:
-------
IV.40
0 people with peripheral vascular diseases such as
atherosclerosis and intermittent claudication
(0.7 million people);
° people with chronic obstructive pulmonary diseases
(17 million people);
0 people with anemia or abnormal hemoglobin types that
affect the oxygen-carrying capacity of the blood
(0.1-.245 million people);
0 people drinking alcohol or taking certain medications
(e.g., vasoconstrictors);
0 the elderly;
0 visitors to high altitudes; and
0 fetuses and infants* (3.7 million total live births
per year).
A comprehensive economic estimate of the benefits from
reduced CO is not possible. The current medical literature does
not provide a dose-response relationship between COHb levels and
specific health effects that can be valued monetarily. However,
analysis relating changes in CO emissions to the distribution
of COHb levels in the U.S. is possible using NHANES II. Work
still in progress indicates that a change in ambient CO levels
may have significant impacts on the distribution of COHb.
* Animal studies showed that pregnant females exposed to CO
reported lower birth weights, increased newborn mortality,
and lower behavioral levels in newborn animals, even when
no effects on the mothers were detected. In addition,
research has reported a possible association between
elevated CO levels and Sudden Infant Death Syndrome
(Hoppenbrouwers et al., 1981).
-------
IV.41
IV.D. Summary of Health and Welfare Benefits
Table IV-6 summarizes the direct estimates of the 1988
benefits of reducing HC and NOX by pollutant and benefit sub-
category. The range of $114 to $579 million in annual benefits
incorporates the estimates of both the top-down and bottom-up
approaches. The point estimate of $377 million was derived by
aggregating the best estimates of each subcategory.
Table IV-6
1988 Benefits of Reducing HC, NOy and
CO Emissions by Direct Estimation
(millions of 1983 dollars)
Benefit Category Range Point Estimate
Ozone
Acute Health $59-233 $146
Agriculture $30-225 $160
Vegetation $ 0- 15 $7.5
Materials Damage $16- 28 $ 24
Chronic Health NA OZC
NOX
Health and Visibility $7.8-62 $35
Hydrocarbons
Health $1.06-16.52 $4.19
Sulfate Deposition (Health, NA NA
Materials Damage, Visibility)
Carbon Monoxide
Acute Health NA
TOTAL $ 114-579 $ 377 +OZC +CMA
NA = Quantitative estimates not available or attempted.
CMA = Non-quantified benefit of reducing acute health effects
from CO.
OZC = Non-quantified benefit of reducing chronic health effects
from ozone.
-------
IV.42
IV.E. Summary of HC, CO, and NOX Benefits
As we noted earlier, there is no consensus on a good, simple
way to value the benefits of eliminating misfueling and its
consequent excess emissions. As a result we have used three
different approaches:
0 the value using the costs of alternative regulations;
0 the value of preserving catalysts; and
0 the value of avoiding damage to health, vegetation, and
materials.
Table IV-7 summarizes the values obtained by each of these three
methods.
The first method computed the value of reduced emissions
by using the cost of HC, CO, and NOX regulations that EPA is
considering promulgating. This revealed the low end of the range
of values that EPA or Congress impute for controlling additional
increments of these pollutants.
The second method of valuation used the cost of catalytic
converters and other emission control equipment disabled by mis-
fueling to approximate the benefits of eliminating misfueling.
Finally, the third method directly calculated some of the
health and welfare benefits of reducing HC and NOX emissions,
by applying the results of research that related improvements
in air quality to improvements in human health, or reductions in
damages to materials and vegetation.
Our health and welfare estimates are probably low because
they do not include all the potential health and ecological
-------
IV.43
effects of ozone and CO. This method of valuation is also less
certain than the other methods. Conceptually, however, it is a
reasonable (and probably the best) way to measure the social
benefits of reducing emissions of HC, NOX, and CO.
We used the mean of this direct estimate and the value of
preserving catalytic converters as the best estimate of the
benefits of reducing misfueling. We obtained a value of $405
million to represent the 1988 benefits of reducing HC/ NOX, and
CO emissions through the elimination of misfueling. Note that
the different methods yielded fairly similar estimates of the
benefits.
Table IV-7
Summary of Benefits in 1988 of Reducing HC, CO, and NOy Emissions
(millions of 1983 dollars)
Value by Directly
Value by Next-Step Value of Perserving Estimating Health
Regulations Catalytic Converters and Welfare
$121 - 452 $432 $114 - 579
(point estimate: $377)
-------
IV. 44
TECHNICAL APPENDIX FOR CHAPTER IV
Accurately estimating the costs and benefits of reducing lead
in gasoline required the use of disaggregated data, some of which
was not readily available. For this reason, we developed a fairly
simple "bottom-up" model to forecast light-duty fleet size and mix,
numbers of misfuelers, and gasoline demands by various categories
of vehicles. This technical appendix describes this model.
Overall Structure of Model and Summary of Estimates
In general terms, the fleet model can be broken into five
major pieces:
0 It ages the existing stock of cars (1982) and light trucks
(1980) — using data from Polk, 1983 — and includes Data
Resources, Inc. (DRI) projections of sales from 1983 to
1988, to estimate the size and composition of the light-
duty fleet in 1988. Appendix Tables IV-1 and IV-2 show
the projection of this fleet into 1988.
0 Misfueling rates by age of vehicle are used to estimate
both the number of misfueled vehicles and those that
would misfuel for the first time in 1988 under current
policies. The sources of misfueling data are surveys
conducted by EPA's Office of Mobile Sources. Our analysis
assumed that current misfueling rates would continue.
0 The model estimates excess emissions due to new misfuelings
in 1988 by aging (retiring) the new misfuelers over the
subsequent 20 years (to 2007), calculating the expected
-------
IV. 45
APPENDIX TABLE IV-1
LIGHT-DUTY VEHICLE PROJECTIONS
(thousands of vehicles)
CARS IN
OPERATION
MODEL YR 1982
(Polk)
1983
1984
1985
1986
1987
1988
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
1974
1973
1972
1971
1970
1969
1968
1967
1966
8,000
8,280
8,825
10,075
10, 155
9,661
8,471
6,190
7,498
7,629
5,989
4,243
3,581
2,822
2,208
1,609
5,220
BESSSOCBS
9,200
7,992
8,255
8,723
9,838
9,927
9,429
8,099
5,790
6,836
6,725
5,098
3,521
2,910
2,282
1,789
1 , 306
4, 176
BSBBBB
10,500
9, 191
7,968
8, 159
8,518
9,618
9,689
9,015
7,576
5,279
6,026
5,724
4,230
2,861
2,354
1,850
1,453
1,045
3,299
BBBBBB
1 1 , 000
10,490
9, 163
7,876
7,968
8,327
9,386
9,263
8,432
6,907
4,653
5,129
4,750
3,438
2,314
1,907
1,502
1,162
826
2,540
BBBBBB
1 1 , 200
10,989
10,458
9,057
7,690
7,789
8, 127
8,974
8,665
7,688
6,088
3,961
4,256
3,860
2,781
1,875
1,549
1,201
918
636
1,905
B8BBSS
1 1 , 600
11,189
10,956
10,337
8,844
7,518
7,602
7,770
8,395
7 , 900
6,777
5, 182
3,287
3,459
3, 122
2,253
1 , 523
1 , 239
949
707
477
0
BaaaaB an
1 1 , 800
1 1 , 588
11, 155
10,829
10,094
8,646
7,337
7,268
7,268
7,653
6,964
5,768
4 , 300
2,671
2,798
2,530
1,829
1,218
979
731
530
0
0
:BBBBBB:
110,456 111,897 114,353 117,033 119,668 121,084 123,957
-------
IV. 46
APPENDIX TABLE IV-2
LIGHT-DUTY TRUCK PROJECTIONS*
(thousands of vehicles)
LDTs IN
MIODEL OPERATION
YEAR 1980 1981
(Polk*.87)
1982
1983
1984 1985
1986
1987
1988
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
1974
1973
1972
1971
1970
1969
1963
1967
1966
1
2
2
2
2
1
1
1
1
1
1
1
24
,936
,931
,844
,580
,139
,494
,976
,927
,608
,127
,022
,048
769
640
556
,597
1,826
1,926
2,907
2.815
2,546
2,099
1,458
1,914
1,849
1,595
1,013
949
961
698
575
496
25,628
1,748
1,817
1,911
2,878
2,778
2,498
2,048
1,412
1,837
1 , 834
1,434
940
870
872
627
513
439
26,456
2,520
1,739
1 , 802
1,891
2,840
2,726
2,437
1,984
1,355
1,822
1,649
1,331
862
790
784
560
454
387
27,933
2,970
2,507
1,725
1,784
1,866
2,787
2,660
2,361
1 , 904
1,344
1,638
1,530
1,220
783
709
700
495
401
340
29,724
3,050
2,955
2,487
1,708
1,760
1,831
2,719
2,576
2,266
1,888
1,208
1,520
1,403
1, 108
703
633
618
437
352
297
31,521
3,
3,
2,
2,
1,
1.
1,
2,
2,
2,
1,
1,
1.
1,
33,
120
035
931
462
685
727
787
633
472
247
697
121
394
274
995
623
560
546
384
308
260
267
3,320
3, 1O4
3,010
2,902
2,429
1,654
1,685
1,731
2,527
2,452
2,020
1,575
1,028
1,265
1,145
888
555
494
480
335
269
0
34,870
3,450
3,303
3,079
2,980
2,863
2,384
1,613
1,632
1,661
2,507
2,204
1,875
1,444
933
1, 137
1,022
785
490
434
419
293
O
0
36,511
* Trucks 0-8500 Ibs.
-------
IV. 47
excess emissions of EC, CO, and NOX in each year
(based on both the extra grams of emissions per mile
traveled and annual miles per vehicle by age). It then
discounts these emissions (at 3% rate) back to the year
of misfueling — 1988.
0 The model estimates gasoline demand in 1988 for four
major categories of demand: those vehicles designed for
and using leaded gasoline, those designed for leaded
gasoline but switching to unleaded premium (for the octane),
those vehicles designed for and using unleaded gasoline,
and those misfueling with leaded gasoline. A fifth
category is "special" uses for heavy trucks, agricultural
equipment, boats, etc. We hold "special" use demand
constant at 9.6% of total gasoline demand, the 1982
percentage.
Table IV-3 is a summary of the results.
APPENDIX TABLE IV-3
SUMMARY OF FLEET MODEL PARAMETERS
Total f of light-duty cars and trucks in 1988: 159,644,000
Incremental # of vehicles assumed to misfuel in 1988: 2,524,000
Total # of vehicles in 1988 misfueling in all years: 19,481,000
Overall misfueling rate: 12.2%
Average miles per gallon for cars and trucks: 20.4
Average miles per year per cars and trucks: 11,436
Total demand for gasoline (million gal/yr) : 100,737 100%
Legal light duty demand for leaded (million gal/yr): 12,485 12.4%
Misfuelers1 demand for leaded (million gal/yr): 10,290 10.2%
Demand for unleaded (million gal/yr): 68,290 67.8%
Other legal demand for leaded (million gal/yr): 9,671 9.6%
-------
IV. 48
Sources of Data and Major Assumptions
We found it necessary to draw actual data from several
different sources to estimate other important pieces of infor-
mation. In general, we used the following hierarchy of sources;
if a preferred source did not provide the data, or did not
provide it in enough disaggregation, we turned to the next-pre-
ferred source.
HIERARCHY OF SOURCES OF DATA
R.L. Polk & Co. (mostly provided in MVMA Facts & Figures)
U.S. DOT/FHA: Highway Statistics 1982
U.S. EPA Office of Mobile Sources: MOBILE II Documentation
The Transportation Energy Book
These sources are all referenced in Chapters III or IV. In
addition, we also derived certain estimates based on the
data these sources presented.
Sensitivities of Our Projections to Alternative Assumptions
Our predictions of total gasoline demand in 1988 are sen-
sitive to the average miles per year traveled by vehicles, to
the projected sales of cars and light trucks in each year to
1988, and to the scrappage rates we used to retire portions of
each cohort in each year. Roughly, changes in these parameters
cause proportional changes in gasoline demand. Appendix
Table IV-4 lists the basic age-related assumptions.
Data concerning average miles per vehicle per year* (MPV)
came from EPA's MOBILE II documentation, representing about a
* Wherever possible with data and method, we disaggregated by
cars and light trucks (0-8500 Ibs. GVW), and by age of vehicle,
-------
IV.49
APPENDIX TABLE IV-4
GENERAL FLEET ASSUMPTIONS
AGE
1 OR <
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
# OF
LDVa*
15,250
14,892
14,235
13,809
12,957
1 1 , 030
8,951
8,900
8,929
10, 160
9, 168
7,643
5,745
3,604
3,934
3,552
2,615
1 , 708
1,413
1,150
7. MIS-
FUELING
IN 1 YR*
5.5
1.7
1.6
1.6
1.6
1.6
1.7
1.6
1.6
1.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
CUMU- V
LATIVE
7. MIS- -
^ FUELING
5.5
7.7
8.8
10.4
12.0
13.6
15.3
16.9
18.5
20.1
20. 1
20.1
20. 1
20. 1
20. 1
20. 1
20. 1
20. 1
20. 1
20. 1
VEHICLE SURVIVAL AVG ANNUAL MILES
RATES * PER VEHICLE **
CARS
0.999
0.996
0.988
0.976
0.978
0.976
0.956
0.935
0.912
0.881
0.851
0.830
0.813
0.809
0.810
0.812
0.800
0.790
0.770
0.750
LOT
0.995
0.987
0.977
0.964
0.946
0.923
0.894
0.858
0.851
0.765
0.710
0.651
0.591
0.531
0.474
0.419
0.370
0.325
0.284
0.248
CARS
14,400
14,275
13,775
13,250
13,250
12,675
12, 175
1 1 , 650
1 1 , 075
10,575
10,050
9,475
8,975
8,450
7,875
7,375
6,850
6,275
5,775
5,000
LOT
15,676
15,276
13,692
12,223
11,001
9,992
9,238
8,488
7,913
7,413
6,929
6,510
6, 163
5,829
5,425
5, 160
4,954
4,625
4,400
4,400
are. SB SEES S
159644
13.5
Note: Light-duty vehicles (LDVs) includes cars and light-duty trucks (LDTs)
* First time misfuelers
** Includes current and past misfuelers. (Source: U.S. EPA Office of Mobile
Sources, 1983d.)
+ Source: R.L. Polk & Co. in MVMA Facts & Figures, 1981
++ Source: U.S. EPA Office of Mobile Sources, 1983d.
-------
IV. 50
1.1% annual growth from actual 1980 MPV figures (estimated by
Polk in MVMA, 1982). In 1988, these figures are about 9% greater
than 1980 figures. Consequently, if one were to use 1980 data,
gasoline demand would be 8-9% lower.
Estimates of the initial number of cars from each model year
came from Polk, as reported by MVMA. The data on light trucks
were acquired by the EPA directly from Polk, but were adjusted
downward by 13%, to transform the category from 0 to 10,000 pounds
to 0 to 8500 pounds. This adjustment was derived by a comparison
of several different sources of data and is used by EPA's Office
of Air and Radiation.
We use transitional probabilities of survival in order to
retire some portion of each cohort as it moved into the next age
category. That is, 99.6% of one year old cars live to be two
years old; 98.8% of two year olds live to be three, etc. For
cars, we averaged the transitional probability of survival for
each age group reported by Polk in MVMA for 1978-1982. For
light trucks, sufficient Polk data were not available; instead,
we used survival rates estimated in Kulp and Holcomb (1982). We
did not use their estimates for cars because it was derived by a
model with which we were not familiar and which used scrappage
rates well above any observed in recent years. We used a 7.4%
scrappage rate in the current analysis. Using Kulp and Holcomb's
estimate of 10.5% would decrease gasoline demand from 100.7
billion gallons to 89 billion gallons in 1988. In addition,
such a change in assumptions would increase the unleaded market
share from 67.8% to 70.5%.
-------
IV. 51
We used Data Resources, Inc. (DRI) projections for sales
of cars and light trucks (TRENDLONG2008B), as reported in U.S.
