&EPA
             United States
             Environmental Protection
             Agency
             Office of
             Policy Analysis
             Washington DC 20460
February 1985
EPA-230-05-85-006
Costs and Benefits of
Reducing Lead in Gasoline

Final Regulatory Impact
Analysis

-------
COSTS AND BENEFITS OF REDUCING LEAD IN GASOLINE:


        FINAL REGULATORY IMPACT ANALYSIS
                 February 1985
                    Authors:
                 Joel Schwartz
                  Hugh Pitcher
                  Ronnie Levin
                   Bart Ostro
               Albert L. Nichols
           Economic Analysis Division
           Office of Policy Analysis.
   Office of Policy, Planning and Evaluation
      U.S. Environmental Protection Agency
              Washington, DC 20460

-------
                       ACKNOWLEDGEMENTS
Contributors;

Jane Leggett
Albert McGartland
George Sugiyama


     The authors thank their colleagues Lane Krahl, Robert
Fegley, and June Taylor Wolcott for their help and expertise.

     For technical assistance, often under very tight deadlines,
the authors thank Terry Higgins, Bill Johnson, and Steve Sobotka
of Sobotka and Company; Chris Weaver and Craig Miller of Energy
and Resource Consultants; Asa Janney and Jim Duffey of ICF;
and Ed Fu.

     Many individuals have provided excellent secretarial
support, in particular Saundra Womack, Georgia Jackson, Hallie
Baldwin, Sylvia Anderson, Michelle Smith, Leslie Hanney, and
Yvette Carter.

     The authors also are grateful to the many colleagues who
provided helpful comments on earlier drafts of this document
and its predecessors.

-------
                             TABLE OF CONTENTS






                                                                  PAGE




EXECUTIVE SUMMARY                                                  E-l






CHAPTER Ii  INTRODUCTION




  I.A.  Background Information                                     1-2



     I.A.I.  Previous Rulemakings                                  1-3



     I.A.2.  Continuing Problems                                   1-4



     I.A.3.  The August 1984 Proposal                              1-8



     I.A.4.  Information Received After August 1984 Proposal       1-9



  I.E.  Need for Government Intervention                           1-13



  I.C.  Alternatives to New Regulations                            1-14



     I.C.I.  No Change in Policy                                   1-14



     I.C.2.  Public Education                                      1-15



     I.C.3.  Stepped-Up Enforcement                                1-16



  I.D.  Market-Oriented Alternatives                               1-18



     I.D.I.  Marketable Permits                                    1-19



     I.D.2.  Pollution Charges                                     1-21



  I.E.  Alternative Standards                                      1-23

-------
                            Table of Contents, page 2



                                                                   PAGE



CHAPTER II:  COSTS OF REDUCING LEAD IN GASOLINE



  II.A.  Price versus Cost Differences                             II-2



  II.B.  The Refinery Model                                        II-6



      II.B.I.  Introduction to the DOE Model                       II-6



      II.B.2.  Overview of Refining Processes                      II-7



  II.C.  Base-Case Assumptions and Cost Estimates                  11-18



      II.C.I.  Base-Case Parameter Values                          11-19



         II.C.I.a.  Gasoline Demand                                11-19



         Il.C.l.b.  Assumptions About Refinery Operations          11-24



      II.C.2.  Base-Case Results                                   11-34



  II.D.  Sensitivity Analyses                                      11-39



      II.D.I.  Level of Aggregation                                11-40



      II.D.2.  Other Parameters                                    11-42



         II.D.2.a.  Assumptions Varied                             11-44



         II.D.2.b.  Results of Sensitivity Analyses                11-48



  II.E.  Impact of Banking on Costs                                11-60



     II.E.I.  Base-Case Banking Results                            11-61



     II.E.2.  Sensitivity Analyses with Banking                    11-68

-------
                            Table of Contents, page 3


                                                                   PAGE

CHAPTER III:  HUMAN EXPOSURE TO LEAD FROM GASOLINE

  III.A.  The Relationship between Lead in Gasoline
          and Lead in Blood                                        II1-2

       III.A.I.  Recent Studies                                    III-2

       III.A.2.  Available Data Sets                               III-5

           III.A.2.a.  Gasoline-Use Data                           III-6

           III.A.2.b.  The NHANES II                               III-6

           III.A.2.c.  The CDC Screening Program for               111-10
                       Lead Poisoning

           III.A.2.d.  Chicago, New York, and Louisville,          111-10
                       Kentucky Data

    III.A.3.  Statistical Analyses of Exposure                     111-10

  III.B.  The Question of Causality                                111-21

       III.B.I.  Experimental Evidence                             111-22

       III.B.2.  Does Cause Precede Effect?                        111-23

       III.B.3.  Replicability and Consistency                     111-24

       III.B.4.  Does a Dose-Response Relationship Exist?          111-25

       III.B.5.  Biological Plausibility                           111-26

       III.B.6.  Control of Confounding Factors                    111-26

           III.B.6.a.  External Validation                         111-27

           III.B.6.b.  Seasonality                                 111-30

           III.B.6.C.  Other Time Trends                           111-31

           III.B.6.d.  Geographic Sampling Pattern                 111-31

           III.B.6.e.  Subgroup Analysis                           111-33

  III.C.   Impact of Rule on Numbers of Children Above              111-34
          Various Blood-Lead Levels

