United States
Environmental Protection
Office of
Poicy Analysis
Washington, DC 20460
March 1984
Draft Final Report
Costs and Benefits of
Reducing Lead in  Gasoline

            COSTS  AND  BENEFITS
               OF  REDUCING
             LEAD  IN GASOLINE
              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)


              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



    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


   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


   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


   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


   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

              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

    TABLE 1-1

    TABLE 1-2

    FIGURE 1-1


    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



II. 3

II. 4

II. 8

    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








                   TABLES, EXHIBITS, AND FIGURES (Continued)

    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




    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

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

      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









                   TABLES, EXHIBITS, AND FIGURES (Continued)

    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






    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




    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


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
VI. 12



                                EXECUTIVE SUMMARY



     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.


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


     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**

 Manufacturing Costs

 Non-monetized Valve Damage
   to Engines that Need Lead


 Maintenance Benefits

 Environmental and Health Benefits

      Conventional pollutants

      Reduced damage by eliminating misfueling

      Non-monetized health benefitst


      Reduced medical care costs

      Reduced cognitive damage

      Non-monetized health benefits''










 * 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
(thousands of metric tons)


     Ozone (resulting from reduced
            HC and NOx emissions)


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

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

1.5% reduction

                    ALL UNLEADED

1.5% reduction


                           CHAPTER I


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.


     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


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.

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.


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


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.


     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


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.


  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

* Scavengers are necessary to remove lead from the engine after
  combustion.  Without these scavengers, engine performance
  would rapidly deteriorate to complete inoperability.


                                       TABLE I-l

                          Comparison of Benefits and Costs of
                             Lead Reduction Options in 1988
                               (millions of 1983 dollars)

 Manufacturing Costs

 Non-monetized Valve Damage
   to Engines that Need Lead

                                                   Low-Lead Option*   All Unleaded**



 Maintenance Benefits

 Environmental and Health Benefits

      Conventional pollutants

      Reduced damage by eliminating misfueling

      Non-monetized health benefitst


      Reduced medical care costs

      Reduced cognitive damage

      Non-monetized health benefits







 * 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.


     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

                                    TABLE 1-2
                                GASOLINE LEAD USE
(thousands of metric tons)


     Ozone (resulting from reduced
           HC and NOX emissions)


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

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

1.5% reduction

                     ALL UNLEADED

1.5% reduction



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


     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.

8 100




0 60 H


                (FEB.  1976  -  FEB, 1980)
                  LEAD USED IN















     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


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


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


     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


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


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.


     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


     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


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.)

                           TABLE II-l

                   Projected Gasoline Demands
                 (millions of barrels per day)

                    With Misfueling         Without Misfueling

1982  (actual)

*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%


     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


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%


     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.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

 * 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.


     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.


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.


     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


     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,


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


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


                           TABLE II-4

           Cost of Reducing or Banning Leaded Gasoline

                               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.


     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


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.


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


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


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.


     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


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


                   EXHIBIT II-l



      GAS FUEL

      a MILITARY
      JET FUEL
             YIELD, VOLUME
             PERCENT OF
            RAW MATERIALS


    > FUEL OIL

'''Included  with  gas  fuel


                             EXHIBIT II-2
                      GASOLINE (LOW OCTANE)
                                                  GASOLINE (HIGH OCTANE)
                                                                                                 REFINERY GAS
                                                                                                 FUEL (CONSUMED



                                                                                                 JET FUEL.
                                                                                                 FUEL OIL 8
                                                                                                 DIESEL FUEL
                                                                                                 FUEL OIL
                                                                                                FUEL (CONSUMED

           * Included with gas  fuel

                         EXHIBIT II-3




                    GASOLINE (LOW OCTANE)
                                                 GASOLINE (HIGH OCTANE)
             YIELD. VOLUME
              PERCENT OF
             RAW MATERIALS



                                                                                                JET FUEL,
                                                                                                FUEL OIL 8
                                                                                                DIESEL FUEL