Long-Term Review (Fall, 1983). Miles per gallon per vehicle
came from the road mileages reported in EPA's Passenger Car Fuel
Economy and were adjusted for change in fuel economy by age.
There are several assumptions that do not influence total
demand for gasoline but do determine the split between leaded
and unleaded grades. Most important are misfueling rates by
age of vehicle, and, in particular, the shape of this curve in
the youngest model year cohorts (i.e., model years 1985-1988).
This part of the fleet is particularly important because:
there will be more of these vehicles in 1988 than older cohorts,
these vehicles will be emitting farther into the future than
older vehicles, and, because of discounting back to the year
of misfueling, they are weighted most heavily. We used EPA's
1982 survey of vehicle tampering for raw data, which provided
misfueling rates by age of vehicle. We used regression analysis
to estimate the relationship between age and incremental mis-
fueling, using several specifications of form. By far, the best
fit was a tri-linear form, with a 5.5% increase in the first
year of the cohort's existence, a 1.66% increase per year for
ages 2 to 9, and with no incremental misfueling in subsequent
years. (In 1982, the time of the survey, vehicles with catalytic
converters had been sold for only seven years, so no data existed
on misfueling beyond the seventh year.)
-------
IV. 52
Listed below are the actual 1982 survey results and the
regression estimates used in the analysis.
APPENDIX TABLE IV-5
MISFUELING RATES BY AGE
(as percentage of model year cohort)
EPA REGRESSION
AGE 1982 SURVEY ESTIMATES
1 or less 5.2% 5.5
2 7.4% 7.2
3 8.1% 8.8
4 12.1% 10.4
5 12.2% 12.0
6 12.4% 13.6
7 14.5% 15.3
8 17.7% 16.9
9 NA 18.5
10 NA 20.1
11 NA 20.1
-------
IV.53
METHOD FOR ESTIMATING DISCOUNTED STREAM OF AVOIDED EMISSIONS
In estimating the discounted streams of avoided emissions,
the following procedure was used:
1. We assumed 87.3 million cars and 31.9 million light-duty
trucks designed to use unleaded fuel would be on the
road in 1988. Using vehicle survival rates, we estimated
that approximately 82% of the total light-duty vehicles
(cars and trucks) would be equipped with catalytic conver-
ters in 1988.
2. We then estimated, from EPA survey data, the proportion
of these vehicles expected to misfuel for the first time
in 1988. These estimates are presented below:
First-Time** Total # of
Model Total # of Vehicles* Misfueling First-Time
Year (thousands of vehicles) Rates Misfuelers
1988-89 15,250 .055 839
1987 14,892 .017 253
1986 14,235 .016 228
1985 13,809 .016 221
1984 12,957 .016 207
1983 11,030 .016 176
1982 8,951 .017 152
1981 8,900 .016 142
1980 8,924 .016 143
1979 10,160 .016 163
3. The projected number of misfueling vehicles was multiplied
by an estimate of the number of miles driven per vehicle in
1988. The average annual mileage factors were specific both
for classt and age of vehicle. These calculations were
* Automobile data from MVMA Facts and Figures '83; light-duty
truck data from R.L. Polk & Co.
**No first-time misfueling was assumed for vehicles older than
model year 1979.
t Automobiles, light-duty trucks between 0 and 6000 Ibs., and
light-duty trucks between 6000 and 8500 Ibs.
-------
IV. 54
repeated for every year of the assumed 20-year life of the
vehicle, with the fleet size being diminished annually
according to contemporary scrappage rates. Annual mileage
per vehicle was adjusted according to vehicle age. In this
way, model year- and vehicle class-specific estimates for
total miles driven after misfueling in 1988 were derived,
with the final year investigated being 2007 (when the 1988
model year fleet was assumed to be retired). This forecast
the mileage from each misfueled cohort in each future year.
4. Each future year's mileage etimates were discounted back to
1988 at a 3% discount rate. The total discounted miles
driven (in billions of miles) are shown below:
Model
Year Automobiles LDTi* LPT?*
1979 3.45 1.00 0.45
1980 4.21 0.72 0.33
1981 4.54 0.82 0.38
1982 5.64 0.97 0.42
1983 7.34 1.52 0.75
1984 10.06 2.06 0.96
1985 12.48 2.43 1.14
1986 14.74 2.85 1.34
1987 18.44 3.68 1.73
1988-89 67.90 13.94 6.57
The 1988-89 numbers are large because of the 5.2% rate of
misfueling in the first year and because 15 months of auto
sales are included in the last category.
5. Discounted future mileage was multiplied by the excess emis-
sions factor developed by EPA (1983d), measured in grams of
pollutant per mile (see below). This yielded total discounted
future emissions of conventional pollutants as a result of
misfueling in 1988.
Model Year CO NOX HC
1981-1988 11.07g/mi 0.71g/mi 1.57g/mi
1979-1980 17.65g/mi ~ 2.67g/mi
6. This result was divided by 1x10^ to calculate the total metric
tons of discounted emissions shown in Table IV-3.
= Trucks between 0 and 6000 Ibs.
LDT2 = Trucks between 6000 and 8500 Ibs.
-------
IV. 55
POSSIBLE BIASES IN AVOIDED EMISSIONS
A. Reasons our Emissions Estimates may be too Low
0 1982 misfueling rates, based on EPA surveys, may be too low
for reasons explained on page IV-3ff and in the 1979 EPA
survey. Most notable is that vehicle inspections for misfuel-
ing were voluntary and in some areas, the rates of drivers
refusing inspections were very high (up to 44% in one non-I/M
area).
0 We held misfueling rates constant over time, but these rates
may be increasing over time.
0 Vehicles are lasting longer than previously; therefore, our
vehicle survival rates may be too low. With longer lifetimes,
older, dirtier, misfueled vehicles would be in operation longer,
and the stream of excess emissions would extend farther into
the future. Furthermore, we retired each cohort after its
twentieth year of operation (with about 7% remaining in the
twentieth year).
0 If vehicles are not well-maintained, excess emissions factors
for misfueling will be higher.
B. Reasons our Emissions Estimates may be too High
0 We assumed that pollution control equipment would be effective
past the five-year manufacturer's warranty, for the life of
the vehicle. Some EPA data indicated that this was true if
vehicles were not misfueled or tampered with.
-------
IV. 56
References
Adams, R., and McCarl, B., "Assessing the Benefits of Alternative
Oxidant Standards on Agriculture: The Role of Response Information,"
completed for U.S. EPA, September 1983.
Anderson, E., et al.f "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, 79:
46-50, 1973.
American Petroleum Institute, Unpublished Carcinogenicity Study
on Unleaded Gasoline in Vapor in Fischer 344 Rats and B6C3F1
Mice, 1982.
Brown, C., "Equalizing Differences in the Labor Market," Quarterly
Journal of Economics, Vol. 94, 1980.
Chock, D., et al., "Effect of NOX Emission Rates on Smog Formation
in the California South Coast Air Basin", Environmental Science
and Technology, Vol. 15, No. 8, August 1981.
Council on Environmental Quality, Environmental Quality - 1980,
Government Printing Office, December 1980.
Crocker, T., and Vaux, H., "Some Economic Consequences of Ambient
Oxidant Impacts on a National Forest," completed for U.S. EPA,
Office of Policy Analysis, August 1983.
Detels, R., et al., "The UCLA Population Studies of Chronic
Obstructive Respiratory Disease," American Journal of Epidemiology,
Vol. 109, 1979.
Drinkwater, B., et al., "Air Pollution, Exercise, and Heat Stress,"
Archives of Environmental Health, 28: 177-181, 1974.
Durand, D., Stable Chaos, Morristown, N.J., General Learning
Corporation, 1971.
ETA Engineering, Inc., "Assessment of Benefits from New Source
Performance Standards for Volatile Organic Compounds," completed
for U.S. EPA, Office of Policy Analysis, September 1983.
Evans, et al., "Ozone Measurement from a Network of Remote Sites,"
Journal of the Air Pollution Control Association, Vol. 32, No. 4,
April 1983.
Faucett Associates, Draft Report: Review and Critique of Previous
OMSAPC Cost-Effectiveness Analysis, March 1983.
-------
IV. 57
Ferris, B.r Jr., "Health Effects of Exposure to Low Levels of
Regulated Air Pollutants," Journal of the Air Pollution Control
Association, Vol. 28, No. 5, May 1978.
Freeman, A.M., Ill, Air and Water Pollution Control; A Benefit-
Cost Assessment, New York: John Wiley and Sons, 1982.
Gerking, S.; Stanley, L.; and Weirick, W., "An Economic Analysis
of Air Pollution and Health: The Case of St. Louis," report to
U.S. EPA, Office of Policy Analysis, July 1983.
Glasson, W., "Effect of Hydrocarbons and NOX on Photochemical
Smog Formation Under Simulated Transport Conditions," Journal
of the Air Pollution Control Association, Vol. 31, No. 11,
November 1981.
Goldstein, E., moderator, Photochemical Air Pollution, inter-
departmental conference sponsored by the Department of Medicine,
University of California, School of Medicine, Davis, California,
1982.
Hammer, D., et al., "Los Angeles Student Nurse Study. Daily
Symptom Reporting and Photochemical Oxidants," Archives of
Environmental Health, Vol. 28, 1974.
Hasselblad, V., and Svendsgaard, D., "Reanalysis of the Los
Angeles Student Nurse Study," U.S. EPA, Health Effects Research
Lab, Research Triangle Park, NC, August 1975.
Heck, W., et al., "A Reassessment of Crop Loss from Ozone,"
Environmental Science and Technology, Vol. 17, No. 12, 1983.
Hoppenbrouwers, T., et al., "Seasonal Relationships of Sudden
Infant Death Syndrome and Environmental Pollutants," American
Journal of Epidemiology, Vol. 113, No.6, 1981.
Horvath, S., et al., "Maximal Aerobic Capacity at Different
Levels of Carboxyhemoglobin," Journal of Applied Physiology,
38: 300-303, 1975.
Kinosian, J., "Ozone Precursor Relationships from EKMA Diagrams,"
Environmental Science and Technology, Vol. 16, No. 12, 1982.
Kopp, R., and Vaughan, W., "Agricultural Benefits Analysis:
Alternative Ozone and Photochemical Oxidant Standards," Resources
for the Future, June 30, 1983; and discussion with the authors.
Lave, L., and Seskin, E., Air Pollution and Human Health,
Baltimore, Johns Hopkins University Press, 1977.
-------
IV. 58
Lebowitz, M., et al., "The Effect of Air Pollution and Weather
on Lung Function in Exercising Children and Adolescents,"
American Review of Respiratory Diseases, Vol. 109, 1974.
Leighton, J.; Shehadi, A.; and Wolcott, R., "The Aggregate
Benefits of Air Pollution Control," prepared for U.S. EPA,
Office of Policy Analysis, by Public Interest Economics Foundation,
Washington, B.C., June 1983.
Linn, W., et al., "Human Respiratory Effects of Heavy Exercise
in Oxidant-Polluted Ambient Air," American Review of Respiratory
Disease, Vol. 123, No. 4, 1981.
Lippmann, M., et al., "Effects of Ozone on the Pulmonary Function
of Children," in: Lee, S., et al., eds, The Biomedical Effects
of Ozone and Related Photochemical Oxidants, Princeton Scientific
Publishers, Inc.; Princeton, N.J.: Advances in Modern Environmental
Toxicology, V: 423-46; 1983.
Makino, K., and Mizoguchi, I., "Symptoms Caused by Photochemical
Smog," Japan Journal of Public Health, Vol. 22, No. 8, 1975.
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.
Portney, P., and Mullahy, J., "Ambient Ozone and Human Health:
An Epidemiological Analysis," completed for U.S. EPA, Office of
Air Quality Planning and Standards, September 1983.
Raven, P., et al., "Effect of Carbon Monoxide and Peroxyace-
tylnitrate on Man's Maximal Aerobic Capacity," Journal of Applied
Physiology, Vol. 36, 1974.
Seigneur, C.; Saxena, P.; and Roth, P., "Preliminary Results of
Acid Rain Modeling," submitted at a Specialty Conference on
Atmospheric Deposition sponsored by the Air Pollution Control
Association, November 7-10, 1982, Detroit, Michigan.
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.
Thaler, R., and Rosen, S., "The Value of Saving a Life: Evidence
from the Labor Market," in Household Production and Consumption,
ed. N. E. Terleckyj, New York, Columbia University Press, 1976.
U.S. Department of Agriculture, Agricultural Statistics, 1980,
Washington, D.C., U.S. Government Printing Office, 1980.
-------
IV. 59
U.S. Department of Health and Human Services, Public Health
Service, "Blood Carbon Monoxide Levels in Persons 3-74 Years of
Age: United States, 1976-80," Advance Data, No. 76, March 17,
1982.
U.S. Department of Health, Education, and Welfare; Public Health
Service, Vital and Health Statistics Series 10, No. 96, "Limitations
of Activity and Mobility Due to Chronic Conditions", 1973a.
U.S. Department of Health, Education, and Welfare; Public Health
Service, Vital and Health Statistics Series 10, No. 84, "Prevalence
of Selected Chronic Respiratory Conditions: United States - 1970,"
1973b.
U.S. EPA, "Response to Public Comments on EPA's Listing of Benzene
Under Section 112 and Relevant Procedures for the Regulation of
Hazardous Air Pollutants," Office of Air Quality Planning and
Standards, forthcoming Summer, 1984a.
U.S. EPA, "VOC/Ozone Relationships from EKMA," memo from Warren
Freas (Air Management Technology Branch) to Alan McGartland
(Benefits Branch), January 27, 1984b.
U.S. EPA, Draft Revised Air Quality Criteria for Ozone and Other
Photochemical Oxidants, Office of Research and Development,
1983a.
U.S. EPA, "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), May 16,
1983b.
U.S. EPA, Motor Vehicle Tampering Survey - 1982, National Enforcement
Investigations Center, Office of Enforcement, April 1983c.
U.S. EPA, 1982 NCLAN Annual Report, Environmental Research Lab,
Corvallis, Oregon, 1982a.
U.S. EPA, National Air Pollutant Emissions Estimates, 1940 -
1980, Monitoring and Data Analysis Division, January 1982b.
U.S. EPA, Air Quality Criteria for Nitrogen Oxides, Office of
Research and Development, 1982c.
U.S. EPA, The Costs of Controlling Emissions of 1981 Model Year
Automobiles, Office of Mobile Source Air Pollution Control, June
1981.
-------
IV. 60
U.S. EPA, "Regulatory Impact Analysis of the National Ambient
Air Quality Standards for Carbon Monoxide," Office of Air Quality
Planning and Standards, 1980.
U.S. EPA, Carcinogen Assessment Group, "The Carcinogen Assessment
Group's Final Report on Population Risk to Ambient Benzene Exposure,"
January 10, 1979.
U.S. EPA, Air Quality Criteria for Ozone and Other Photo-chemical
Oxidants, Office of Research and Development, April 1978.
Whittemore, A., and Korn, E. L., "Asthma and Air Pollution in
the Los Angeles Area," American Journal of Public Health, Vol. 70,
1980.
Zagraniski, R.; Leaderer, B.; and Stolwuk, J., "Ambient Sulfates,
Photochemical Oxidants and Acute Adverse Health Effects: An
Epidemiologic Study," Environmental Research, Vol. 19, 1979.
-------
CHAPTER V
BENEFITS OF REDUCING LEAD: CHILDREN WITH HIGH BLOOD LEAD
Our analysis of the 1988 health benefits of reducing lead
is presented in two parts. This chapter describes the benefits
associated with reducing the number of children with blood lead
levels above 30 ug/dl. Currently the Centers for Disease Control
(CDC) considers this level the criterion for lead toxicity when
combined with FEP levels of 50 ug/dl or more (CDC, 1978). Chapter
VI addresses the benefits for children with blood lead below 30
ug/dl. We have focused our analysis on children. Although adults
experience adverse effects from lead, these effects generally
occur at higher lead levels than in children and, as described
below, children have higher body lead burdens for the same exposure,
Blood lead levels above 30 ug/dl are associated with adverse
cognitive effects, anemia, kidney damage, hypertension, and other
pathophysiological consequences. Several of these effects have
only been documented at blood lead levels well above 30 ug/dl.