       III.C.1.  Estimation Procedure                              111-35

       III.C.2.  Incidence Versus Prevalence                       II1-39

-------
                            Table of Contents, page 4

                                                                       PAGE

CHAPTER IV:  BENEFITS OF REDUCING CHILDREN'S EXPOSURE TO LEAD

  IV.A.  Pathophysiological Effects                                    IV-3

      IV.A.I.  Effects of Lead on Pyrimidine Metabolism                IV-7

      IV.A.2   Effects on Heme Synthesis and Related Hematological     IV-8
               Processes

         IV.A.2.a.  Mitochondrial Effects                              IV-8

         rv.A.2.b.  Heme Synthesis Effects                             IV-8

         IV.A.2.C.  Impact of Lead on Red Blood Cell Abnormalities     IV-10

                IV.A.2.c.l.  Effects of Lead Exposure on Blood Cell    IV-11
                             Volume and Hemoglobin Content

                IV.A.2.c.2.  The Relationship Between Blood Lead       IV-18
                             and FEP

                IV.A.2.c.3.  The Relationship Betwen FEP Levels and    TV-21
                             Anemia

       IV.A.3.  Lead's Interference with Vitamin D Metabolism          IV-24
                and Associated Physiological Processes

  IV.B.  Neurotoxic Effects of Lead Exposure                           IV-28

      IV.B.l.  Neurotoxicity at Elevated Blood-Lead Levels             IV-28

      IV.B.2.  Neurotoxicity at Lower Blood-Lead Levels                IV-30

         IV.B.2.a.  Cognitive Effects of Moderate Blood-Lead Levels    IV-33

      IV.B.3.  The Magnitude Impact of Lead on IO                      IV-41

  IV.C.  Fetal Effects                                                 IV-42

  IV.D.  Monetized Estimates of Children's Health Benefits             IV-47

      IV.n.l.  Reduced Medical Costs                                   IV-47

      IV.D.2.  Reduced Costs of Compensatory Education                 IV-52

      IV.D.3.  Summary of Estimated Benefits                           IV-53

-------
                    Table of Contents, page 5

                                                                   PAGE

CHAPTER V:  HEALTH BENEFITS OF REDUCING LEAD:
            ADULT ILLNESSES RELATED TO BLOOD PRESSURE

  V.A.  The Relationship Between Blood Lead and Blood Pressure      V-2

     V.A.I.  Earlier Studies                                        V-2

     V.A.2.  Analysis of NHANES II Data                             V-5

        V.A.2.a.  Blood Pressure Measurements                       V-7

        V.A.2.b.  Initial Analysis                                  V-7

     V.A.3.  Tests of Robustness                                    V-9

        V.A.3.a.  Nutritional and Biochemical Variables             V-9

        V.A.S.b.  Interaction Terms                                 V-15

        V.A.3.C.  Marginally Insignificant Variables                V-18

        V.A.S.d.  Nonnutrition Variables                            V-19

        V.A.3.e.  Other Age Groups                                  V-23

     V.A.4.  Summary of Blood Lead - Blood Pressure Results         V-23

  V.B.  Benefits of Reduced Cardiovascular Disease                  V-26

     V.B.I.  Reductions in Hypertension and Related Morbidity
             and Mortality                                          V-26

        V.B.1.a.  Hypertens ion                                      V-27

        V.B.l.b.  Myocardial Infarctions, Strokes, and Deaths       V-29

     V.B.2.  Monetized Benefit Estimates                            V-35

        V.B.2.a.  Hypertension                                      V-35

        V.B.2.b.  Myocardial Infarctions                            V-37

        V.B.2.C.  Strokes                                           V-40

        V.B.2.d.  Mortality                                         V-42

     V.B.3.  Summary of Blood Pressure Benefits                     V-44

-------
                    Table of Contents, page 6

                                                                     PAGE

CHAPTER VI:  BENEFITS OF REDUCING POLLUTANTS OTHER THAN LEAD

  VI .A.  Emissions Associated with Misfueling                        VI-4

  VI.B.  Health and Welfare Effects Associated with Ozone            VI-9

      VI.B.I.  Effects of a 1 Percent Reduction in Ozone             Vl-11

         VI.B.I.a.  Health Effects of Reducing Ozone                 VI-11

         VI.B.l.b.  Ozone Agricultural Effects                       VI-22

         VI.B.I.e.  Ozone Effects on Nonagricultural Vegetation      VI-26

         VI.B.l.d.  Ozone Materials Damage                           VI-28

         VI.B.I.e.  Summary of Benefits of a 1 Percent               VI-29
                    Change in Ozone

      VLB.2.  Linking NOX and HC Reductions to Ozone Effects        VI-31

         VLB.2.a.  The Process of Ozone Formation                   VI-31

         VI.B.2.b.  Quantitative Estimates of Impacts of             VI-33
                    HC and NOX on Ozone

         VI.B.2.C.  Ozone-Related Effects Per Ton of HC and          VI-39
                    NOX Controlled