        *   Percent by weight
        **  Included with  gas  fuel

                          EXHIBIT I1-4

                                                                      RECOVERY a
                                                                             (HIGH OCTANE)
                     GASOLINE (LOW OCTANE
                                                  GASOLINE (HIGH OCTANE)
               LIGHT GAS OIL
               HEAVY GAS
                                                  GASOLINE (MEDIUM OCTANE)
                                              GASOLINE (LOW OCTANE)
                                         	I NAPHTHA
                                         —I GAS OIL
         ___  FOR HYDROGEN
          «-J  TREATING
         "T1  REFINERY
          ^  GAS FUEL

                                                                                                  JET FUEL.
                                                                                                  FUEL OIL a
                                                                                                  DIESEL FUEL
             LIQUID FUEL
                                                                                                 , HIGH SULFUR
         *   Percent by weight
         **  Included  with  gas  fuel

                              EXHIBIT II-5
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

   Extraction (deasphalting of feed-
     stock for lubricating oil manu-
     facture or for catalytic

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)

A. Conversion on the Basis of
   Molecular Weight

   Thermal cracking (visbreaking,

   Catalytic cracking



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)

                      Hydrogen treatment (hydrotreating)

                      Caustic treatment (Merox, Bender)

                      Clay treatment (of lubricating oils)

                      Acid treatment


function of the material that the unit is "fed."  Each unit

incurs some costs that vary with how hard it is run -- called


     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


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


     Exhibit II-8 presents projected capacity in 1988 for

various processing units in the model.

                        EXHIBIT II-6
                    CRUDE DISTILLATION UNIT
                                           Crude Oil Type
                                       Saudi Arab     Mexican
                                     Light    Heavy      Maya


                                     Yields  (Fraction of  Intake)
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) OTHER
DIST (500-620) HI SULFER
HY GAS OIL (800-BTMS) (2.0% S)





                                     Operating Cost Coefficients
                                      (Per Barrel of Throughput)
Note: The negative signs (-) indicate consumption of crude oil,
      fuel oil, power, steam, etc.

                         EXHIBIT II-7
                     CATALYTIC REFORMING UNIT
                  (200 PSIG Operating Pressure)
                                90 RON  100 RON    90 RON   100  RON
REF FEED (250-325) PARF
REF FEED (250-325) NAPH


                                  Yields  (Fraction of  Intake)
H2 (100 PCT FOE)




                                   Operating Cost Coefficients
                                     (Per Barrel of Thoughput)
Note: The negative signs (-) indicate consumption  of  crude  oil,
      fuel oil, power, steam, etc.


                       EXHIBIT II-8
               (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


                           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

     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

     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.

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

     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-

     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%

     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.


                    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


atmosphere or sent to a CO boiler where carbon monoxide is burned

to produce C02«  The regenerated catalyst is returned to the


     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-

cracking produces by-product H2S, which is diverted to sulfur


     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


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


Dupont Petroleum Chemicals, Tech Brief 17909, December 1979.

Energy Information Administration, Annual Report to Congress,

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


       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.


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


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


     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,


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.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

                                                                   TABLE III-l

Pahnke & Bettoney
(DuPont, 1969)

(rpted. in Pahnke &
Bettoney, 1971)
Gray & Azhari (1972)

MY 1967:
MY 1968:
Gray & Azhari (1972)

Wintringham et.al. ,
(Ethyl, 1972)

Baton Rouge:

Hickling Partners
(Environment Canada)

Municipal Fleets

Changes in








2.9 times





3d avg.


as many
w/leaded vehicles

2.2 more replace-
ments for
net of
other technology




.0033 .187

.220 ' .275


weigh tec


1 avg.

.155 .289
.004 .358

2.4 times as
many for leaded
vehicles (they
exclude Toronto

(Not Applicable)





Not reported



(Not Applicable)





1 to 6 yrs.


23,810 leaded
24,990 unleaded

(Not Applicable)

Personal Use


Commuting and
business use

Personal Use

Employee Fleet
(Business and
Personal Use)

Municipal Service

(Not Applicable)

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)

South New Jersey
and Wilmington,


Chicago and

Eastern states
in Mid-Atlantic

Baton Rouge


(Not Applicable)


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


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


     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


     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


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


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


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


     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


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


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


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


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).


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).


In fact, using the data from MVMA (1984)  reveals an estimate of

$328 million* — over four times the value derived In our


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.


     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.


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


     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


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


     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


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



     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


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.


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


     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.


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.


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.


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


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


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


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.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.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

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.)


     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:


     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.


     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).