In the next chapter, we discuss the physiological and cognitive
effects that occur below 30 ug/dl.
It should be noted that while our discussion of reducing
lead emissions has focused on airborne lead, airborne lead is
eventually deposited in the environment on land, water, buildings,
etc. Children, as a class, are most at risk from all sources of
lead — inhaled or ingested. Small children who crawl and "mouth"
objects and hands are especially likely to ingest lead. Fetuses
and young children are more vulnerable than the population as
a whole. The absorption and retention rates, and the partitioning
of lead in hard and soft tissues all contribute to the fact that
-------
V.2
children possess greater lead body burdens for a given exposure.
Children have also been shown to display a greater sensitivity
to lead toxicity, and their inability to recognize symptoms may
make them especially vulnerable. In the late 1970s data indi-
*»
cated that well over 10% of black children had blood lead levels
above 30 ug/dl (Mahaffey et al., 1982a).
The first section of this chapter presents the evidence
•
supporting the relation of blood lead to gasoline lead. Next,
two aspects of the effects of lead exposure that we have been
able to monetize are discussed. First, we assessed the costs
associated with medical treatment and follow-up care for the
children who have elevated blood lead levels. Second, we
considered the cognitive and behavioral impacts of high blood
lead levels (above 40 ug/dl) in children. This chapter also
presents the methodology by which we predicted the changes in
the number of children above 30 ug/dl (and other thresholds) as a
function- of changes in the total amount of lead used in gasoline.
The monetized benefits of reducing the number of children
with blood lead levels above 30 ug/dl in 1988 fall into two
categories: 1) the avoided costs of testing for and monitoring
children with elevated blood lead levels, and medically treating
children with very elevated levels; and 2) the costs associated
with the cognitive effects of lead exposure above 30 ug/dl.
The benefits computed in this chapter are a linear function
of the reduction in the number of children above 30 ug/dl of blood
lead. For each policy option, we estimated these reductions by
-------
V.3
by using the techniques discussed in the statistical methodo-
logies section (V.E.). The results are shown below in Table V-l.
TABLE V-l
Reduction in Number of Children above 30 ug/dl in 1988
Low-Lead Option All Unleaded
43,000 45,000
V.A. The Relationship between Gasoline Lead and Blood Lead
Several recent articles have shown persuasively that blood
lead levels for a given age group 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 V-l on the next page shows this
relationship graphically.
In 1982, Billick 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 his earlier results.
A recent paper by EPA's Office of Policy Analysis (Schwartz,
Janney, and Pitcher, 1984), presented the results of a study
concerning the relationship between blood lead levels and gasoline
lead. Three different data sets were employed for this analysis,
-------
V.4
FIGURE V-l
CHILDREN'S BLOOD-LEAD LEVELS VARY
DIRECTLY WITH LEVELS OF LEAD IN GASOLINE
35
BLACK
HISPANIC —
GASOLINE LEAD
"w 30
a.
LU
LU
-J
§ 25
LU
_J
Q
20
LU
CC
|" 15
O
LU
\ /
6.0
5.0
C5
>
C/5
O
r-
Z
m
m
Q
G)
4.0 f
3
O
3.0 2
1 J^ J I I I I
. I.
I I I
1970 1971 1972 1973 1974 1975
QUARTERLY SAMPLING DATE
1976
-------
V.5
including the second National Health and Nutrition Examination
Survey (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. External estimates of environmental lead
from other sources clearly indicated that paint and other dietary
lead were not the primary sources of the observed decline in
blood lead levels. (An earlier paper on this subject was
presented by Schwartz at the International Conference on Heavy
Metals in the Environment (1983) in Heidelberg, West Germany.)
Critics questioned whether the association between gasoline
lead and blood lead levels could have been due to the sequence
in which the NHANES II survey team moved from one site to the
next. Repeated tests of the results, using variables for each
location, indicated that specific locations or geographic regions
did not confound the relationship between blood lead and gasoline
lead. Furthermore, performing separate regressions for urban
areas, rural areas, adults, children, blacks, and whites indi-
cated these factors could not be substituted for gas lead to
explain the changes in blood lead levels.
A third study, by Annest et al. (1983) of the U.S. Public
Health Service, also used data from the NHANES II, finding that
the only reasonable explanation for the decline in blood lead
levels was the decline in the amount of lead in gasoline.
Finally, the Draft Lead Criteria Document cited two studies
(Fachetti and Geiss, 1982; and Manton, 1977) which, by introducing
-------
V.6
tetraethyl lead with a different isotope ratio into gasoline,
were able to directly measure the contribution to blood lead
levels from gasoline. Both of these papers showed that gasoline
accounted for about 5-10 ug/dl of blood lead.
V.B. Medical Benefits of Reducing High Blood Lead Levels
To estimate the benefits from reduced numbers of children
with blood lead levels above 30 ug/dl, we assumed that all
children whose lead levels were elevated above this limit would
receive follow-up medical attention and/or immediate medical
treatment. Unfortunately, however, many -- perhaps even most --
children with elevated lead levels are not detected, although
their lives and health are adversely affected. It should be
noted, therefore, that children with blood lead levels greater
than 30 ug/dl who go untreated bear a burden which we valued
equal to the cost of follow-up and/or treatment. Furthermore,
the dollar estimate of average medical management costs (testing
and monitoring) assumed a prototypical method of determining
treatment; these costs were representative of the costs associ-
ated with treatment and follow-up techniques in general use,
although the exact procedures may vary.
We have distinguished between three basic follow-up and
treatment categories: children with blood lead levels over 30
ug/dl but with free erythrocyte protoporphyrin (FEP) levels
below 50 ug/dl, children in the Centers for Disease Control's
-------
V.7
(CDC) lead toxicity category II, and children in CDC categories
III and IV.* Treatment and follow-up practices may differ for
each.
For children with over 30 ug/dl of blood lead, given FEP
levels below 50 ug/dl, we assumed one follow-up blood test and
the associated overhead costs. From the regression presented by
Piomelli et al. (1982) on the probability of elevated FEP versus
blood lead, we estimated that 60% of the children over 30 ug/dl
had FEP levels above 50 ug/dl. However, Mahaffey and coworkers
(1982) cited data from the CDC screening program indicating that
75% of all screened children over 30 ug/dl of blood lead also
had FEP levels above 50 ug/dl. Since the CDC sample was both
larger and more representative of the entire nation than that
used by Piomelli et al., we have placed slightly greater emphasis
on this result and assumed 70% of the children above 30 ug/dl
would be classified lead toxic by CDC. To further estimate the
fraction of the most severely lead toxic children in categories
III and IV, we examined the results of the CDC screening program
for 1977-81. They showed a relatively constant 33% of all lead
toxic children were in categories III or IV; the remaining 67%
* CDC classifies children as "lead toxic" if they have blood lead
levels above 30 ug/dl and FEP levels above 50 ug/dl. Children
between 30-49 ug/dl blood lead and 50-109 ug/dl FEP are category
II. Category III is either children > 50 ug/dl blood lead and
<250 ug/dl FEP or > 110 ug/dl FEP and 30-50 ug/dl blood lead.
Children >50 ug/dl blood lead and >250 ug/dl FEP or children
>70 ug/dl blood lead are category IV.
-------
V.8
must, therefore, have been in category II.* Therefore, for all
children with blood lead levels above 30 ug/dl, 30% have FEP
levels below 50 ug/dl, 47% are in category II, and 23% are in
CDC's categories III and IV.
We assumed that children in category II would receive six
regularly scheduled blood tests, and that about half of these
children would also have a county sanitarian visit their homes to
evaluate possible sources of lead exposure. (The CDC screening
program data indicated that 65% of the homes of all lead toxic
children were visited. Assuming all category III and IV children
had home visits, this suggested a 50% rate for category II.)
We also assumed, per the CDC's recommendations, that
detailed medical histories, physical examinations, and an assess-
ment of nutritional status would be performed by a physician.
For children in categories III and IV, we assumed a three-day
hospital stay for testing, and that a county sanitarian would
* The quarterly prevalence data for the percent of all lead toxic
children who were category III or IV are:
Year 1st Quarter 2nd Quarter 3rd Quarter 4th Quarter
1977
1978
1979
1980
1981
32%
31%
30%
29%
30%
32%
32%
33%
34%
33%
34%
35%
37%
38%
31%
31%
34%
35%
(source: Morbidity and Mortality Weekly Reports)
Approximately 7,000 children per quarter were found to be lead
toxic. Note that the percent in categories III and IV was
highest in the 3rd quarter (July, August and September) when
gasoline lead emissions are highest.
-------
V.9
inspect their homes. On the basis of CDC recommendations, it was
assumed these children would have six monthly follow-up blood tests
after discharge, and another six quarterly follow-ups. Finally,
we assumed the children in these severely afflicted categories
would receive a neurological examination, and that one-third
of them would undergo provocative ethylenediaminetetraacetic
acid (EDTA) testing and chelation therapy to remove lead from
the body.
EPA has estimated the cost of blood tests to be $30. We
assumed (1) a one-time administrative overhead charge of $50 for
every child who entered the system, (2) a physician's cost of $50
per visit, and (3) a home inspection by a county sanitarian cost
of $60, including overhead. We have used 1982 hospital costs per
adjusted inpatient day from the Department of Health and Human
Services publication Hospital Statistics (1983). Having regressed
the trend in these costs since 1972 against the GNP deflator, we
obtained an average rate of increase in real costs and projected
costs per day in 1988 (including lab tests, etc.) to be $425 (in
1983 dollars). For each of the major hospitalization stages,
physician's costs of $250 have been estimated, including a neuro-
logical work-up. Using these figures we estimated the average
medical costs for children over 30 ug/dl to be $950 per child.
Table V-2 shows the medical cost savings in 1988 of reducing
the number of children over 30 ug/dl. Because we have not esti-
mated welfare losses (such as work time lost by parents), the
adverse health effects of chelation (such as the removal of help-
ful minerals), or such non-quantifiables as the pain from the
-------
V.10
treatment, our estimate of the benefits of reduced treatment is
conservative. As mentioned above, we have taken these medical
costs as a measure of avoidable damage for all the incremental
cases of lead toxicity, whether detected or not.
TABLE V-2
Medical Cost Savings in 1988
(1983 dollars)
Low-Lead All Unleaded
$41 million $43 million
Our analysis has assumed 30 ug/dl as the criterion for
defining when a child is at risk for undue lead exposure or toxi-
city and may require pediatric care. (This is the criterion now
used by CDC, in conjunction with elevated FEP levels.) If that
criterion is lowered, greater numbers of children would receive
medical management, thereby increasing the medical expense savings
from lowering blood lead levels. This is not an unlikely event,
as the Draft Lead Criteria Document (1983) indicated:
"If, for example, blood lead levels of 40-50 ug/dl in
"asymptomatic" children are associated with chelatable lead
burdens which overlap those encountered in frank pediatric
plumbism, as documented in one series of lead exposed children,
then there is no margin of safety at these blood levels for
severe effects which are not at all a matter of controversy.
Were it both logistically feasible to do so on a large scale
and were the use of chelants free of health risk to the
subjects, serial provocative chelation testing would appear
to be the better indicator of exposure and risk. Failing
this, the only prudent alternative is the use of a large
safety factor applied to blood lead which would translate to
an "acceptable" chelatable burden. It is likely that this
blood lead va]ue would lie well below the currently accepted
upper limit of 30 ug/dl, since the safety factor would have
to be large enough to protect against frank plumbism as well
as more subtle health effects seen with non-overt lead
intoxication." (Chapter 13, p. 15) (emphasis added)
-------
V.ll
For example, the estimated number of children whose blood lead
levels would be expected to drop from above 25 ug/dl to below
this figure as a result of an all unleaded gasoline scenario in
1988 is 150,000, over three times the figure of 45,000 used in
the present analysis, derived from a criterion of 30 ug/dl blood
lead.
V.C. Cognitive and Behavioral Effects
Many studies have noted neurological effects in children
with elevated blood lead levels. De la Burde and Choate's results
(1972, 1975) have been summarized by the Draft Lead Criteria
Document as showing persisting neurobehavioral deficits at blood
lead levels of 40-60 ug/dl. In the 1975 study, seven times as
many high lead children were found to have repeated grades in
school or were referred to school psychologists as low lead
control children. The control children were drawn from the same
clinic population and were matched for age, sex, race, parents'
socioeconomic status, housing density, mother's IQ, number of
children below six in the family, presence of father in the
family, and mother working.
Although the children examined in the work of de la Burde
and Choate included some with blood lead levels between 30 and
40 ug/dl, the issue of whether the cognitive deficits occurred
at those levels was not clear from the results. Several addi-
tional studies cited in the Draft Criteria Document, as well as
recent work by Odenbro et al. (1983), indicated a significant
association between these blood lead levels and neurological/
-------
V.12
cognitive effects in children (e.g., Needleman et al., 1979;
McBride et al., 1982; Yule et al., 1981; Yule et al. , 1983;
Smith et al., 1983; Yule and Lansdown, 1983; Harvey et al.,
1983; and Winneke et al., 1982). All of these studies generally
support these results, even though individually the probability
of a false positive was not always less than 5% and the possibility
of uncontrolled covariates existed. Nevertheless and despite
the difficulties with the specific studies, the combined weight
of the evidence showed that cognitive deficits occurred at blood
lead levels over 30 ug/dl, with the work of de la Burde and
Choate indicating that the most serious damage may be associated
with blood lead levels over 40 ug/dl. (A more detailed analysis
of the studies is presented in Section VI.E.)
V.D. Estimating Avoided Costs of Compensatory Education
The evidence for cognitive effects of lead in children
above 30 ug/dl is fairly strong, and the studies by cle la Burde
and Choate gave direct evidence of poorer classroom performance
by children with higher lead levels, particularly those over
40 ug/dl. It also showed that the cognitive effects remained
three years later.
To value avoiding such cognitive damage, we could posit
that children involuntarily exposed to enough lead to make them
seven times more likely to be forced to repeat a grade should
be given enough supplementary educational assistance to bring
their school performance back to what it otherwise would have
been. Therefore, we could use the cost of such compensatory
education as a proxy for the avoided cost.
-------
V.13
Of course, it is probably impossible to completely restore
these high lead children's performance. Therefore, lifetime
work and production may be affected. However, tutoring, reading
teachers, school psychologists, and the like can help improve
their achievement in school.
Given the finding of at least a three year persistence in
the cognitive effects of lead, we assumed that the cost of
correcting these cognitive effects would be at least three years
of compensatory education. We judged that de la Burde and
Choate's exposed population corresponded to children in CDC's
categories III and IV, as well as some category II children.
From January 1977 until mid-1981, one-third of the children
identified by the CDC screening program as being lead toxic
(over 30 ug/dl blood lead and 50 ug/dl FEP) were in CDC cate-
gories III and IV, the more severe categories of lead toxicity.
From this we estimated that one-third of the children above
30 ug/dl would fall in the category of those severely enough
affected to need compensatory education to recover their
previously expected performance levels. Children with lower
internal lead/FEP levels were assumed not to need this
education. Therefore, an average of one year of compensatory
education would be required per child with blood levels over
30 ug/dl to compensate for the deficits.*
* We assumed that the number of person-years of compensatory
education divided by the number of children would be about
one. In other words, if one-third of these children require
three years of compensatory education, there is an average
of one year of education for each of the children.
-------
V.14
As a rough approximation of the expense of such compensa-
tion, we have used the cost of part time special education for
children who remain in regular classrooms. The staff of the
Department of Education's Office of Special Education Programs
(OSEP) felt this level of effort was appropriate for these
children. According to a report written for OSEP (Kakalik et
al., 1981), a child needing this form of compensatory education
incurred additional costs of $3,064 per year in 1978 dollars,
or $4,290 in 1983 dollars (using the GNP deflator). This
figure was quite close to Provenzano's (1980) estimate of the
special education costs for non-retarded lead exposed children.
We applied these costs to our estimate of the number of children
who would fall below 30 ug/dl as a result of our two hypothetical
policy options to calculate the benefits in 1988 of reducing
their cognitive damage. These benefits are displayed below in
Table V-3.*
TABLE V-3
Benefits in 1988 of Reduced Cognitive Losjsej;
(1983 dollars)
Low-Lead All Unleaded
$184 million $193 million
* We have not assumed that all these children would be classi-
fied as having learning disabilities, but rather that they
would all perform worse than they would have otherwise.