  VI.C.  Health and Welfare Effects Not Related to Ozone             VI-43

      VI.C.1.  Hydrocarbons                                          VI-43

         VI.C.I.a.  Impact on Sulfates                               VI-43

         Vl.C.l.b.  Impact on Benzene and Other Aromatics            VI-44

      VI.C.2.  Nitrogen Oxides                                       VI-47

         VI.C.2.a.  Visibility Benefits from Reduced NOx             VI-48

         VI.C.2.b.  Health Benefits of Reducing NOx                  VI-50

         VI.C.2.C.  NOX Effects on Vegetation                        VI-52

         VI.C.2.d.  NOX Effects on Materials                         VI-53

         VI.C.2.e.  Acid Deposition Benefits                         VI-54

-------
                  Table of Contents, page 7


                                                                 PAGE

    VI.C.3.  Carbon Monoxide                                     VI-55

       Vl.C.S.a.  Health Effects of CO                           VI-56

       VI.C.3.b.  Change in Numbers of People Above
                  2.9 Percent COHb                               VI-59

    VI.C.4.  Ethylene Dibromide Emissions                        VI-64

VI.D.  Monetized Benefit Estimates                               VI-66

    VI.D.I.  Value of Quantified Health and Welfare Benefits     VI-66

    VI.D.2.  Implicit Value Based on Cost of Control
             Equipment                                           VI-68

    VI.D.3.  Summary of Benefits of Controlling Conventional
             Pollutants Other than Lead                          VI-71

-------
                   Table of Contents, page 8

                                                                        PAGE

CHAPTER VII:  VEHICLE MAINTENANCE, FUEL ECONOMY,
              AND ENGINE DURABILITY BENEFITS

  VILA.  Maintenance Benefits                                          VII-3

      VII.A.I.  Exhaust Systems                                         VII-3

      VII.A.2.  Reduced Fouling and Corrosion of Spark Plugs            VII-9

      VII.A.3.  Extended Oil Change Intervals                           VII-11

      VII.A.4.  Summary of Maintenance Benefits                         VII-14

  VII.B.  Improved Fuel Economy                                         VII-18

      VII.B.I.  Energy Content                                          VII-18

      VII.B.2.  Reduced Fouling of Oxygen Sensors                       VII-19

      VII.B.3.  Summary of Fuel Economy Benefits                        VII-20

  VII.C.  Engine Durability                                             VII-20

      VII.C.I.  Valve-Seat Recession                                    VII-22

          VII.C.I.a.  Laboratory and Track Studies of Valve-
                      Seat Recession                                    VII-22

          Vll.C.l.b.  Fleet Studies of Valve-Seat Recession             VII-27

          VII.C.I.e.  Other Types of Engines                            VII-35

          Vll.C.l.d.  Alternatives to Lead to Avoid Potential
                      Valve Recession                                   VI1-36

      VII.C.2.  Negative Effects of Lead on Engine
                Durability                                              VII-38

      VII.C.3.  Summary of Engine Durability Effects                    VII-43

-------
                              Table of Contents, page 9


                                                                          PAGE

CHAPTER VIIIi COST-BENEFIT ANALYSIS OF ALTERNATIVE
              PHASEDOWN RULES

  VIII.A.  Summary of Cost and Benefit Estimates                          VIII-2

  VIII.B.  Comparisons of Alternative Lead Levels                         VIII-7

  VIII.C.  Impact of Banking on Costs and Benefits                        VIII-27
           of Final Rule

  VIII.D.  Summary                                                        VIII-33



  REFERENCES                                                              R-l

  APPENDIX A:  Refinery Processes                                         A-l

  APPENDIX B:  Fleet Model                                                B-l

  APPENDIX C:  Supplementary Regressions of Blood Lead on
               Gasoline Lead                                              C-l

-------
                               LIST Of TABLES
CHAPTER I:

    TABLE 1-1
Alternative Phasedown Schedules
                                                    PAGE
1-24
CHAPTER II:

    TABLE II-l


    TABLE I1-2



    TABLE I1-3

    TABLE I1-4


    TABLE I1-5


    TABLE 11-6



    TABLE I1-7


    TABLE 11-8


    TABLE 11-9



    TABLE 11-10

    TABLE 11-11


    TABLE 11-12


    TABLE 11-13

    TABLE 11-14
Functional Characterization of Refinery
Processes                                          11-13

Sample Process Data Table from Refinery Model:
Yields and Operating Cost Coefficients for         11-15
Crude Distillation Unit

Year-by-Year Estimates of Gasoline Demand          11-23

Estimated U.S. Refinery Processing Unit
Capacities for 1988                                11-26

Prices of Crude Oil and Petroleum Products
in 1983 and 1985                                   11-31

Cost of the 0.10 gplg Standard with New Oil
Prices:  New Model Run versus Repricing,
Assuming No Misfueling                             11-33

Base-Case Results for 1985 - 1988, with
Partial Misfueling                                 11-36

Year-by-Year Estimates of Costs of Meeting
Alternative Rules, with Partial Misfueling         11-38

Costs of Meeting the 0.10 gplg Standard:
Comparison of National and Regional Results        11-43
for 1986

Parameters Examined in Cost Sensitivity Analyses   11-45

Effects of Varying Individual Parameters/
Assumptions:  PADDs I-IV/VI                        11-49

Effects of Varying Multiple Parameters/
Assumptions:  PADDs I-IV/VI                        11-52-53