                                                        TABLE  III-2


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
No extra valve problems with unleaded
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

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
More expensive valve problems with unleaded


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-


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.


     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


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.


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,


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.


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.


                           Table III-3


                                Thousands of Vehicles in 1988
Type of Vehicle

Cars Pre-'Vl

LDTt Pre-171

Cars 1971-1980

LDTt 1971-1975

Heavy-Duty Trucks


   High Risk*



Not calculated

 Low Risk*


Not calculated

                 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

t  Light-duty trucks.


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


                                    TABLE II1-4

                           SUMMARY OF MAINTENANCE SAVINGS
                            (1983 dollars, in millions)
Billions of gallons of
leaded gasoline demand in 1988

    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
$ 36
$ 44
$ 80

    Exhaust Systems

    Spark Plugs

    Oil Changes

   $ 36l



$ 443

$755 Million


 $ 80

                                   $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

    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.



                           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.


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,

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.

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


     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


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.


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


                              TABLE IV-2

      1982 Misfueling Rates by Age of Vehicle and by I/M Status
Misfueling Rates

I/M Areas

Non-I/M Areas
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.


     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


                               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



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.


     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
 * 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,

 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.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


      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.


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

*  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.C.  Benefits Estimated Directly from Health and Welfare

     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


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.


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.


     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.


     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:
                      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


                       where T =  b (H x N)-5

Taking the derivative and collecting terms:

      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.


     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


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,


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


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.


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


     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.


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


     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.


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


     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


     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.

                            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
Attack Rate
Value of
$ 70
$ 70
$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.


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


     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.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


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.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


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


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

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.


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-


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)


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.


     Probably the greatest concern about CO exposure is its

effect on the cardiovascular system.   The effect of CO thus far


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


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:


     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.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


  Acute Health                       $59-233            $146
  Agriculture                        $30-225            $160
  Vegetation                         $ 0- 15            $7.5
  Materials Damage                   $16- 28            $ 24
  Chronic Health                        NA                OZC


  Health and Visibility              $7.8-62             $35


  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.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

Table IV-7 summarizes the values obtained by each of these three


     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

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


                             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


     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
MODEL YR    1982

10, 155

1 , 306
4, 176

9, 191
8, 159

1 1 , 000
9, 163

1 1 , 200
8, 127
1 1 , 600
7 , 900
5, 182
3, 122
1 , 523
1 , 239
BaaaaB an
1 1 , 800
1 1 , 588
11, 155
4 , 300
           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
 YEAR      1980    1981
1984    1985





1 , 834
1 , 802
1 , 904
1, 108

3, 1O4
1, 137
 * 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


     Table IV-3 is a summary of the results.

                       APPENDIX TABLE IV-3


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.


     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,


                                APPENDIX TABLE IV-4

                             GENERAL  FLEET  ASSUMPTIONS

1 OR <
# OF
1 1 , 030
10, 160
9, 168
1 , 708
7. MIS-
IN 1 YR*
7. MIS- -
20. 1
20. 1
20. 1
20. 1
20. 1
20. 1
20. 1
20. 1
20. 1



12, 175
1 1 , 650
1 1 , 075

6, 163
5, 160
                                        are. SB SEES S
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


    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:

         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
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

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

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,

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,

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,"

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,

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,

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

                              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,

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


     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


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


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,

                               FIGURE  V-l
                                          HISPANIC —
                                    GASOLINE LEAD	
"w 30


§  25


|"  15


                                                    \  /



4.0  f
                                                                  3.0 2
                1  J^ J I   I  I  I
                                             .  I.
                                                          I  I  I
     1970    1971     1972     1973     1974     1975

                      QUARTERLY SAMPLING DATE


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


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


(CDC) lead toxicity category II, and children in CDC categories

III and IV.*  Treatment and follow-up practices may differ for


     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.


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


   (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.


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


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)


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


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/


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.


     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.


     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.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


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,


     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.

                                 FIGURE  V-2
o in
!2" _J

~x m

                AVERAGE BLOOD LEAD LEVELS   (micrograms/deciliter)







                    TOTAL  LEAD  USED PER 6 MONTH PERIOD  (1000 tons)

  17 -i
6 12-
y 11 H
S  9
                               CORRELATION COEFFICIENT « 0.95
                                           (P < .0001)
50      60      70      80      90      100

                           FIGURE V-4
                               r—-r—-r-—i	1	1
                                                   OF GRAMS

                                                .1  O BLOOD LEAD

                                                   A GASLEAD


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.