Thus, compensatory education costs were used as a proxy for
the cost of restoring their cognitive functioning.
-------
V.15
V.E. Statistical Methodologies
In this section we present the regression results and
forecast procedures that underlie the estimates of the changes
in the number of children at risk of elevated blood lead levels
used in this and the subsequent chapter. First, however, we
review the evidence of a relationship between blood lead levels
and the amount of lead in gasoline. Following this, we describe
the data base used for our regression work and the regression
results. Finally, there is a discussion of our forecasting
procedures and a consideration of the implications of forecast-
ing prevalence rather than incidence.
V.E.I. The NHANES II Data
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
selected according to 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 were 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
-------
V.16
blood samples, including all children six months through six
years and half of those between seven through 74 years. The
non-respondent rate for blood samples was 39% and did not
correlate with race, sex, annual family income, or degree of
urbanization.* A study of the potential non-response biases
indicated that this was not a significant problem (Forthofer,
1983).
Lead concentrations in the blood of sampled persons and
control groups were determined by atomic absorption spectro-
photometry using a modified Delves Cup micro-method. Specimens
were analyzed in duplicate with the average of the two measure-
ments 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 twenty NHANES II blood samples. No temporal trend was
evident in the blind quality control measurements.
The NHANES II data did, however, display a marked relation-
ship between blood lead and gasoline lead, as is shown in Figures
V-2 and V-3. A similar pattern existed between average blood lead
levels for black children in Chicago and lead use in local Chicago
gasoline during the same period. This is evident in Figure V-4.
* Because children were less likely to respond, they were double
sampled, and 51% of the children did not provide blood for lead
determinations in the NHANES II data set. The weights used to
adjust the data to the national population accounted for both
the oversampling and under-response of the children.
-------
V.17
FIGURE V-2
o in
!2" _J
UJ
o
00
CO
~x m
Q^H
LJ
C^)
AVERAGE BLOOD LEAD LEVELS (micrograms/deciliter)
CO
I
CM
i
o
o
o
00
T-
O
O
t
O3
~T-
O
~T-
O
O
00
CD
r-»
o>
CO
en
TOTAL LEAD USED PER 6 MONTH PERIOD (1000 tons)
-------
AVERAGE NHANES II BLOOD LEAD LEVELS VS.
LEAD USED IN GASOLINE PRODUCTION
17 -i
a:
a
a
15-
o
2
13-
LU
6 12-
y 11 H
o
o
§1°
LL!
o
S 9
LJ
8-1
T
40
O
c
»
ra
<
i
OO
CORRELATION COEFFICIENT « 0.95
(P < .0001)
60
50 60 70 80 90 100
TOTAL LEAD USED PER 6 MONTH PERIOD (1000 TONS)
110
-------
V.19
FIGURE V-4
AVERAGE BLOOD LEAD LEVELS FOR BLACK CHILDREN IN CHICAGO
AND GASOLINE LEAD IN CHICAGO
r—-r—-r-—i 1 1
BILLIONS
OF GRAMS
Legend
.1 O BLOOD LEAD
A GASLEAD
CALENDAR YEAR QUARTERS
-------
V.20
V.E.2. Reduction in Number of Children Below Critical Thresholds
The NHANES II data was used to estimate both linear
regressions relating blood lead to gas lead and the percentage
of children who would be expected to have blood leads above
various thresholds.
To estimate these percentages, logistic regressions were
performed separately for white and black children to see how
the odds of having blood lead levels above a 30 ug/dl threshold
varied with gasoline lead. These regressions were performed on
data from individual children. The dependent variable was the
natural log of the odds of being above the threshold while the
independent variables were various demographic factors* and
gasoline lead. The original selection of demographic factors
for consideration was based on linear regressions on individual
blood lead levels, discussed in detail in the paper by Schwartz,
Janney, and Pitcher (1983).**
To predict how the number of children above each threshold
would change as the amount of lead in gasoline was reduced, a
mechanism was needed to forecast the distribution of blood lead
* The demographic variables were selected by backwards stepwise
elimination. We also used a procedure that maximized R2 for
any given number of variables and a procedure that minimized
the difference between Cp and the number of independent vari-
ables. They produced the same model as backwards elimination
** The regressions were all performed on individual data using
the SAS procedure SURREGR to estimate the coefficients.
SURREGR is a special procedure designed to estimate the
variances in regressions using clustered stratified samples.
Demographic control variables were eliminated by backwards
stepwise elimination until all the remaining variables were
significant at the 95% confidence level. See Schwartz,
Janney, and Pitcher (1983) for further detail.
-------
V.21
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 distribution
could be used to determine the percentage of the population
above any 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.
If the distribution X is normal with mean u and standard
deviation s (X:N (u,s)), then Y = exp X is log-normal with a
mean of a and a standard deviation of b where
a = exp (u + 1/2 s2} and
b = exp (2u + s2) (exp (s2) -1)
Further, if eg and VQ are percentiles of the log-normal and its
corresponding normal distribution, we have
eg = exp (u + vg s)
We used the logistic regressions to estimate eg in equation (2)
and the SURREGR regressions to estimate a in equation (1) which
yielded
(1) a = exp (u + 1/2 s2)
(2) 6g = eXp (U + Vg S)
Solving these equations for u and s produced a quadratic equation
-------
V.22
0 = (In (eg) - In (a)) - vg s + .5s2
which had the solution s = -v + (v 2 - 2 (In (e) - In (a)) 5.
Then u = ln(a) - 1/2 s2. Only the smaller root yielded sensible
values for u and s. 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 chose to use a logistic equation to estimate the percent-
age of children over 30 ug/dl to control for problems of multiple
sources of exposure. If we had simply used the regressions
explaining the mean and assumed a constant standard deviation,
we would have predicted that removing lead from gasoline would
have resulted in there being no children above 30 ug/dl. This
seemed unreasonable since paint, food, and water are known alter-
nate sources of lead, and are sometimes associated with high
blood lead levels.
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 number of children above
various thresholds represented the sum of distributions for each
age category. The regression results are shown in the Appendix
to this chapter.
For children from six months to seven years of age, we used
logistic regressions for the percent above 30 ug/dl 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 work,
-------
V.23
Because our figures included only children under age
thirteen and no adults, these results significantly under-
estimated the benefits from reduced lead levels in the entire
population. To make our predictions, we used our projections of
lead used each year under the various scenarios. These are
shown on Table V-4 below.
TABLE V-4
Estimated Lead Used for Gasoline in 1988*
(metric tons per day)
Base Case Low-Lead All Unleaded
97.6 6.1 0
*Computed using gasoline demand in Table II-l and assuming
1.1 g/gal, 0.1 g/gal, and 0 g/gal.
From this, we could also predict changes in mean blood
lead levels. These are shown in Table V-5 below.
TABLE V-5
Changes in Mean Blood Lead in 1988**
for Black and White Children aged 5 and under
(micrograms per deciliter)
Base Case Low-Lead All Unleaded
Incremental Projected Incremental Projected
Decline Level Decline Level
White 7.93 1.93 6.00 2.13 5.80
Black 14.31 1.72 12.59 1.89 12.42
**Derived using gasoline lead values in Table V-4 and the
regressions presented in the Appendix to this chapter.
-------
V.24
To compute the number of children above the various
thresholds in 1988, we needed estimates of the population at
different ages. These were produced by linearly interpolating
the Bureau of the Census population projections (mid-range
forecast) for 1985 and 1990 and are shown in Table V-6.
TABLE V-6
1988 Population Projections*
Ages Blacks Whites
1/2 year - 7 years 4,573,000 23,259,000
8 years - 13 years 3,797,000 16,528,000
*Bureau of the Census, 1982
These results also tend to underestimate the extent of the
problem because the NHANES II survey, upon which our model was
based, omitted children under six months of age. This was
especially significant because we are learning that the damage
to this infant population from elevated blood lead levels may be
more severe than that of older children.
V.E.3. Incidence Versus Prevalence
Our predicted decreases in the number of children above a
given threshold were for a specific point in time; our costs were
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.
-------
V.25
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 raised the difficult epidemiological issue of preva-
lence versus incidence. Prevalence means 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
the integral of the incidence of cases times their duration, or
prevalence is approximately incidence times average duration.
This issue became 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 wanted to estimate may 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) found fairly stable
blood lead levels in individual children with high levels in
Chicago. For these children, levels remained high for more than
a few days, usually for months or years. However, if the average
elevation of blood lead was six months, the actual number of
-------
V.26
.children affected in a year would be twice the average prevalence
for the year. This obviously would affect our benefit estimates.
Because 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 remained ele-
vated 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
effects. To do this we looked at 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.
As our first source we looked at the CDC screening program.
This program screened approximately 100-125,000 children per
quarter of the year to detect lead toxicity. Approximately
6-7,000 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 gaso-
line 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 previous
cases.
-------
V.27
We have also 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 manage-
ment 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 were generally followed for several
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 number of cases of children above 30 ug/dl
by as much as a factor of two.
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 was 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.
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
-------
V.28
relationship to gasoline lead, and whether any seasonal dummy
variables were significant in explaining the large observed
seasonal variations in blood lead. Schwartz, Janney, and Pitcher
(1983) found that the lag structure of average blood lead levels'
dependence on gasoline lead extended about 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 (450,000 children screened) over a ten
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 and coworkers analyzed
this 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.
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 number of children above 30 ug/dl during 1988 is
low, as is our estimate of avoided medical expenses.
V.E.4. Assessing the Accuracy of our Forecasting Procedures
The NHANES II data we used to estimate the regressions in our
forecasting model corresponded with a range of gasoline lead usage
from 193 to 550 metric tons per day. The options we are consider-
ing have gasoline lead usage rates of 97.6, 6.1, and 0 metric
tons per day,* values that are below the range associated with
*See Table V-4.
-------
V.29
the NHANES II data set. An obvious concern was the applica-
bility of results gathered from NHANES II data to the policy
options under consideration.
To examine the hypothesis that the gasoline lead coefficient
changed at lower gasoline lead values, we regressed blood lead
levels for white children for just the last two years of the
NHANES II period. (During this time period both blood and gaso-
line lead levels were lowest.) The gasoline lead coefficient
changed by 3%, which was not significantly different from that
derived for the full period. For blacks, the small sample size
did not allow separate estimates for different periods. While
there is no reason to believe that the functional form of the
dependence was different for blacks and whites, we used an alter-
nate procedure that did not require a reduction in sample size
to check the linearity of blood lead's dependence on gasoline
lead for blacks.
The log of blood lead was regressed against age, income,
sex, and degree of urbanization, and against the log of gasoline
lead. This produced a model in which blood lead was a function
of gas lead to some power B, where B was the coefficient of log
(gaslead) in the regression. We performed this regression to
estimate the power law of blood lead's relation to gasoline.
Had we just regressed log (blood lead) on log (gaslead),
we would have artificially forced blood lead to be zero when
gasoline lead was zero. While studies of the bones of ancient
Nubians indicated that prehistoric lead levels were essentially
-------
V.30
trivial, studies of remote populations today (e.g., in the
Himalayas) suggested that general environmental contamination
produced 3-5 ug/dl blood lead levels in the absence of any gaso-
line or local industrial emissions (Piomelli et al., 1980).
Since background levels in the United States were likely
to be higher than those of remote populations, we tested models
with intercepts ranging from 6 to 10 ug/dl. They yielded
exponents ranging from 0.82 to 1.08 for the dependence of black
children's blood lead levels on gasoline lead. The model with
the highest R2 had an intercept of 8 ug/dl and an exponent of
0.98. The fact that the exponent values that fit the data best
were very close to unity implied that blood lead is equal to
(gaslead)l — i.e., the relationship was linear.
Finally, we tested a model where blood lead was related to
the square root of gasoline lead, and it did not fit as well as
the linear model. We believe, therefore, that the assumption
that blood lead levels in black children are a linear function
of gasoline lead is reasonable.
V.F. Conclusion
We have monetized two health related effects of reducing
the amount of lead in gasoline. The projected benefits in 1988
estimated from these two effects alone are presented in Table
V-7.
-------
V.31
TABLE V-7
Monetized Benefits of Reduced Numbers
of Children Above 30 ug/dl Blood Lead Level in 1988
(millions of 1983 dollars)
Low-Lead All Unleaded
$225 million $236 million
There are additional effects that we have not monetized
which have also been associated with blood lead levels above
30 ug/dl.
0 We have not estimated the value of adverse effects in
adults or infants under six months. As we mentioned
above, new data have indicated that fetuses and newborn
infants may be most vulnerable to lead effects.
0 Non-neurological effects such as kidney damage, anemia,
and other medical problems have not been assessed.
0 Behavioral problems have not been addressed.
(These can adversely alter attention span or
take more overt forms such as serious behavioral
abnormalities, perhaps affecting the education
of other children in the classroom.)
0 Finally, we mentioned certain non-quantifiable
problems earlier such as the pain associated
with some medical procedures, lost work (and
leisure) time by family members, and the potential
long-term social costs from the lower employment
-------
V.32
potential of individuals whose learning
abilities have been impaired. As a result,
the health benefits presented in Table V-7
are likely to be much less than the real
cost to society.
-------
V.33
TECHNICAL APPENDIX TO CHAPTER V
In addition to the regressions shown in Schwartz, Janney,
and Pitcher (1983), we have used the regressions presented in
this appendix for our forecasts. We used the following variables
in these regressions:
Variable Name
Gaslead
Poor
Age 1
Age 2
Age 3
Age 4
Age 5
Age 6
Age 7
Income 1
Income 2
Teen
Description
Lead used in gasoline, in hundreds of
of tons/day, lagged one month
1 if Income 1 (see below); 0 otherwise
1 if age >_ 6 months and < 2 years;
0 otherwise
1 if age _> 2 years and _£ 3 years;
0 otherwise
1 if age _> 4 years and j£ 5 years;
0 otherwise
1 if age _> 6 years and <_ 1 years;
0 otherwise
1 if age _>. 8 years and _< 9 years;
0 otherwise
1 if age _>. 10 years and <_ 11 years;
0 otherwise
1 if age _> 12 years and _< 13 years;
0 otherwise
1 if family income < $6,000;
0 otherwise
1 if family income < $15,000 and
> $6,000; 0 otherwise
1 if age _> 14 years and < 18 years;
0 otherwise
Male
1 if gender is male; 0 if female
-------
V.34
Variable Name
Teen Male
Adult Male
Small City
Rural
Drinker
Heavy Drinker
Northeast, Midwest,
South
Education
Description
1 if gender is male and age _> 14 years;
and < 18; 0 otherwise
1 if gender is male and age _> 19 years;
0 otherwise
1 if residence is in city with population
_< 1,000,000; 0 otherwise
1 if residence is a rural area as defined
by the Bureau of the Census; 0 otherwise
1 if alcohol consumption is _> 1 drink/
week and _< 6 drinks/week; 0 otherwise
1 if alcohol consumption is >^ 1 drink/
day; 0 otherwise
Are regions of the country as defined
by the Bureau of the Census.
Is 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
-------
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
V.35
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 0.0000
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
-8.1667
0.6331
1.2174
1.4332
1.7168
1.1405
1.2322
0.2160
0.2935
0.7978
0.7415
0.7503
*A11 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 zero otherwise.