Sensitivity Analyses for 1986:  PADD V             11-58

Sensitivity Analyses for 1985:  PADDs I-IV/VI      11-59

-------
                           List of Tables, page 2
    TABLE I1-15

    TABLE 11-16


    TABLE 11-17

    TABLE 11-18


    TABLE 11-19
Alternative Phasedown Patterns with Banking

Refining Costs Under Alternative Phasedown
Patterns, with Partial Misfueling

Impact of Banking on Marginal Costs of Octane

Sensitivity Analyses for 1986 with Banking:
Alternative 1, PADDs I-IV/Vl

Sensitivity Analyses for 1986 with Banking:
Alternative 2, PADDs I-IV/VI
 PAGE

11-63


11-65

11-67


11-69


11-70
CHAPTER III:

    TABLE III-l

    TABLE II1-2



    TABLE III-3a



    TABLE III-3b



    TABLE II1-4



    TABLE II1-5

    TABLE II1-6



    TABLE II1-7
NHANES II:  Regression Results for Whites         111-15

Logistic Regression on Probability of
Blood Lead > 30 ug/dl for Children                I11-16
6 months to 7 years

Regression of CDC Screening Data:
Percent of Children with Lead Toxicity on         111-18
Gasoline Lead

Regression of CDC Screening Data:  Change
in Lead Toxicity on Change in Gasoline
Lead                                              II I-18

Black Children in Chicago:  Regression of
Average Blood-Lead Levels on Gasoline             II1-20
Lead Levels

Lead in the Diet                                  111-28

Chicago Data:  Probability of Blood Lead
> 30 ug/dl With and Without Lead Paint Hazard     111-30
in the Hone

Estimated Reductions in the Numbers of
Children Over Various Blood-Lead Levels,          111-38
Assuming No Misfueling

-------
                           List of Tables, page 3
CHAPTER TV:

    TABLE IV-1


    TABLE IV-2


    TABLE IV-3



    TABLE IV-4



    TABLE IV-5


    TABLE IV-6


    TABLE IV-7
                                                                       PAGE
Blood Lead Levels of Children  in the United
States, 1976-80                                    IV-4

Variables Considered in the Regressions
of FEP, MCV, MCH, and Anemia                       IV-12

Computation of Joint P-Value from Epidemiological
Studies of Cognitive Effects from Low Level        IV-39
Lead Exposure in Children

Year-by-Year Estimates of Gain in Person -
10 Points Under Alternative Rules, Assuming        IV-43
No Misfueling

Estimated Decrease in Number of Fetuses
Exposed to 25 ug/dl of Blood Lead                  IV-47

Percent of Children Requiring Chelation            IV-50
Therapy

Year-by-Year Monetized Benefits of Reducing        IV-55
Children's Exposure to Lead Under Alternative
Rules, Assuming No Misfueling
CHAPTER V:

    TABLE V-l


    TABLE V-2



    TABLE V-3




    TABLE V-4


    TABLE V-5


    TABLE V-6



    TABLE V-7
Variables Included in the Stepwise                  V-ll
Regression Analyses

Regression of Diastolic and Systolic
Blood Pressures in White Males Aged                 V-13
40 to 59

Weighted Logistic Regression on Probability
of Diastolic Blood Pressure Greater Than            V-14
or Equal to 90 mm Hg in Men Aged 40 to 59

Nonnutrition Variables Tested in the Stepwise       V-20
Regression

Regression of Diastolic and Systolic Blood
Pressures in White Males Aged 49 to 50              V-24

Weighted Logistic Regression Probability
of Blood Pressure Greater Than or Equal
to 90 mm Hg in Men Aged 40 to 59                    V-25

Reductions in Cases of Hypertension in Males
Aged 40 to 59, Assuming No Misfueling               V-28

-------
                           List of Tables, page 4
    TABLE V-8      Reductions in Numbers of Cases of
                   Cardiovascular Disease and Deaths in Males
                   Aged 40 to 59, Assuming No Misfueling

    TABLE V-9      Benefits of Reducing Strokes

    TABLE V-10     Year-by-Year Estimates of Blood Pressure
                   Benefits, Assuming No Misfueling
                                                    PAGE



                                                    V-34

                                                    V-41


                                                    V-45
CHAPTER VI:

    TABLE VI-1

    TABLE VI-2

    TABLE VI-3


    TABLE VI-4


    TABLE VI-5




    TABLE VI-6


    TABLE VI-7


    TABLE VI-8


    TABLE VI-9



    TABLE VI-10


    TABLE VI-11


    TABLE VI-12


    TABLE VI-13
Increase in Emissions Due to Misfueling             VI-6

Misfueling Rates in 1983                            VI-8

Year-by-Year Estmates of Reductions in
Emissions, Assuming No Misfueling                   VI-10

Regression Results for Portney and Mullahy
Study on Respiratory Symptoms Related to Ozone      VI-15

Regression Coefficients from Hasselblad
and Svendsgaard Study on Respiratory and
Non-respiratory Symptoms Related to Ozone           VI-19

Estimated Health Effects of a 1 Percent
Reduction in Ozone                                  VI-23

Annual Agricultural Benefits of a 1 Percent
Ozone Reduction                                     VI-25

Summary of Estimated Effects of a 1 Percent
Reduction in Ozone                                  VI-30