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


                       (1)   a = exp (u + 1/2 s2)

                       (2)   6g = eXp (U + Vg S)

Solving these equations for u and s produced a quadratic equation


                 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,


     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.


     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.


Conversely, if children remain above 30 ug/dl for more than a

year, these cases may be counted twice and we will overstate


     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.,


     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


.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



     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


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.


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


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



                            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


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.



     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



Age 1
Age 2

Age 3

Age 4

Age 5

Age 6

Age 7

Income 1

Income 2

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
1 if gender is male; 0 if female

Variable Name

Teen Male

Adult Male

Small City



Heavy Drinker
Northeast, Midwest,

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.
1 if age < 6 ; 0 otherwise

Age 1
Age 2
Age 3

                    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
Age 1
Age 2
Age 3
*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.


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
Age 5
Age 6
                                               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
Age 5
Age 6
Fraction of concordant pairs of predicted probabilities
and responses = 0.710

                    SURREGR Regression Results
Whites;  children 6 months to 7 years
Dependent variable: individual blood lead levels
Income 1
Income 2
Age 1
Age 2
Age 3
Age 4
Teen male
Adult male
Small City
Heavy Drinker
Education level
Std. Error
0 . 0 29 6

Whites:  6 months to 13 years
Dependent variable: individual
Income 1
Income 2
Teen male
Small City
Adult male
Age 4
Age 5
Age 6
Age 7
Heavy Drinker
Education level
Beta Std
. Error
lead levels



Blacks:  6 months to 7 years
Dependent variable: individual blood lead levels
Income 1
Income 2
Age 1
Age 2
Age 3
Age 4
Adult male
Heavy drinker
Education level
Std. Error

Blacks;  6 months to 13 years
Dependent variable
Income 1
Income 2
Adult male
Age 4
Age 5
Age 6
Age 7
Heavy Drinker
: individual
Beta Std
South -0.1173
Education level
blood lead
. Error



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.

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

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

U.S. Environmental  Protection Agency (1983) Review Draft Air
     Quality Criteria for Lead, Volumes III and IV  (Research
     Triangle Park, North Carolina).


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:

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
                     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

     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

      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%


   all ages             4.0%
   6 months-5 years     2.7%
   6-17 years           8.0%
   18-74 years          2.3%


31 .0%


*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.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


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


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


     ...[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)


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


                             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


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

                          Figure VI-1

                 (Age 6 Months to 8 Years)
   20 -
 53 10
                   With Transferrin Saturation
               One Standard Deviation Below Average
                                     With Average
                                  Transferrin Saturation
                     Blood Lead Level (/tg/dl)


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


(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.


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

    By Age and Race at Average Transferrin Saturation Levels
c 30

"5 24


2—6 yrs.

                                  0.5 — 2 yrs.
20     25     30     35
45    50
                       FEP Level (jtg/dl)


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


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.


     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


noted in rats with blood lead levels below 30 ug/dl (Draft

Criteria Document, Section  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


 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.


     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


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).


     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
McBride et al.
Yule et al.
Smith et al.
Yule and Lansdown
Harvey et al.
Winneke et al.
Joint P-Value

Needleman et al.
Total of 189
for Studies:

> 25.10)

= .014

Lead Levels
Teeth (ppm)
Control Exposed IQ Difference
< 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

.03 7.01
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


      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 -
£ -1-
0 _
U ^
! -3 -
£ -4-
§ -l-
£ -7-
a -d -
q -10 -
b -11 -
3 -12 -
£ -13 -
-14 -


i :





, '

; ;




' •

1 1

-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


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.


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.


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


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


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


     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


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


     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


  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.


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


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.

                           TABLE VI-5

        Decreased Number of Children (under 14 years old)
             Above Apparent Threshold Levels in 1988

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

     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.)


                           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
;.l. Distributional Aspects of
aged 13
1 ,085
and under)
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.


      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


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


options suggest that reducing the pathophysiological effects of

lead exposure would be a significant public health benefit of

reducing lead in gasoline.

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.


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.


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 :

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.).

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