-------
V.36
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 Beta Std. Error Chi square P
Intercept
Gaslead
Poor
Age 5
Age 6
-6.0148
0.9786
0.2356
0.6158
0.2397
2.4044
0.4943
0.5289
0.6304
0.6208
6.26 0.0124
3.92 0.0477
0.20 0.6560
0.95 0. 3 28 6
0.15 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 Beta Std. Error Chi square
Intercept
Gaslead
Poor
Age 5
Age 6
-8.9395
1.0674
0.8355
1.4199
1.2041
1.6782
0.3374
0.4883
0.5810
0.5904
28.38
10.01
2.93
5.97
4.16
0.0000
0.0016
0.0871
0.0145
0.0414
Fraction of concordant pairs of predicted probabilities
and responses = 0.710
-------
V.37
SURREGR Regression Results
Whites; children 6 months to 7 years
Dependent variable: individual blood lead levels
Variable
Intercept
Gaslead
Income 1
Income 2
Age 1
Age 2
Age 3
Age 4
Teen
Male
Teen male
Adult male
Small City
Rural
Drinker
Heavy Drinker
Northeast
Midwest
South
Education level
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
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 . 0 29 6
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
—
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
-------
V.38
Whites: 6 months to 13 years
Dependent variable: individual
Variable
Intercept
Gaslead
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
Beta Std
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
. 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
lead levels
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
—
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
-------
V.39
Blacks: 6 months to 7 years
Dependent variable: individual blood lead levels
Variable
Intercept
Gaslead
Income 1
Income 2
Age 1
Age 2
Age 3
Age 4
Teen
Male
Adult male
Drinker
Heavy drinker
Education level
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
Std. Error
2.4116
0.1432
0.2593
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
—
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
-------
V.40
Blacks; 6 months to 13 years
Dependent variable
Variable
Intercept
Gaslead
Income 1
Income 2
Kid
Teen
Male
Adult male
Age 4
Age 5
Age 6
Age 7
Drinker
Heavy Drinker
Northeast
Midwest
: individual
Beta Std
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
South -0.1173
Education level
-0.826
blood lead
. 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
levels
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
—
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
-------
V.41
REFERENCES
Annest, J.L.; Pirkle, J.L.; Makuc, D.; Neese, J.W.,;
Bayse, D.D.; Kovar , M.G. (1983) Chronological trend in
blood lead levels between 1976 and 1980. (Boston) New
England Journal of Medicine. 308: 1373-1377.
Billick, I.H. (1982) Prediction of pediatric blood lead levels
from gasoline consumption (submitted to docket for public,
hearing on lead phase-down proposed rule making, April 15,
1982). Available from: U.S. Environmental Protection
Agency, Central Docket Section, Washington, DC; docket no.
A-81-36; document number IVA.4.
Billick, I.H.; Curran, A.S.; Shier, D.R. (1979) Analysis of
pediatric blood lead levels in New York City for 1970-1976,
Environmental Health Perspectives 31: 183-190.
De la Burde, B.; Choate, M.S., Jr. (1972) Does asymptomatic
lead exposure in children have latent sequelae?
Journal of Pediatrics (St. Louis) 81: 1088-1091.
De la Burde, B.; Choate, M.S., Jr. (1975) Early asymptomatic
lead exposure and development at school age.
Journal of Pediatrics (St. Louis) 87: 638-642.
Fachetti, S.; Geiss, F. (1982) Isotopic lead experiment:
status report. Luxembourg: Commission of the European
Communities; Publication no. EUR 8352 EN.
Forthofer, R.N. (1983) Investigation of nonresponse bias in
NHANES II. American Journal of Epidemiology (Baltimore)
117: 507-515.
Harvey, P.; Hamlin, M.; Kumar R. (1983) The Birmingham blood
lead study. Presented at: annual conference of the
British Psychological Society, symposium on lead and
health. Available for inspection at: U.S. Environmental
Protection Agency, Environmental Criteria and Assessment
Office, Research Triangle Park, NC.
Kakalik, J.S. et al. (1981) The Cost of Special Education, Rand
Corporation (Report No. N-1792-ED).
Mahaffey, K.R.; Annest, J.L.; Roberts, J.; Murphy, M.S. (1982)
National estimates of blood lead levels: United States
1976-1980: association with selected demographic and
socioeconomic factors. New England Journal of Medicine
307: 573-579.
-------
V.42
Manton, W.I. (1977) Sources of lead in blood: identification
by stable isotopes. Archives of Environmental Health
32: 149-159.
McBride, W.G.,; Black, B.P.; English, B.J. (1982) Blood lead
levels and behaviour of 400 preschool children. Medical
Journal Aust. 2: 26-29.
Needleman, H.L.; Gunnoe, C.; Leviton, A.; Reed, R.; Peresie, H.;
Maher, C.; Barrett, P. (1979) Deficits in psychological
and classroom performance of children with elevated dentine
lead levels. New England Journal of Medicine 300: 689-695.
Odenbro, A.; Greenberg, N.; Vroegh, K.; Bedreka, J.; Kihlstrom,
J.E. (1983) Functional disturbances in lead-exposed
children. Ambio 12: 40-44.
Piomelli, S.; Corash, L.; Corash, M.B.; Seaman, C.; Mushak, P.;
Glover, B.; Padgett, R. (1980) Blood lead concentrations in
a remote Himalayan population. Science 210: 1135-1137.
Piomelli, S.; Seaman, C.; Zullow, D.; Curran, A.; Davidow, B.
(1982) Threshold for lead damage to heme synthesis in urban
children. Proceedings of the National Academy of Science
U.S.A. 79: 3335-3339.
Provenzano, G. (1980) The social cost of excessive lead exposure
during childhood. In: Low Level Lead Exposure,
H.L. Needleman, editor. Raven Press1, New York.
Schwartz, J.D.; Janney, A.; Pitcher, H. (1983) The relationship
between gasoline lead and blood lead. (submitted for
publication)
Smith, M.; Delves, T.; Lansdown, R.; Clayton, B.; Graham, P.
(1983) The effects of lead exposure on urban children: The
Institute of Child Health/Southhampton study. London,
United Kingdom: Department of the Environment.
U.S. Department of Commerce, Bureau of the Census (1982)
Projections of the population of the United States, 1982 to
2050. Current Population Reports series P-25, No. 922.
U.S. Department of Health and Human Services (1983) Hospital
Statistics.
U.S. Environmental Protection Agency (1983) Review Draft Air
Quality Criteria for Lead, Volumes III and IV (Research
Triangle Park, North Carolina).
-------
V.43
Winneke, G.; Hrdina, K-G.; Brockhaus, A. (1982) Neuropsychological
studies in children with elevated tooth-lead concentrations.
part I: pilot study. Int. Arch. Occup. Environ. Health 51:
169-183.
Yule, W.; Lansdown, R. (1983) Lead and children's development,
recent findings. Presented at: International Conference:
Management and Control of Heavy Metals in the Environment;
September; Heidelberg, West Germany.
Yule, W.; Lansdown, R.; Millar, I.E.; Urbanowicz, M.A. (1981)
The relationship between blood lead concentrations,
intelligence and attainment in a school population: a
pilot study. Dev. Med. Child Neurol. 23: 567-576.
Yule, W.; Urbanowicz, M.A.; Lansdown, R.; Millar, I. (1983)
Teachers' ratings of children's behaviour in relation to
blood lead levels. British Journal of Developmental Psychology
(in press).
-------
CHAPTER VI
BENEFITS OF REDUCING LEAD: CHILDREN
WITH MODERATE BLOOD LEAD
In this chapter, we discuss the known pathophysiological
effects of lead that may occur in children below 30 ug/dl of
blood lead. As noted in the introduction to Chapter V, we
focused our analysis on children because, on the whole, they are
more sensitive and vulnerable to lead than adults. We discuss
the hematological and neurological effects in particular, as well
as the expected change in the number of children at potential
risk of those effects under our policy alternatives.
Our benefit estimates present only changes in the numbers of
children at risk of these effects; we have not associated any
dollar values with reducing these exposures. Although no monetary
estimate of adverse effects is provided, the social costs (to the
individuals affected and society as a whole) associated with even
low blood lead levels is probably substantial.
The scientific literature presents evidence of a continuum
of biological effects associated with lead across a broad range
of exposure. Even at low exposure levels, the Draft Lead Criteria
Document (EPA, 1983) found that:
biochemical changes, e.g., disruption of certain enzymatic
activities involved in heme biosynthesis and erythropoietic
pyrimidine metabolism, are detectable. With increasing lead
exposure, there are sequentially more pronounced effects on
heme synthesis and a broadening of lead effects to additional
biochemical and physiological mechanisms in various tissues,
such that increasingly more severe disruption of the normal
functioning of many different organ systems becomes apparent.
In addition to impairment of heme biosynthesis, signs of
disruption of normal functioning of the erythropoietic and
nervous systems are among the earliest effects observed in
response to increasing lead exposure. At increasingly higher
exposure levels, more severe disruption of the erythropoietic
-------
VI.2
and nervous systems occurs; and other organ systems are also
affected so as to result in the manifestation of renal
effects, disruption of reproductive functions, impairment
of immunological functions, and many other biological effects.
At sufficiently high levels of exposure, the damage to the
nervous system and other effects can be severe enough to
result in death or, in some cases of non-fatal lead poisoning,
long-lasting sequelae such as permanent mental retardation.
(Chapter 12, pages 1-2)
While the hematopoietic, nervous, and renal systems are generally
considered to be the most sensitive to lead, lead has a signifi-
cant impact on reproductive and developmental processes as well.
Table VI-1 presents blood lead levels from the second
National Health and Nutrition Examination Survey (NHANES II).
TABLE VI-1
BLOOD LEAD LEVELS OF PERSONS
Aged 6 Months - 74 Years in the United States 1976-80*
(percent in each cell)
<10 ug/dl
All Races
all ages 22.1%
6 months-5 years 12.2%
6-17 years 27.6%
18-74 years 21.2%
Wh i t e
all ages 23.3%
6 months-5 years 14.5%
6-17 years 30.4%
18-74 years 21.9%
Black
all ages 4.0%
6 months-5 years 2.7%
6-17 years 8.0%
18-74 years 2.3%
10-19
ug/dl
62.9%
63.3%
64.8%
62.3%
62.8%
67.5%
63.4%
62.3%
59.6%
48.8%
69.9%
56.4%
20-29
ug/dl
13.0%
20.5%
7.1%
14.3%
12.2%
16.1%
5.8%
13.7%
31 .0%
35.1%
21.1%
34.9%
30-39
ug/dl
1.6%
3.5%
0.5%
1.8%
1.5%
1.8%
0.4%
1.8%
4.1%
11.1%
1.0%
4.5%
40-69
ug/dl
0.3%
0.4%
0.0%
0.3%
0.3%
0.2%
0.0%
0.4%
1.3%
2.4%
0.0%
1.8%
*Table 1 Advance Data #79 May 12, 1982, from Vital and Health
Statistics, National Center for Health Statistics (Supplemental
Exhibit 4.) NOTE: These results were produced after adjusting
the data for age, race, sex, income, degree of urbanization,
probability of selection, and non-response to the NHANES survey
-------
VI.3
VI.A. Pathophysiological Effects
Pathophysiological effects are found at blood lead levels
well below 30 ug/dl, particularly in children. There is evidence
that blood lead levels under 30 ug/dl 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 (Angle
et al., 1982). Hernberg and Nikkanen (1970) found 50%
of ALA-D inhibited at about 16 ug/dl.
2. Elevated levels of zinc protoporphyrin (ZPP or PEP) in
erythrocytes (red blood cells) at about 15 ug/dl. This
probably indicates a general interference in heme
synthesis throughout the body, including interference
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 EEC
patterns (Otto et al., 1981, 1982) which begin to occur
at about 15 ug/dl, and which appear to persist over a
two-year period. Also, the relative amplitude of syn-
chronized 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 from
about 15 ug/dl on (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 (Draft Lead Criteria Docu-
ment, p 13-34? 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 12-32).
6. Inhibition of vitamin D pathways, which has been detected
as low as 10 to 15 ug/dl (Rosen et al., 1980, 1981; Mahaffey
et al., 1982b). Further, as blood lead levels increased,
the inhibition became increasingly severe, and the lead
absorption rate was enhanced.
-------
VI. 4
These levels approximate the lowest observed effect levels to
date and do not necessarily represent the affirmative findings
of a threshold.
The types of specific effects listed above as occurring at
blood lead levels below 30 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 production, and (d) lead-induced
perturbations in central and peripheral nervous system functioning,
The medical significance of such effects is not yet fully under-
stood. But current knowledge regarding the deleterious nature
of such effects and the 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. Drawing on material in Chapter
12 of the Draft Lead Criteria Document, we discuss the potential
consequences of these findings below.
Heme, in addition to being part of hemoglobin, is the obli-
gatory prosthetic group for diverse hemoproteins in all tissues,
both neural and non-neural. Hemoproteins play important roles
in generalized functions such as cellular energetics, as well as
in more specific functions such as oxygen transport and detoxifi-
cation of toxic foreign substances (e.g., drug detoxification in
the liver). Available data (on elevated ALA and FEP levels,
inhibited ALA-D, etc.) show clear and significant inhibition
in the heme biosynthetic pathway at low blood lead levels, with
-------
VI.5
statistically significant effects detectable at 10-15 ug/dl.
This heme biosynthetic disturbance may result in the impairment
of many normal physiological processes and/or the reduced reserve
capacity of many cells or organs to deal with other types of
stress (e.g., infectious diseases).
The best known effect of lead on erythrocytic pyrimidine
metabolism is the pronounced inhibition of PY-5-N activity. This
enzyme figures 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).
As noted earlier, the disruption of this function by lead has
been noted at levels of exposure beginning at 10 ug/dl. At blood
lead levels of 30-40 ug/dl, this disturbance is sufficient to
materially contribute to red blood cell lysis (destruction) and,
possibly, decreased hemoglobin production contributing to anemia
(Draft Lead Criteria Document, p 12-27f).
Another serious consequence of lead exposure is the impair-
ment of the biosynthesis of the active vitamin D metabolite,
1,25(OH)2 vitamin D, which is detectable at blood lead levels
of 10-15 ug/dl. Interference with vitamin D production 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.
-------
VI. 6
The reduced uptake and utilization of calcium has two
compounding consequences. There is interference 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
calcium availability. The latter can create a feedback response
further exacerbating the vitamin D production inhibition, reduced
calcium availability, and consequently even greater lead absorp-
tion and greater vulnerability to increasingly more severe lead-
induced health effects (Draft Lead Criteria Document, p 10-32f).
These effects are especially dangerous for young (preschool age)
children who are developing rapidly. These children, even in
the absence of lead, generally are deficient in calcium 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.
The negative correlation between blood lead and serum 1,25-
(OH)2D, the active form of vitamin D, appears to be an example
of lead's disruption of mitochondrial activity at low concentra-
tions. While serum levels of 1,25-(OH)2 vitamin D decreased
continuously as blood lead levels increased from an apparent
threshold of 10-15 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
-------
VI.7
levels remained unchanged. This indicated that lead may inhibit
renal 1-hydroxylase, the enzyme that converts the precursor to
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 mitochondrial systems detected
at 15 ug/dl has uncompensated consequences.
If cytochrome P-450 is being inhibited at the low levels
that the reduced renal 1-hydroxylase activity suggests, we must
consider the possibility that other physiological functions
related to cytochrome P-450 may also be disrupted. In particular,
reduced P-450 content has been correlated with impaired activity
of the liver detoxifying enzymes, aniline hydroxylase and amino-
pyrine demethylase, which help to detoxify medications, hormones,
and other chemicals.
While cytochrome P-450 inhibition has been found in animals,
and in humans at higher lead levels, this damage has not yet been
detected in children at low blood lead levels (i.e., 10 to 15
ug/dl). The disruption of vitamin D biosynthetic pathways at
these levels is suggestive of an effect.
The elevation of ALA levels is another indication of lead's
interference in mitochondrial functioning. In vitro studies have
shown that ALA can interfere with several physiological processes
involved in the GABA-ergic neurotransmitter system, including a
possible role as a GABA-agonist. There appears to be no thres-
hold concentration for ALA at the neuronal synapse below which
presynaptic inhibition of GABA release ceases. We do not know
at what blood lead level detectable interference with brain
-------
VI. 8
functions by ALA begins in-vivo*, nor the level at which the
neural interference becomes "critical". However, 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).
Lead in the brain is likely to enhance brain ALA concentrations
because neurons are rich in mitochondria, the subcellular site
of ALA production. Blood ALA elevations begin to be detectable
at 15 ug/dl of blood lead. Since ALA is a neurotoxin, th.e poten-
tial implications for brain function are disturbing. The fact
that EEC patterns also begin to change at this blood lead level
is an additional source of concern.