Estimated Ozone Reductions from 1 Percent
Reduction in Rural and Metropolitan HC              VI-34
and NOX

Estimated Ozone Reductions due to Eliminating
Misfueling                                          VI-41

Quantified Ozone-Related Effects due to
Eliminating Misfueling in 1986                      VI-42

Reductions in Benzene Emmissions in 1986,
Assuming No Misfueling                              VI-46

Monetized Benefits Per Ton of Hydrocarbons
Controlled                                          VI-69

-------
                           List of Tables, page 5
    TABLE VI-14
 Year-by-Year Estimates of Benefits of Reduced
 Emissions of Conventional Pollutants, Assuming
 No Misfueling
PAGE

VI-74
CHAPTER VII:

    TABLE VII-1


    TABLE VI1-2


    TABLE VI1-3


    TABLE VI1-4


    TABLE VI1-5


    TABLE VI1-6


    TABLE VI1-7
Summary of On-Road Studies of Spark Plug             VI1-5
and Exhaust System Wear

Estimated Maintenance Benefits Per Mile of
Reducing Lead in Gasoline                            VI1-15

Year-by Year Estimates of Maintenance
Benefits, Assuming No Misfueling                     VII-17

Year-by-Year Estimates of Fuel Economy
Benefits, Assuming No Misfueling                     VII-21

Summary of Findings of Track and Dynamometer
Studies of Lead Levels and Valve Recession           VII-24

Summary of Findings of Consumer and Fleet
Studies of Lead Levels and Valve-Seat Recession      VII-29

Vehicle and Engine Types in U.S. Army
Unleaded Gasoline Test                               VI1-33
CHAPTER VIII:

    TABLE VIII-1   Costs and Monetized Benefits of 0.10 gplg
                   in 1986, Assuming No Misfueling:  Comparison      VIII-3
                   of Current and Draft RIA Estimates

    TABLE VI11-2   Non-monetary Measures of Health and
                   Environmental Benefits of 0.10 gplg in 1986,      VIII-6
                   Assuminig No Misfueling:  Comparison of
                   Current and Draft RIA Estimates

    TABLE VIII-3a  Costs and Monetized Benefits of Alternative
                   Lead Levels in 1986, Assuming No Misfueling       VIII-8

    TABLE VIII-3b  Costs and Monetized Benefits of Alternative
                   Lead Levels in 1986, Assuming Full Misfueling     VIII-9

    TABLE VIII-3c  Costs and Monetized Benefits of Alternative
                   Lead Levels in 1986, with Partial Misfueling      VIII-15

-------
                       List of Tables, page 6
TABLE VIII-4a


TABLE VIII-45


TABLE VIII-4C


TABLE VII1-5


TABLE VI11-6
TABLE VIII-7a


TABLE VIII-7b



TABLE VIII-Vc


TABLE VI11-8



TABLE VI11-9

TABLE VIII-10



TABLE VIII-11
Costs and Monetized Benefits of Alternative
Lead Levels in 1987, Assuming No Misfueling

Costs and Monetized Benefits of Alternative
Lead Levels in 1987, with Full Misfueling

Costs and Monetized Benefits of Alternative
Lead Levels in 1987, with Partial Misfueling

Costs and Monetized Benefits of Alternative
Lead Levels in 1985, Assuming Full Misfueling

Present Values of Costs and Monetized
Benefits: Comparison of Proposed,
Alternative, and Final Schedules for
1985-1987

Year-by-Year Costs and Monetized Benefits
of Final Rule, Assuming No Misfueling

Year-by-Year Costs and Monetized Benefits
of Final Rule, Assuming Full Misfueling

Year-by-Year Costs and Monetized Benefits
of Final Rule, with Partial Misfueling

Present Values of Costs and Benefits of
Final Rule, 1985-1992

Alternative Phasedown Patterns with Banking

Costs and Monetized Benefits of Alternative
Phasedown Patterns, with Partial Misfueling

Present Values of Costs and Benefits of
Alternative Phasedown Patterns, 1985-87,
with Partial Misfueling
 PAGE


 VIII-16


 VIII-17


 VI11-18


 VIII-20


 VIII-21
VII1-23


VIII-24


VIII-25


VIII-26

VIII-28


VIII-30


VIII-32

-------
                              LIST of FIGURES
CHAPTER I:

    FIGURE 1-1
Comparison of 1982 and Current
Projections of Lead Use in Gasoline
                                                    PAGE
                                                                        1-5
CHAPTER II:

    FIGURE II-l



    FIGURE II-2

    FIGURE I1-3

    FIGURE I1-4



CHAPTER III:

    FIGURE III-l



    FIGURE II1-2



    FIGURE II1-3

    FIGURE II1-4



    FIGURE II1-5

    FIGURE II1-6
Schematic Diagram of a Simple Oil Refinery
(Topping Plant)

Schematic Diagram of a Hydroskimming Refinery

Schematic Diagram of a Fuels Refinery

Schematic Diagram of a High Conversion Refinery
Relationship Between Gasoline Lead and Blood
Lead in New York City

Lead Used in Gasoline Production and Average
NHANES II Blood Lead Levels

NHANES II Data:  Blood Lead versus Gasoline Lead

CDC Data: Gasoline Lead versus Percent of
Children with Lead Toxicity

Chicago Data:  Gasoline Lead versus Blood Lead

New York City Data: Gasoline Lead versus
Blood Lead
 II-9


 11-10

 11-11

 11-12
II1-3


III-8


III-9


III-ll

II1-12


II1-13
CHAPTER IV:

    FIGURE IV-1


    FIGURE IV-2
Multi-Organ Impact of Lead's Effects               IV-2
on the Heme Pool

The Relationship Between MCV and Lead After
Adjusting for All Other Significant Variables      IV-14

-------
                          list of Figures, page 2
    FIGURE IV-3



    FIGURE IV-4



    FIGURE IV-5



    FIGURE IV-6



    FIGURE IV-7

    FIGURE IV-8
The Relationship Between MCH and Blood Lead
After Adjusting for All Other Significant
Variables

Prediction of Percent Children with MCV < 74 fl
as a Function of Blood Lead, After Controlling
for All Other Significant Variables

The Relationship Between FEP and Blood Lead
After Controlling for All Other Significant
Variables

Prediction of Percent of Children with Anemia
as a Function of FEP Levels at Normal
Transferrin Saturation Levels

Effects of Lead on 10

Flow Diagram for Children with Blood Lead
Levels above 25 ug/dl
PAGE


IV-15



IV-17



IV-20



IV-23


IV-37


IV-49
CHAPTER V:

    FIGURE V-l


    FIGURE V-2


    FIGURE V-3
Adjusted Systolic Blood Pressure versus
Blood Lead

Adjusted Diastolic Blood Pressure versus
Blood Lead

Adjusted Rates of Death and Heart Attacks
versus Blood Pressure:  Framingham Data
 V-16


 V-17


 V-30
CHAPTER VIII:

    FIGURE VIII-1  Impact of Lead Levels in Misfueling Under         VIII-11
                   Three Assumptions

    FIGURE VIII-2  Net Benefits (Including Blood-Pressure-           VIII-12
                   Related Effects) as Functions of Lead Level
                   and Misfueling

    FIGURE VIII-3  Net Benefits (Excluding Blood-Pressure-           VIII-13
                   Related Effects) as Functions of Lead and
                   Misfueling

-------
                        EXECUTIVE SUMMARY


     The Environmental Protection Agency (EPA) is promulgating a
rule to reduce the amount of lead in gasoline from its current
limit of 1.10 grams per leaded gallon (gplg) to 0.50 gplg on July
1, 1985 and to 0.10 gplg effective January lf 1986.  EPA's primary
objective in promulgating this rule is to minimize the adverse
health and environmental effects of lead in gasoline.  To
increase flexibility in meeting the phasedown schedule, EPA has
proposed to allow refineries that reduce lead ahead of schedule
in 1985 to "bank" those extra lead rights for use in 1986 or
1987.  The Agency also is considering the possibility of a
complete ban on leaded gasoline, to take effect as early as 1988.
This Regulatory Impact Analysis addresses only the final
phasedown rule; a separate document summarizes the costs and
benefits of a possible ban.


Basis for Action

     Section 211(c)(l) of the Clean Air Act gives EPA's
Administrator broad authority to "control or prohibit the
manufacture ... or sale of any fuel additive" if its emission
products (1) cause or contribute to "air pollution which may be
reasonably anticipated to endanger the public health or welfare,"
or (2) "will impair to a significant degree the performance of
any emission control device or system ... in general use."
Reductions in the lead content of gasoline are justified under
both of these criteria.

     Lead in gasoline has been shown to increase blood lead
levels, which in turn have been linked to a variety of serious
health effects, particularly in small children.  Recent studies
linking lead to blood pressure in adult males also are a source
of concern about the health effects of lead in gasoline; because
these studies have just been published, however, EPA will not
rely upon that evidence until there has been a greater
opportunity for scientific review and public comment.

     Lead in gasoline also impairs the effectiveness of
pollution-control catalysts.  A 1983 EPA survey of vehicles in
use showed that about 15.5 percent of the vehicles that should
use unleaded gasoline to protect the effectiveness of their
pollution-control catalysts are misfueled with leaded gasoline,
resulting in significant excess emissions of hydrocarbons (HC),
nitrogen oxides (NOX), and carbon monoxide (CO).  In addition to
these health and environmental effects, reducing lead in gasoline
will reduce vehicle maintenance costs associated with the
corrosive effects of lead on engines and exhaust systems, and
will improve fuel economy.

-------
                              E-2


     EPA has considered a variety of alternatives for reducing
the health and environmental problems caused by lead in gasoline,
and has concluded that the most effective approach is to reduce
the amount of lead in gasoline as quickly as possible.  EPA is
confident that the phasedown schedule mandated by the rule can be
met by the refining industry with existing equipment.  The Agency
is not promulgating a complete ban on lead in gasoline at this
time because of continuing concern about the possible effects of
such a ban on certain engines that may rely on lead for
protection against valve-seat recession.  EPA will consider the
valve-seat issue, along with the recent evidence on lead and
blood pressure, in reaching a final decision on a ban.


Costs of Reducing Lead in Gasoline

     Since the 1920s, petroleum refiners have added lead to
gasoline as a relatively inexpensive way of boosting octane.  To
meet octane requirements with less lead, refiners have several
options, including additional processing in reforming or
isomerization units and the use of alternative additives, such
as MMT or alcohols.  At the margin, however, each of these
alternatives is more expensive than lead for producing octane.
Higher refining costs constitute virtually all of the rule's
social costs.