In addition to the effects of lead on the brain and central
nervous sysem, there is evidence that peripheral nerves are
affected as well. Silbergeld and Adle (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. The Draft Criteria Document
notes:
...[lead causes] a blockade of calcium binding to the
synaptosomal membrane reducing calcium-dependent choline
uptake and subsequent release of acetylcholine from the
nerve terminal. Calcium efflux from neurons is mediated by
the membrane (Na+, K+)-ATPase via an exchange process with
sodium. Inhibition of the enzyme by lead, as also occurs
with the erythroctye....', increases the concentration of
calcium within nerve endings (Goddard and Robinson, 1976).
As seen from the data of Pounds et al. (1982a), lead can
also elicit retention of calcium in neural cells by easy
entry into the cell and by directly affecting the deep
calcium compartment within the cell, of which the mito-
chondrion is a major component. (Section 12.2.3)
-------
VI.9
This disruption of cellular calcium transport may also contribute
to the effects of lead on peripheral nerve conduction velocity.
Landrigan et al. (1976) have noted a significant correlation
between blood lead and decreasing conduction velocity in children
in a smelter community. This effect may indicate advancing
peripheral neuropathy.
VLB. Hematological Effects of Lead
High levels of blood lead are known to produce anemia.
Previously it was an unresolved question whether blood lead
below 30 ug/dl increased the risk of anemia in children. We
addressed this question in two ways. First, we examined the
relationship between blood lead levels and various measures of
anemia, and the inhibition of heme synthesis as evidenced by
elevated free erythrocyte protoporphyrin (FEP) levels. Second,
because FEP is a more stable indicator of a person's lead
exposure over several months than a single blood lead determi-
nation, we also analyzed the relationship between elevated FEP
levels and anemia. We found that blood lead and FEP levels were
associated with increased risks of anemia in children, even
below 30 ug/dl of blood lead.
For this analysis we again used data from the NHANES II
survey. Among the hematological information collected was mean
corpuscular 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 relation-
ship between blood lead levels and the presence of hematological
abnormalities.
-------
VI. 10
VI.B.1. Effects 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 those currently considered to be safe.
Linear regressions were performed of MCV and MCH on blood
lead levels in children, controlling for race, age, income, and
iron status (i.e., the level of iron in their blood). 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 erythrocyte hemoglo-
bin levels. As previous work led us to expect', percent transfer-
rin saturation was a superior control for iron status compared
to serum iron and was used throughout our analysis.
The regressions for both MCV and MCH found blood lead to be
a significant explanatory variable (p < .0001 and .0033, respec-
tively) for the decreases in each.
Because small decreases in MCV and MCH are of unknown sig-
nificance, we also analyzed the probability of children having
abnormally low MCV or 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 both in mean cell volume
being low (MCV < 80 femptoliters [fl], p < .0001), and in mean
-------
VI.11
cell volume being seriously low (MCV < 74 fl, p < .0001). Blood
lead levels were also significantly associated (p < .023) with
the percent of children having MCH less than 25 pico grams (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 abnormal MCV using only those children
whose blood lead levels were less than 25 ug/dl. The regression
coefficient for blood lead was unchanged and significant
(p < .014). Thus1, blood lead levels under 25 ug/dl were associated
with increased risks of microcytic anemia.
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% saturated for children in the NHANES II
survey) and children with transferrin saturation levels one
standard deviation below average (13.6%). The results are shown
in Figure VI-1. Note that at 25 ug/dl of blood lead almost 10%
of the children with average iron levels and 17% 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% 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%). Since logistic regressions
-------
VI.12
Figure VI-1
PERCENT OF CHILDREN WITH MCV BELOW 74
(Age 6 Months to 8 Years)
25
20 -
15
0)
53 10
Q.
With Transferrin Saturation
One Standard Deviation Below Average
With Average
Transferrin Saturation
10
15
20
25
30
35
Blood Lead Level (/tg/dl)
-------
VI.13
gave the same results when we used only children with blood lead
levels under 25 ug/dl, and since the 95% confidence limits on the
relative risk did not include 1.0, these results showed increased
risks of hematological abnormalities in children at blood lead
levels of 20 ug/dl and below.
VI.B.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 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. (The
authors have not yet published these findings.) 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,
we tested several alternative specifications. We considered a
linear model, we examined 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 (Trans-
ferrin saturation)6, and a model where FEP was proportional to
-------
VI.14
(Blood lead)Bl and (Transferrin saturation)B2. The model that
that fit best was exp(Blood lead) times (Transferrin saturation)^.
We examined the possibility of different additive intercepts in
this model and found the highest correlation coefficient and
F-statistic for a zero additive constant. This model suggested
the relationship: PEP = 36.73 (Transferrin saturation) -0.11684
exp(0.01183 Blood lead).*
While others have found sex differences in the response of
FEP to blood lead, sex was not a significant variable in any of
our models for children. This was probably a result of the fact
that the sex difference in the response of FEP to blood lead is
smaller in children. We also suspect that the differences in
adults are due predominantly to sex differences in iron status,
which we controlled for directly.
We also investigated the relationship between the probability
of elevated FEP levels and blood lead, and verified previous
findings. Again using NHANES II data, we performed logistic
regressions 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, and obtaining a better fit with
blood lead. The 50 ug/dl FEP level is considered to indicate
severe enough interference with heme processes that medical
attention is usually required even when not coupled with elevated
blood lead levels.
Again, we checked to see whether the relationship between the
risk of elevated FEP and blood lead held at lower blood lead levels,
* Transferrin saturation is expressed in tenths of a percent.
-------
VI.15
repeating the regression only for children with blood lead levels
under 30 ug/dl. Using maximum likelihood analysis, blood lead was
again extremely significant (p < .0001). The coefficient of blood
lead for the low group was .178 +_ .04 compared to .175 H^ .018 for
the regression with all blood lead levels, a trivial difference
between the two cases. This indicated that the risk of seriously
elevated PEP levels was strongly related to blood lead, even at
blood lead levels well below the currently defined safety level.
Piomelli and coworkers1 studies have suggested a threshold
for lead-induced increases in PEP levels of about 15 ug/dl. Taking
17.5 ug/dl of blood lead as our reference level, our regression
predicted that the relative risk of PEP levels over 50 ug/dl was
1.55 (1.42-1.70 at 95%) at 20 ug/dl of blood lead, and was 3.73
(2.55-4.89 at 95%) at 25 ug/dl of blood lead. This was true
across all transferrin saturation levels.
VLB.3. The Relationship Between PEP Levels and Anemia
Since the average lifetime of erythrocytes is approximately
120 days, a single blood lead level measured concurrently with
hematocrit levels, MCV, and MCH cannot adequately evaluate the role
of lead in the impairment of red cell production. Such a single
measurement is a poor proxy for the blood lead levels over the
previous 120 days, as these levels may not have been constant.
By contrast, FEP, once created, remains in red cells for their
lifetime. While FEP levels are affected by iron status as well
as blood lead, using iron status as an independent variable
along with FEP restricts FEP to principally being a surrogate for
-------
VI. 16
average blood lead levels when studying its association with
anemia. Because FEP levels are exponentially associated with
blood levels, log(FEP) was used as a proxy for lead exposure over
the relevant period.
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 < .0001 in all cases) to hematocrit
levels, hemoglobin levels, and MCV. We then performed logistic
regressions on the probability of abnormal levels of hematocrit,
hemoglobin, and MCV as a function of log(FEP), with the same controls,
They also showed that FEP was an excellent predictor (p < .0001)
of the probability of abnormally low levels of all three indicators.
Again, we repeated our regressions using only children with FEP
values of less than 33 ug/dl, and FEP was still very significant
(p < .0001). The coefficients 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.
FEP levels of less than 33 ug/dl are generally associated
with blood lead levels under 30 ug/dl. Figure VI-2 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 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
-------
VI. 17
Figure VI-2
PERCENT OF CHILDREN WITH ANEMIA
By Age and Race at Average Transferrin Saturation Levels
48
42
36
c 30
o
U
"5 24
•*-•
c
0)
o
(5
Q.
18
12
Black
2—6 yrs.
White
0
12-5
0.5 — 2 yrs.
20 25 30 35
40
45 50
FEP Level (jtg/dl)
-------
VI.18
hematocrit levels of less than 33% for ages 0.5-2, less than 34%
for ages 2-6, and less than 35% for ages 6-8 — the minimum normal
range levels recommended by the Journal of Pediatrics (1977).
These definitions are supported by the work of Yip et al. (1981).
Figure VI-2 shows that as FEP levels increase from 20 ug/dl
to 50 ug/dl, an additional 20% of children aged 2-6 years would
develop anemia at normal iron levels. Our earlier regressions
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 interference
in heme synthesis, it cannot be viewed as the cause of these
abnormal hematocrits. The causal association must be with what-
ever produced the excess FEP. As the data portrayed in Figure
VI-2 were for normal iron levels, the anemia appeared to be the
outcome of the lead exposure underlying the FEP values.
In summary, blood lead levels below the currently defined
"undue lead exposure" range of 30 ug/dl (and, indeed, even below
25 ug/dl) seem to be associated with increased incidence of anemia
in children and increased interference with heme synthesis pro-
ducing elevated levels of free erythrocyte protoporphyrin. This
suggests that both the levels of blood lead and FEP used in the
current Centers for Disease Control definition of undue lead
exposure may be inadequate to protect children from the risk of
anemia. In addition, the reduced mean cell volumes and the lower
hematocrits again indicate that lead's effect on heme synthesis has
uncompensated effects at levels below 30 ug/dl. This further
-------
VI. 19
strengthens the case for considering elevated FEP levels, which
mark lead's interference with normal body activity, as a patho-
physiological effect.
VI.C. Fetal Effects
A growing concern in the public health community is that
the most sensitive population for lead exposure is not children,
but fetuses and newborn infants. This concern is supported by
both animal studies and, recently, human data.
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 post-natal 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 (3-4 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, 1977; Bull et al.,
1979) have shown that the prenatal exposure of rats to 0.2% lead
chloride in the mother's drinking water markedly reduced the
cytochrome C content in the cerebral cortex, and possibly
produced an uncoupling of the electron transport chain in the
-------
VI.20
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).
Human data are scarcer. Needleman et al. (1984) have analy-
zed data from over 4,000 live births at Boston Women's Hospital
and found an association between some congenital anomalies and
umbilical cord blood lead levels. Holding other covariates
constant, the relative risk of a child's demonstrating a minor
malformation at birth increased by 50% as lead levels increased
from 0.7 ug/dl to 6.3 ug/dl (the mean cord lead level). This
increased an additional 50% 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 preliminary analysis by Needleman and coworkers (1984)
also found an association between increasing cord lead levels
and deficits in the child's subsequent perfermance on the Bayley
development scales, after controlling for covariates. Again,
the cord lead levels in this study were very low.
Finally Erickson et al. (1983) found lung and bone lead
levels in children who died from Sudden Infant Death Syndrome
were statistically significantly higher than in children who
died of other causes, after controlling for age.
VI.D. Neurological Effects
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 documented.
-------
VI.21
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 (Draft Lead Criteria Document, Section 12.3.4).
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 adenosine triphosphate (ATP) energy
for the entire cell. Given the high energy demands of neurons,
selective damage to the nervous system seems plausible.
Paralleling these cellular or biochemical effects were
electrophysiological changes indicating the perturbation of
peripheral and central nervous system functioning observed in
children with blood lead levels of approximately 15 ug/dl (Otto
et al., 1981, 19S2; Uenignus et al., 1981). These included
slowed nerve conduction velocities, as well as persistent
abnormal EEC patterns. Aberrant learning behavior has been
-------
VI.22
noted in rats with blood lead levels below 30 ug/dl (Draft
Criteria Document, Section 12.4.3.1.3). This behavior evidenced
both reduced performance on complex learning problems and signs
of hyperactivity and excessive response to negative feedbacks
(Winneke, 1977, 1982).
Finally, the cognitive effects of lead in children showed
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
somewhat lower blood lead levels, de la Burde and Choate (1972,
1975) found lesser but still significant cognitive effects,
including a 4-5 point difference in mean IQ and reduced attention
spans. Several studies discussed in more detail later in this
chapter have found smaller effects at lower blood lead levels.
The precise biological mechanisms connected with these effects
are not yet clearly defined.
While some of these effects have only been observed at higher
blood lead levels, in animals, or in vitro, they all showed 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
-------
VI.23
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 those larger scale
studies with what has been discovered on the molecular and cellu-
lar levels.
VI.D.I. Cognitive and Behavioral Effects
Many studies have noted neurological effects in children
with elevated blood lead levels. A brief discussion of these is
presented in Section V.B. of this paper, concentrating on those
examining the effects of blood lead levels above 30 ug/dl. In
this section, we will examine the effects below 30 ug/dl.
VI.D.I.a. Assessment of the Relationship Between IQ or
Cognitive Function and Low Blood Lead Levels
The answer to the question of whether the relationship between
blood lead and cognitive performance extends to levels below 30
ug/dl is tremendously important. If IQ is affected at blood lead
levels below 30 ug/dl, the benefit of reducing lead emissions is
very large because of the many children who would be at risk.
The literature on cognitive effects at low lead levels is
extensive. However, most of the studies have methodological flaws
of varying importance and few display indisputable results con-
cerning the relationship between IQ effects and changes in low
lead levels. The Draft Lead Criteria Document 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.
-------
VI.24
The summary table in Chapter 12 of the Criteria Document
(pp 55-58) indicated that virtually all of the 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 falsely finding an effect due
to chance was more than 5%, 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.) However, because the reported
sample sizes were small, it was not likely that small effects
would have been detected. The consistent pattern in all the
studies of high lead groups doing less well indicated that the
combined evidence of a significant effect was stronger than the
evaluations of the individual studies suggested.
In developing a better test for the existence of a specific
effect, we limited the studies we examined for two reasons.
First, because we were interested in low level exposure effects
in the whole population, we used only the six general population
studies. The smelter study by Winneke et al. (1982) was also
included, as blood lead levels appeared to be in the same range
as the general population studies. Second, because we were
interested in general effects, we chose to look only at Full
Scale 10 measures. While not all studies used the same IQ test,
the Full Scale IQ measures employed were close enough to allow
us to compare differences between groups and across studies.
We used the Fisher aggregation procedure (Fisher, 1970,
p.99) to develop an estimate of the combined significance of the
-------
VI.25
observed 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. Unfortunately, as indicated above, they
were not reported where they were larger than 0.05, so we had to
calculate several p-values from the data presented.
For each study we used the standard deviation of the IQ
measure to compute the standard deviation for the difference in
the mean IQs across groups. From the ratio of the 10 difference
to this standard deviation, we could compute a p-value. We
could directly apply this method to the study by Smith et al.
(1983). In this study, one of the best methodologically, all
of the 10 effects were reported as "not significant". However,
when we computed the p-values, we found that the p-value was
0.051 when comparing high and low lead groups for the Full Scale
IQ.* Similar computations for the Verbal and Performance IQs
produced p-values of 0.068 and 0.105, respectively.**
* W. Yule, in a personal communication at the the International
Conference on Heavy Metals in the Environment (Heidelberg,
September 1983), said that a recomputation paying more attention
to round-off and computational errors found a one-tailed
p-value of less than 0.05.
**The mean IQs for the low and high groups given in Smith et al.
(1983) were quoted with 95% confidence intervals. For the
sample size (145, 155) for these groups we can assume normality.
[The sample size is taken from Table 13 of Smith_ et al. (1983).]
Thus, for the low group, 2.0 IQ poin_ts = -1.96_si ^nd^ for
the high group, 1.9 IQ points = 1.96s2, where BI and 82 were
the standard deviations for the low and_the high groups,_
respectively. This implied values for sj of 150.98 and 32 of
145.66. Combining these variances yielded an overall variance
of 148.23. Weighting this by the sum of the inverses of the
sample sizes gave the variance for the difference of the means,
which was 1.978. Taking the square root of these yielded a
standard deviation of 1.407. Dividing the difference between the
high and low group (2.3) by 1.407 produced a normal statistic of
1.635, which has an associated p-value of 0.051 (Bryant, 1966).
-------
VI.26
For other studies, where we could not determine the standard
deviation for the test procedure, we assumed it was equal to 15.
This is the commonly cited standard deviation for IQ, although it
varies slightly from test to test. Because this standard devia-
tion was somewhat higher than the standard deviations in the
studies that reported such values (the study groups were more
homogeneous than the general population), our calculations probably
produced p-values larger than the true p-values.
We used these p-values and the Fisher procedure to compute
a joint probability for the observed results, presented in Table
VI-2. The resulting probability of 0.014 indicates that it was
very unlikely that we could get the observed pattern of results
if there were really no effect. The overwhelming preponderance
of the data (all studies show high lead groups with lower cog-
nitive 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
at least one case (Smith et al., 1983), 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 remaining IQ effects
were regressed on lead in the next step. Many of these covariates
(e.g., parental care, income, IQ) negatively correlate with lead
exposure, and this procedure attributed all of the joint variation
to the nonlead variable.
-------
VI. 27
TABLE VI-2
Computation of Joint P-Value from Epidemiological Studies
of Cognitive Effects from Low Level Lead Exposures in children
Study9
McBride et al.
(1982)
Yule et al.
(1981)
Smith et al.
(1983)
Yule and Lansdown
(1983)
Harvey et al.
(1983)
Winneke et al.
(1982a)
Joint P-Value
Needleman et al.
(1979)
Sample
Control
100
20
145
80
Sizes
Exposed
100
21
155
82
Total of 189
26
for Studies:
100
26
P(xf2
58
Blood
Control
0.5-9
7-10
7-12
N/A
> 25.10)
Internal
(ug/dl)
Exposed
19-30
17-32
13-24
N/A
= .014
Lead Levels
Teeth (ppm)
Control Exposed IQ Difference
,.6C
< 2.5 8 2.2C
N/A N/A .7<3
2.4 7 5C
< 10 < 20 4C
P-Value -2 In p
.32 2.28
.029 7.08
.051 5.95
.22 3.03
.34 2.15
.10 4.61
25.10
.03 7.01
32.11
Joint p-value including Needleman P(X£4 > 32.02) = .004
a Citations refer to Draft Lead Criteria Document, October 1983
b Peabody Picture Vocabulary IQ Test
c Welchsler Intelligence Scale for Children-Revised
d British Ability scales
-------
VI.28
Another study (Harvey et al., 1983) had IQ measurements on
131 children but only 71 degrees of freedom in the t-test for
lead. If this study included 59 covariates in the analysis,
such an over specification clearly would bias downward the
significance of lead as well.
We have treated one major study (Needleman et al., 1979)
differently because a recent critique raised questions about the
appropriateness of the p-value reported in the study (Draft
Criteria Document, Appendix 12C). Pending the resolution of that
issue, we have presented our results both with and without this
study. Even when it was omitted, the p-value (.014) was clearly
significant. Recent reanalysis by Needleman using the alternative
specifications for his model suggested by the review committee
still found a significant lead effect. Including this study
would lower the joint p-value to .004.
Figure VI-3 presents an alternative method to evaluate the
results from these studies. For each study the figure shows the
90% confidence interval for the full scale IQ difference between
the high and low lead groups. The mean is represented by the
square, the upper limit by an X, and the lower limit by a +. For
two of the studies the confidence interval does not include zero,
which is computationally equivalent to finding that a one-tailed
hypothesis test would reject a null hypothesis of no effect at
the 5% level. However, as the figure shows, all of the studies
found that the high lead group had a lower mean 10, and that the
-------
VI. 29
FIGURE VI-3
Mean IQ Difference Between High Lead Groups
and Controls, Adjusted for Socioeconomic Factors
(90% Confidence Intervals)
3 -T
2 -
§
-v i -
•5
£ -1-
0 _
U ^
! -3 -
£ -4-
§ -l-
£ -7-
_,
a -d -
q -10 -
b
b -11 -
3 -12 -
£ -13 -
-14 -
1
i :
l
i
i
%
, '
; ;
i
i
/
\
i
' •
1 1
i
-15 -1 1 1 1 1 i i '
etBalde Yule Smith Yule and Harvey Winneke Needleman
e<,.pft", et al. et al. Lansdown et al. et al. et al.
UV»/; (1981) (1983) (1983) (1983) (1982) (1979)
i = Mean
X« Upper Confidence Limit
+= Lower Confidence Limit
-------
VI.30
range of effects consistent with all of them was a loss of one
to three and a half IQ points.
Therefore, we accepted the implication of the joint proba-
bility computation that there was an association between cognitive
deficits and differences in lead exposure even at low levels.
Ignoring such a risk without considering the potential cost of
the error associated with that risk would have been inappropriate
in determining the desirability of implementing a policy.
VI.D.l.b. Policy Implications of Significance Tests
In making policy decisions, we should be concerned with the
cost of not implementing an appropriate policy (false negative)
as well as with the cost of implementing an inappropriate policy.
Policy makers must balance the risks of each type of error,
weighted by its costs. Because of this, even if the p-value for
the joint test had been slightly larger than the arbitrary 0.05
level,* we still would have considered the cognitive effect.
For example, for the Smith et al. study, we computed that the
probability of falsely asserting the null hypothesis was true
* Over the years the significance level of 0.05 has become the
basis for rejecting the scientific ("null") hypothesis. While
adherence to the strict definition of "statistical significance"
has been important in science, it must be remembered that this
p-value is arbitrary and may not be the appropriate sole
criterion for regulatory decision making.
-------
VI.31
was .587.* This false negative rate would be even higher in the
other studies considered here owing to their smaller sample sizes.
The cost of not avoiding small cognitive effects for
millions of children is high. If insisting on a p-value of less
than 0.05 before accepting that a cognitive effect exists means
a substantial risk of a false negative, then the potential cost
of the wrong decision may be too large. In this case, given the
relatively small sample sizes, the small cognitive deficits one
might expect, and the standard deviation of the test procedure,
the risk of a false negative is high. We believe that this
association, coupled with the biochemical studies, animal studies,
and high level effects discussed in the introduction to VI.D.
suggests a causal relationship.
The existence of a large chance of a false negative for
outcomes where costs are potentially high suggested the need
for carefully considering the entire process by which the
validity of the hypothesis was evaluated. In particular, where
the choice of a null hypothesis gives credence to one point of
view, which is not justified given the power of the test, a
hypothesis test may not be appropriate. An alternative method
is to look at confidence intervals around the estimated parameter
* We computed the false positive from the Smith et al. data
as follows. To estimate the p-value, we derived a standard
deviation for the difference of the high and low lead groups
of 1.407. At a 5% chance of rejecting the null hypothesis
when it was true, the normal one-tailed statistic was 1.65.
Therefore, we would reject the null hypothesis only for
differences greater than (1.407) (1.65) = 2.32. If the
difference in the groups were two 10 points, the probabil-
ity of the difference being below 2.32 is given by
p (z < [2.32-2]/ 1.407) = .587.
-------
VI.32
values. As an example, consider the confidence intervals in
Figure VI-3, which shows graphically that elevated lead levels
have a negative effect on 10.
This conclusion, plus the above joint probability estimate
of p = 0.014 for the general population studies, led us to accept
the existence of cognitive effects of low level lead exposure.
VI.D.2. Estimating Avoided IQ Loss Associated with
Reduced Blood Lead Levels
We used two hypotheses to evaluate the extent of 10 loss.
Both were based on the Smith et al. study which used tooth lead
as the measure of lead intoxication, where particular attention
was paid to measuring and controlling for covariates. Their
"high lead group" had teeth with lead levels of 8.0 ug/g or
more, a relatively low cutoff level.
Our first hypothesis, assuming a step function with a thres-
hold, was that a group of children whose teeth lead levels were
above 8 ug/g would have an average 10 2.3 points lower than the
average IQ of children in the control group, whose lead exposure
resulted in tooth lead levels below 2.5 ug/g.
To convert tooth lead to blood lead, we used three methods.
First, we followed Steenhout and Pourtois (1981) and Steenhout
(1982), who used regression analysis to estimate the increase in
tooth lead (t) concentration that would result from various
blood lead (BL) levels over time. Her model was:
Tooth Lead(t') = I q(t)BL(t)dt.
J fcO
-------
VI.33
For adults, q = 0.045, a constant, and the model reduced to:
Tooth lead(t') = q BL At.
At the International Conference on Heavy Metals and the
Environment (September 1983), held in Heidleberg, West Germany,
Steenhout presented additional results. For children, the rate
of tooth lead accumulation per unit of blood lead was much higher
than for adults and appeared to decline exponentially to the adult
level with age. Steenhout's best fit of the data was:
Tooth lead(t') = / [0.045 + 0.2 exp(-t/4.5)ppm]BL(t)dt
where t was measured from Steenhout's "midgrowth stage". Replacing
BL(t) by BL, we could solve for BL. This analysis obtained a EL
of 5.0 ug/dl or less for Smith's low exposure group, and 16 ug/dl
or more for her high group.
Second, we used Winneke's data (Winneke, 1979) which showed
mean blood lead levels equaled 2.5 times mean tooth lead levels.
This gave blood lead levels of 6.25 ug/dl for the control group
and 18 ug/dl for the high group, which was consistent with
Steenhout's results.
Finally, we examined Smith's data on blood lead levels for a
non-random sample of her survey population. These showed blood
lead levels of 11.5 ug/dl for 20 low lead children and 15.1 ug/dl
for her high group. While this yielded about the same results for
the high group, it showed much higher levels for the control group.
We are not sure what caused this discrepancy, although the number
-------
VI.34
of low lead children was very small. In any case, the three dif-
ferent procedures for imputing blood lead suggested a threshold
for 10 loss at about 15 ug/dl.
The second hypothesis assumed that, instead of a step func-
tion with a threshold occurring at 15 ug/dl of blood lead, there
was a linear function relating IQ loss to blood lead level. For
our estimate of the effect, we also used the Smith et al. study
and assumed the low tooth lead group had an average blood lead
level of 3 ug/dl and the high group had an average blood lead
level of 18 ug/dl. We used the estimated blood lead levels
based on Steenhout's procedure and divided the difference in IQ
by the difference in blood lead to yield a slope of 0.15 IQ/ug/dl
of blood lead. Other studies, such as the 1981 study by Yule et
al., had coefficients as high as 0.7 10 points per ug/dl of
blood lead. Using Smith's limited blood lead data would suggest
a slope of 0.64. To be conservative, we have used the 0.15
slope.
We computed the total lost IQ points for several hypotheses,
but did not attach monetary values to the lost IQ.
VI.D.3. Thresholds for Effects of Blood Lead on IQ and the Size
of the Affected Population
In assessing the size of the population at risk, two alterna-
tive hypotheses were again possible. The first was a no thres-
hold model. It assumed that the effect of lead on IQ was a
continuous function, with increasing risk and effect as blood
-------
VI.35
lead levels rose. Under this assumption, adverse effects on
either IQ or behavioral patterns, such as disruptive behavior or
shortened attention span, occurred at lower lead levels and
increased at higher lead levels, and despite individual dif-
ferences, the extent of effect was related to the extent of
exposure.
Alternatively, many people believe that cognitive deficits
from lead exposure occur only above a specific threshold, i.e.,
that blood lead levels below some value will not affect either
intelligence or behavior patterns that may reduce educational
attainment. Several alternative threshold values are possible.
Because there is little dispute concerning cognitive effects
above 30 ug/dl, selecting that blood lead level was one option.
On the other hand, ALA levels are elevated at 15 ug/dl and EEC
patterns also show persistent changes at that level. This evidence
suggested that blood lead levels of 15 ug/dl may be a threshold.
This is buttressed by the Smith et al. study, where we have deter-
mined that children whose exposure averaged above 16 ug/dl had
lower IQ levels than children whose exposure averaged below 5
ug/dl. While the cognitive damage may have occurred at earlier
ages when blood lead levels were higher, the work of Harvey, who
surveyed two year old children in Birmingham, England, indicated
that blood lead levels among two-year-olds averaged 15.6 ug/dl.
This was only slightly higher than the average among Smith's
older children. Furthermore, the study by Yule et al. (1981)
indicated that children with blood leads of 7-10 ug/dl had
higher IQs than those with blood leads of 17-32 ug/dl. Yule and
-------
VI.36
Landsdown (1983) can be taken as supporting this level or even
indicating that the threshold may be somewhat lower.
Alternatively, the study by McBride et al. (1982) showed a
small difference between children above 19 ug/dl and those below
10. This data suggested that the threshold may be around 20 ug/dl
Because all three thresholds (15, 20, and 30 ug/dl) were
possible, we have calculated the number of children potentially
at risk in 1988 for each of the three. These estimates are shown
in Table VI-3.
TABLE VI-3
Decrease in Number of Children in 1988
Above Thresholds for Cognitive Effects
Possible Threshold Low-Lead Option All Unleaded
15 ug/dl 1,475,000 1,552,000
20 ug/dl 476,000 500,000
30 ug/dl 43,000 45,000
As noted, accepting the hypothesis of a cognitive effect at
a given threshold does not imply that all the children above the
threshold are affected or that all below the threshold are free
of the effect.
In addition to computing the number of children at risk, we
estimated the total effect on intelligence in 1988, expressed as
the number of children at risk times the mean change in IQ. Our
estimate of the change in 10 was 2.2 10 points.* We then computed
* The 2.2 10 figure is the difference between the average IQ of
the Smith et al. middle group of children and the average 10
of that study's high lead group. We had found the average
blood lead level of the children below 15 ug/dl was nearer
that of the children in the Smith et al. middle group.
-------
VI.37
the change in person-lQ points (i.e., the number of people at risk
times the average 2.2 10 points lost) as a result of the policy
options. For simplicity, we used the same 2.2 IQ point decrement
for the other two thresholds. The results are shown on Table VI-4,
Finally, assuming there is no threshold, we converted the
changes in mean blood lead levels to changes in 10 using an esti-
mate of the rate of change of 10 per ug/dl. We assumed that the
mean of the IQ change for any child was dependent on the mean
change in blood lead levels. (As shown in Table VI-1', black and
white children had different blood lead levels and this was
considered in our calculation.) The estimated changes in 10
points in 1988 for children aged 6 months to 7 years are shown
in Table VI-4. These were computed using the coefficient of
0.15 IQ/ug/dl derived earlier from the Smith et al. study, from
changes in mean blood lead levels given in Table V-5, and from
population figures in Table VI-6 for children aged 0 to 7.
TABLE VI-4
Possible Change in Person-IQ Points in 1988
as a Function of Threshold Levels for
Children 6 Months to 7 Years
Threshold Low-Lead All Unleaded
15 ug/dl 2,867,000 3,018,000
20 ug/dl 986,000 1,035,000
30 ug/dl 92,000 97,000
No threshold 7,913,000 8,728,000
Because there are so many children at risk, any reasonable
monetary value ascribed to avoiding the loss of one person's IQ
-------
VI.38
point would produce very large benefits for these changes in
person-IQ points. For example, if parents were willing to pay
$100 per IQ point to remove the possibility of such a loss, we
would have estimated benefits for the all unleaded case ranging
from $9.2 to $302 million for the thresholds listed above and up
to $873 million if there were no threshold. Thus, even if only
small changes in IQ are found to be associated with lead
exposure, the large number of children affected would make the
benefits of avoiding such effects extremely large.
VI.E. Estimating the Reduction in the Number of Children at Risk
Reducing or eliminating leaded gasoline will reduce the
number of children at risk for the pathophysiological effects
from elevated blood lead levels. Table VI-5 presents the
decrease in the number of children above the "minimum observed
effect level," or "apparent thresholds," for various health
effects in 1988. In many cases, these apparent thresholds
reflect the limitations of current experimental measurement
techniques and not a finding that no effect exists at lower
levels. Therefore, our estimates are likely to be conservative.
Our estimates of the decreased number of children with abnormal
physiological functioning are based on statistical methods
described in section V.E.
-------
VI.39
TABLE VI-5
Decreased Number of Children (under 14 years old)
Above Apparent Threshold Levels in 1988
Apparent
Medical Effect Threshold Low-Lead All Unleaded
Inhibition of PY-5-N 10 ug/dl 4,257,000 4,486,000
Inhibition of ALA-D 10 ug/dl
Inhibition of vitamin D 10-15 ug/dl
Elevated ZPP 15 ug/dl
EEC changes 15 ug/dl 1,475,000 1,553,000
Elevated ALA levels 15 ug/dl
Inhibition of globin 20 ug/dl 476,000 500,000
synthesis
Even if we take the thresholds in Table VI-5 as true thres-
holds, it is very unlikely that all individuals with blood lead
concentrations above a given threshold will suffer a particular
effect, and it is unlikely that all those below the threshold are
free from the effect. The specific blood lead level at which a
particular effect begins to occur varies from person to person.