     We estimated the additional refining costs using a model of
the petroleum refining industry developed for the Department of
Energy (DOE).  This model employs a linear programming framework
to represent U.S. refining operations, and can find the minimum-
cost method for producing a particular product slate subject to
various constraints (including the amount of lead allowed in
gasoline).  The model has been used by both EPA and DOE in
previous rulemakings, and several verification checks have
indicated that it performs well.

     To estimate the costs of alternative rules, we first ran the
model specifying the existing lead limit of 1.10 gplg.  We then
ran the model with a tighter constraint on lead.  In both types
of runs, the model also was constrained to meet projected demands
for gasoline and other refined products.  The difference between
the costs at the two lead limits is the estimated cost of the
rule.  This procedure limits the possibility of underestimating
costs if the model assumes more flexibility than is in fact
possible; any overoptimization affects both estimates, and thus
has little impact on the difference between the two.  In
addition, we placed many constraints on the model to reflect
real-world limitations in the ability of the refining industry to
fully optimize production.

-------
                             E-3
     Our base case results suggest that the final rule will cost
less than $100 million for the second half of 1985, when the 0.50
gplg limit will apply.  For later years, when the 0.10 gplg limit
will apply, the estimated costs range from $608 million in 1986
to $441 million in 1992.  (The estimated costs fall over time
because of projected declines in the demand for leaded gasoline,
even in the absence of this new rule.)

     We ran extensive sensitivity analyses to probe the limits of
feasibility.  Those analyses focused on 1986, because that is the
first year in which the 0.10 gplg will apply, and refineries will
not be able to undertake substantial new construction by then.
Those sensitivity analyses show that the rule remains feasible
(i.e., product demands can be met with existing refining capacity)
under most conditions.  Only when extremely unlikely combinations
of multiple adverse conditions are assumed does feasibility
appear to be in doubt, and then only for the peak-demand summer
months.  Additional sensitivity analyses show that even in those
worst-worst cases, the 0.50 gplg limit in 1985 does not approach
infeasibility.

     We also examined the potential impact on costs of the
banking rule that EPA recently proposed and which it may
promulgate shortly.  Under the banking rule, refineries that
reduced their lead use below applicable limits in 1985 (1.10 gplg
prior to July 1 and 0.50 gplg for the second half of the year)
could bank those extra reductions and use them in 1986 or 1987.
Banking would allow refineries to follow their own least-cost
schedules of lead reduction, so long as their total usage over
the three years did not exceed the amount allowed by this rule.
Because banked rights would be freely transferable among
refineries, they also would increase individual refineries'
flexibility by allowing those refineries with relatively high
costs to buy rights from refineries with lower costs.  Our
estimates suggest that banking would reduce the present value of
the rule's cost by about $200 million over the 1985-to-1987
period.  Moreover, it appears that banking would alleviate the
potential infeasibility found in the most extreme sensitivity
analyses.


Benefits of the Rule

     We estimated benefits in four major categories:
(1) children's health and cognitive effects associated
with lead; (2) blood-pressure-related effects in adult males due
to lead exposure; (3) damages caused by excess emissions of HC,
NOX, and CO from misfueled vehicles; and (4) impacts on mainten-
ance and fuel economy.  In each category, our estimates do not
cover all of the likely benefits because of gaps in the data or
difficulties in monetizing some types of benefits.  Nonetheless,
the estimates are substantial, and far exceed the costs.

-------
                             E-4


     Human exposure to lead from gasoline.  To predict the lead-
related health effects of the rule, we began by estimating its
impact on lead in individuals' blood.  People are exposed to lead
from gasoline through a variety of routes, including direct
inhalation of lead particles when they are emitted from vehicles,
ingestion of lead-contaminated dust or inhalation of such dust
when it is stirred up, and ingestion of food that has been
contaminated with lead.  Although it is difficult to estimate the
contributions of these individual pathways, several large data
sets make it possible to estimate the total contribution of lead
in gasoline to concentrations of lead in human blood.

     These data sets include records of lead-screening programs
from the Centers for Disease Control (CDC), records from
screening programs in individual cities, and, most importantly,
the Second National Health and Nutrition Examination Survey
(NHANES II), which provides blood lead measurements (and other
important information) on a large representative sample of the
U.S. population surveyed during the late 1970s.  By linking these
data to data on gasoline lead use, it is possible to estimate
statistically how blood lead levels respond to changes in
gasoline lead.

     Several studies have shown remarkably strong and consistent
relationships between gasoline lead and blood lead.  Figure E-l
plots those two measures over time using data from NHANES II.
Note how strong the relationship is; blood lead tracks both the
seasonal variations in gasoline lead (rising during the summer
months, when more gasoline is used) and the long-term downward
trend in gasoline lead.  Multiple regression analyses show that
this relationship continues to hold after controlling for other
factors (such as socioeconomic status, nutritional factors, and
exposure to other sources of lead).  Such studies suggest that
during the 1970s, gasoline was responsible on average for about
half of the lead in blood (other sources of lead include lead
paint, stationary sources, and lead solder in cans).
Experimental studies, where the isotopes in gasoline lead have
been modified so that its presence in blood can be distinguished
from lead from other sources, have yielded similar conclusions.