In the general population, such variation generally produces an
S-shaped curve of the percent of people with the effect as a func-
tion of blood lead level or other exposure index. In Table VI-5
we approximated the dose-response curve with a step function
instead of a continuous curve; the numbers, therefore, only
roughly estimate the true values.
We also used regressions to predict the distribution of
blood lead levels in 1988. These values are given in Table
VI-6. (Details of how these numbers were calculated are
contained in Section E of Chapter V.)
-------
VI.40
TABLE VI-6
Estimated Distribution of Blood Lead Levels in 1988
(in thousands of
Blood lead
Base Case
Low- Lead
All Unleaded
Blood lead
Base Case
Low- Lead
All Unleaded
<10ug/dl
3,386
4,496
4,559
<10ug/dl
34,608
37,764
37,921
children
Black
10-15
ug/dl
2,588
2,131
2,096
Non-Black
10-15
ug/dl
4,326
2,001
1,884
;.l. Distributional Aspects of
aged 13
15-20
ug/dl
1,191
790
766
15-20
ug/dl
1 ,085
486
458
and under)
20-30
ug/dl
490
267
256
20-30
ug/dl
397
186
177
>30ug/dl
36
17
16
>30ug/dl
52
29
28
Lead Exposure
One feature often overlooked in analyzing the pathophysio-
logical changes induced by lead is the close correlation between
the occurrence of high lead levels and high levels of other
stressors, which, like lead, both have direct adverse effects
and reduce the reserve capacity of the body to deal with environ-
mental insults. When two or more stressors act in concert, the
severity of the adverse impacts increases and makes it much more
likely that the reduced reserve capacity produced by lead will,
in fact, produce adverse consequences.
-------
VI.41
People who have the highest blood lead levels tend to be
children, in general; black children, in particular; and poor
people. Children are often deficient in iron and calcium, the
adverse effects of which are exacerbated by lead. Children's
nervous systems are more sensitive to toxins, and they are just
beginning their cognitive development. Blacks tend to have higher
hypertension rates, which may also be associated with or exacer-
bated by lead (Beevers et al., 1976). Blacks also tend to have
lower vitamin D levels which are further reduced by lead, and
tend to be poor. Poor people usually have a lower level of
vaccination, well baby care, and preventive medicine in general.
Poor people are more likely to be sick and/or malnourished, have
inadequate medical care, and be under greater stress, both physical
(e.g., poor heating and sanitation) and psychological.
Poor people, on average, are less successful in school so
even marginal central nervous system or cognitive effects of lead
may have more serious implications for this group. Many of the
people at high risk of lead exposure have a high risk of experi-
encing these other factors. For them lead effects that would be
sub-clinical in the absence of these other factors may not
be sub-clinical.
VI.F. Conclusion
We examined several different ways to value the benefits in
1988 of reduced lead exposure through reduced use of lead in
gasoline. In Table VI-7 we present a summary of the estimated
-------
VI.42
benefits for children under age fourteen of reducing the adverse
effects resulting from exposure to lead from gasoline.
TABLE VI-7
Summary of the 1988 Health Benefits of
Reducing Low Level Lead Exposure
Low-Lead All Unleaded
Reduction in number of
children (under 14 years
of age) at risk of:
At 10 ug/dl 4,257,000 4,486,000
Inhibition of PY-5-N
Inhibition of ALA-D
At 15 ug/dl 1,475,000 1,553,000
Inhibition of vitamin D
Elevated ZPP
EEC changes
Elevated ALA levels
At 20 ug/dl
Inhibition of globin synthesis 476,000 500,000
Average loss of 2.2 10 points 43,000 to 45,000 to
1,475,000 1,552,000
Percent change in children's
mean blood lead levels:
Whites 24% 27%
Blacks 12% 13%
The size of the populations potentially at risk for the low
level effects preceding overt manifestations of clinical symptoms
of lead poisoning is large. Although we have not attached any
dollar values, the changes that would occur under our two policy
-------
VI.43
options suggest that reducing the pathophysiological effects of
lead exposure would be a significant public health benefit of
reducing lead in gasoline.
-------
VI.44
REFERENCES
Angle, C. R.; Mclntire, M. S.; Swanson, M. S.; Stohs, S. J. (1982)
Erythrocyte nucleotides in children - increased blood lead
and cytidine triphosphate. Pediatric Research 16: 331-334.
Beevers, D. G.; Erskine, E.; Robertson, M.; Beattie, A. D. ;
Campbell, B. C.; Goldberg, A.; Moore, M. R.; Hawthrone, V. M.
(1976) Blood-lead and hypertension. Lancet 2(1): 1-3.
Benignus, V. A.; Otto, D. A.; Muller, K. E.; Seiple, K. J. (1981)
Effects of age and body lead burden on CNS function in young
children. II. EEC spectra. Electronencephalogr. Clin.
Neurophysiol. 52: 240-248.
Bryant, E. C. (1966) Statistical Analyses (2nd ed.) McGraw Hill,
New York.
Bull, R. J.; Lutkenhoff, S. D.; McCarty, G. E.; Miller, R. G. (1979)
Delays in the postnatal increase of cerebral cytochrome concen-
trations in lead-exposed rats. Neuropharmacology 18: 83-92.
Bull, R. J.; McCauley, P. T.; Taylor, D. H.; Croften, K. M. (1983)
The effects of lead on the developing central nervous system
of the rat. Neurotoxicology 4(1): 1-17.
Crofton, J. M.; Taylor, D. H.; Bull, R. J.; Sivulka, D. J.;
Lutkenhoff, S. D. (1980) Developmental delays in exploration
and locomotor activity in male rats exposed to low level
lead. Life Science 26; 823-831.
De la Burde, B.; Choate, M. S., Jr. (1972) Does asymptomatic lead
exposure in children have latent sequelae? Journal of
Pediatrics (St. Louis) 81: 1088-1091.
De la Burde, B.; Choate, M. S., Jr. (1975) Early asymptomatic lead
exposure and development at school age. Journal of Pediatrics
(St. Louis) 87: 638-642.
Dresner, D. L.; Ibrahim, N. G.; Mascarenhas, B. R.; Levere, R. D.
(1982) Modulation of bone marrow heme and protein synthesis by
trace elements. Environmental Research 28: 55-66.
Erickson, M. M.; Poklis, A.; Gantner, G. E.; Dickinson, A. W.;
Hillman, L. S. (1983) Tissue mineral levels in victims of
sudden infant death syndrome. I. Toxic metals-lead and
cadmium. Pediatric Research 17: 779-784.
Fisher, R. A. (1970) Statistical Methods for Research Workers.
Hafner Press, New York.
-------
VI.45
Gross-Selbeck, E.; Gross-Selbeck, M. (1981) Changes in operant
behavior of rats exposed to lead at the accepted no-effect
level. Clinical Toxicology 18: 1247-1257.
Harvey, P.; Hamlin, M.; Kumar, R. (1983) The Birmingham blood lead
study. Presented at: annual conference of the British
Psychological Society, symposium on lead and health. Available
for inspection at: U.S. Environmental Protection Agency,
Environmental Criteria and Assessment Office, Research Triangle
Park, NC.
Hernberg, S.; Nikkanen, J. (1970) Enzyme inhibition by lead under
normal urban conditions. Lancet 1: 63-64.
Holtzman, D.; Shen Hsu, J. (1976) Early effects of inorganic lead
on immature rat brain mitochondrial respiration. Pediatric
Research 10: 70-75.
Journal of Pediatrics (editorial) (1977) New approaches to screening
for iron deficiency. 90: 678.
Kammholz, L. P.; Thatcher, L. G.; Blodgett, F. M.; Good, T. A.
(1972) Rapid protoporphyrin quantitation for detection of
lead poisoning. Pediatrics 50: 625-631.
Lamola, A. A.; Joselow, M.; Yamane, T. (1975a) Zinc protoporphyrin
(ZPP): a simple, sensitive, fluorometric screening test for
lead poisoning. Clinical Chemistry (Winston-Salem, N.C.)
21: 93-97.
Lamola, A. A.; Piomelli, S.; Poh-Fitzpatrick, M. B.; Yamane, T.;
Harber, L. C. (1975b) Erythropoietic protoporphyria and lead
intoxication: the molecular basis for difference in cutaneous
photosensitivity. II: Different binding of erythrocyte
protoporphyrin to hemoglobin. Journal of Clinical Investigation
56: 1528-1535.
Landrigan, P. J.; Baker, E. L., Jr.; Feldman, R. G.; Cox, D. H . ;
Eden, K. V.; Orenstein, W. A.; Mather, J. A.; Yankel, A. J.;
von Lindern, I. H. (1976) Increased lead absorption with anemia
and slowed nerve conduction in children near a lead smelter.
Journal of Pediatrics (St. Louis) 89: 904-910.
Lauwerys, R.; Buchet, J-P.; Roels, H.; Hubermont, G. (1978)
Placental transfer of lead, mercury, cadmium, and carbon
monoxide in women. I: comparison of the frequency distri-
butions of the biological indices in maternal and umbilical
cord blood. Environmental Research 15: 278-289.
Mahaffey, K. R.; Annest, J. L.; Roberts, J.; Murphy, M. S. (1982)
National estimates of blood lead levels: United States, 1976-
1980: association with selected demographic and socioeconomic
factors. New England Journal of Medicine 307 (10): 573-579.
-------
VI.46
Mahaffey, K. R.; Rosen, J. P.; Chesney, R. W.; Peeler, J. T.; Smith,
C. to.; De Luca, H. F. (1982) Association between age, blood
lead concentration, and serum 1,25-dihydroxcholecalciferol
levels in children. American Journal of Clinical Nutrition
35: 1327-1331.
McBride, W. G.; Black, B. P.; English, B. J. (1982) Blood lead
levels and behaviour of 400 preschool children. Medical
Journal Aust. 2: 26-29.
McCauley, P. T.; Bull, R. J. (1978) Lead-induced delays in
synaptogenesis in the rat cerebral cortex. Proceedings of
the American Society of Experimental Biology 37: 740.
Meredith, P. A.; Moore, M. R.; Campbell, B. C.; Thompson, G. G.;
Goldberg, A. (1978) Deltaaminolevulinic acid metabolism in
normal and lead-exposed humans. Toxicology 9: 1-9.
Moore, M. R.; Meredith, P. A. (1976) The association of delta-
aminolevulinic acid with the neurological and behavioral
effects of lead exposure. In: Hemphill, D. D., ed. Trace
Substances in Environmental Health - X; [proceedings of
University of Missouri's 10th annual conference on trace
substances in environmental health]; June; Columbia, MO.
Columbia, MO: University of Missouri-Columbia; pp. 363-371.
Needleman, H. L.; Gunnoe, C.; Leviton, A.; Reed, R.; Peresie, H.;
Maher, C.; Barrett, P. (1979) Deficits in psychological and
classroom performance of children with elevated dentine lead
levels. New England Journal of Medicine 300: 689-695.
Needleman, H. L.; Rabinowitz, M.; Leviton, A.; Linn, S.; Schoenbaum,
S. (1984) The relationship between prenatal exposure to lead
and congenital anomalies. Journal of the American Medical
Soc iety (in press).
Otto, D. A.; Benignus, V. A.; Muller, K. E.; Barton, C. N. (1981)
Effect of age and body lead burden on CNS function in young
children. I: Slow cortical potentials. Electroencephalogr.
Clin. Neurophysiol. 52: 229-239.
Otto, D.; Benignus, V.; Muller, K.; Barton, C.; Seiple, K.; Prah, J.j
Scroeder, S. (1982) Effects of low to moderate lead exposure on
slow cortical potentials in young children: two year follow-up
study. Neurobeha. Toxicol. Teratol. 4: 733-737.
Piomelli, S.; Davidow, B.; Guinee, V. F.; Yound, P.; Gay, G. (1973)
The FEP (free erythrocyte protoporphyrin) test: a screening
micro-method for lead poisoning. Pediatrics 51: 254-259.
-------
VI. 47
Piomelli, S.; Seaman, C.; Zullow, D. ; Curran, A.; Davidow, B. (1977)
Metabolic evidence of lead toxicity in "normal" urban children.
Clinical Research 25: 459A.
Piomelli, S.; Seaman, C.; Zullow, D.; Curran, A.; Davidow, B. (1982)
Threshold for lead damage to heme synthesis in urban children.
Proceedings of the National Academy of Science U.S.A. 79 :
3335-3339.
Roels, H.; Buchet, J-P.; Lauwerys, R.; Huermont, G. ; Bruauz, P.;
Claeys-Thoreau, F.; Lafontaine, A.; Van Overschedle, J. (1976)
Impact of air pollution by lead on the heme biosynthetic
pathway in school-age children. Archives of Environmental
Health 31: 310-316.
Rosen, J. G.; Chesney, R. W.; Hamstra, A.; De Luca, H. F.; Mahaffey,
K. R. (1980) Reduction in 1 , 25-dihydroxyvitamin D in children
with increased lead absorption. In: Brown, S. S.; Davis, D. S.,
eds. Organ-directed Toxicity: Chemical Indices and Mechanisms.
New York, NY: Pergamon Press; pp. 91-95.
Sassa, S.; Granick, J. L.; Granick, S.; Kappas, A.; Levere, R. D.
(1973) Studies in lead poisoning. I: Microanalysis of erythrocyte
protoporphyrin levels by spectrof luormetry in the detection
of chronic lead intoxication in the subclinical range. Biochem.
Med. 8: 135-148.
Schwartz, J. D.; Janney, A.; Pitcher, H. (1983) The relationship
between gasoline lead and blood lead. (submitted for publication)
Smith, M.; Delves, T.; Lansdown, R. ; Clayton, B.; Graham, P. (1983)
The effects of lead exposure on urban children: the Institute
of Child Health/Southampton study. London, United Kingdom:
Department of the Environment.
Steenhout, A. (1982) Kinetics of lead storage on teeth and bones:
an epidemiologic approach. Archives of Environmental Health
37: 224-231.
Steenhout, A.; Pourtois, M. (1981) Lead accumulation in teeth as a
function of age with different exposures. British Journal of
Industrial Medicine 38: 297-303.
U.S. Centers for Disease Control (1978) Preventing lead poisoning
in young children. Journal of Pediatrics Vol 93.
U.S. Department of Health and Human Sevices (1982) Vital and Health
Statistics. Advance Data #79, May 12, 1982.
U.S. Environmental Protection Agency (1983) Review Draft Air
Quality Criteria for Lead (Research Triangle Park, NC.).
-------
VI.48
White, J. M.; Harvey, D. R. (1972) Defective synthesis of A and B
globin chains in lead poisoning. Nature (London) 236: 71-73.
Winneke, G.; Hrdina, K-G.; Brockhaus, A. (1982a) Neuropsychological
studies in children with elevated tooth-lead concentrations.
Part I: Pilot study, Int. Arch. Occupational and Environmental
Health 51: 169-183.
Winneke, G.; Lilienthal, H.; Werner, W. (1982) Task dependent
neurobehavioral effects of lead in rats. Arch Toxicol: Supp.
5: 84-93.
Winneke, G.; Brockhaus, A.; Baltissen,
and systemic effects of long term
rats. Arch Toxicol. 37:247-263.
R. (1977) Neurobehavioral
blood lead elevation in
Yule, W.; Lansdown, R. (1983) Lead and children's development:
recent findings. Presented at: International conference:
Management and Control of Heavy Metals in the Environment;
September; Heidelberg, West Germany.
Yule, W.; Lansdown, R.; Millar, I. B.; Urbanowicz, M. A. (1981)
The relationship between blood lead concentrations, intel-
ligence and attainment in a school population: a pilot
study. Dev. Med. Child. Neurol. 23: 567-576.
*U.S. GOYEMIMBIT PBIMTHG 07?IOE 1 1984 0-442-296/18372
------- |