     Statistical analyses indicate that gasoline lead not only
raises the average level of lead in blood, but also contributes
substantially to the incidence of lead toxicity in children.
Based on an analysis of NHANES II, we predict that the 0.10 gplg
limit will roughly halve the number of children with blood lead
levels above those recognized as harmful.  Since 1978, the CDC
has recommended that children with blood lead levels above 30
micrograms per deciliter (ug/dl) receive follow-up testing
and possible treatment.  The CDC recently reduced that recom-
mended level to 25 ug/dl.  We estimate that in 1986 alone,
the rule will prevent 172,000 children from exceeding 25 ug/dl
blood lead.

-------
FIGURE E-l
   110-1
100-
  c
  to
  8

  a
    70-
  •0
  w
  5d
  M
  O

  0 60
  o
  o
  o
  £
  3
  in
50-
      LEAD USED IN GASOLINE PRODUCTION  AND

      AVERAGE  NHANES II  BLOOD  LEAD  LEVELS

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

                       GASOUNE
     AVERAGE

     BLOOD

     LEAD LEVELS
•/
          1976
                1977
                1978

                YEAR
                                   1979
1980
                                                    16  g
                                                    14
                                                       13
                                                        12
                                                    11
                                                    10
                                                        0
                                                       g
                                                       H-
                                                       n
                                                       H
                                                       o

                                                       M

                                                       §
                                                       Ul
                                                          ID
                                                          n
                                             c*
                                             ro

-------
                             E-6


     Children's health and cognitive effects.  Elevated blood
lead levels have been linked to a wide range of health effects,
with particular concern focusing on young children.  These
effects range from relatively subtle changes in biochemical
measurements at low doses (e.g., 10 ug/dl) to severe retardation
and even death at very high levels (e.g., 100 ug/dl).  Lead can
interfere with blood-forming processes, vitamin D metabolism,
kidney functioning, and neurological processes.  The negative
impact of lead on cognitive performance (as measured by IQ tests,
performance in school, and other means) is generally accepted at
moderate-to-high blood-lead levels (30 to 40 ug/dl and above).
Several studies also suggest cognitive effects at lower levels.
Changes in electroencephlogram readings have been found at levels
as low as 10 to 15 ug/dl.

     For children's health effects, we estimated benefits in two
categories:  medical care for children exceeding the new CDC
cutoff and compensatory education for a subset of those children
who may suffer cognitive effects from exposure to lead.  These
estimates are conservative in that they do not include many
benefit categories (e.g., lasting health and cognitive damage not
reversed by medical treatment and compensatory education), nor do
they attribute any benefits to reducing lead levels in children
whose blood lead levels would be below 25 ug/dl even in the
absence of the rule.

     To estimate reductions in medical care expenses, we relied
on recently published recommendations for testing and treating
children with blood lead levels above 25 ug/dl.  Such treatment,
we estimate, costs about $900 per child over 25 ug/dl.  (This
average reflects lower costs for most of these children, but much
higher costs for the subset requiring chelation therapy.)

     The estimates for compensatory education assumed three years
of part-time compensatory education for roughly 20 percent of the
children above 25 ug/dl; that averages about $2600 per child
above that blood-lead level.  Thus we estimated a total of $3500
in monetized benefits for each child brought below 25 ug/dl.  Our
estimates of aggregate benefits in this category ranged from
about $600 million in 1986 to roughly $350 million in 1992.


     Blood-pressure-related benefits.  Lead has long been
associated with elevated blood pressure, but until recently most
of the studies have focused only on hypertension and relatively
high lead levels (typically found only in those occupationally
exposed to lead).  Two recent studies, however, have found a
continuous relationship between blood lead and blood pressure
using data from a large representative sample of the U.S.
population (NHANES II, the same data set used to estimate the
relationship between gasoline lead and blood lead).  These
studies show a strong relationship that has proved robust in the
face of exhaustive statistical tests involving many possible
confounding factors and alternative specifications of the model.

-------
                             E-7


     These findings, if verified, have importanb implications for
the benefits of the phasedown rule.  EPA has not relied upon them
in setting the final phasedown rule, because they are too recent
to allow widespread review and comment.  They will be an
important element, however, in the decision on a final ban, so we
present estimates of blood-pressure-related benefits in this RIA
for informational purposes.

     To calculate the benefits, we used logistic regression
equations estimated from NHANES II to predict how the rule would
affect the number of hypertensives.  These estimates cover only
males aged 40 to 59, because the effect of lead on blood pressure
appears to be strongest for men and because estimates for that
age range are not confounded by a strong covariance between age
and blood lead.  We estimate that the rule will reduce the number
of hypertensives in that group by about 1.8 million in 1986.  We
valued reductions in hypertension based on estimates of the costs
of medical care, medication, and lost wages; they yielded a value
of $220 per year per case of hypertension avoided.

     We also estimated how reductions in blood pressure would
affect the incidences of various cardiovascular diseases, based
on our projections of changes in blood pressure as a result of
the rule and estimates of the relationships between blood
pressure and heart attacks, strokes, and deaths from all causes.
The latter estimates were derived from several large-scale
epidemiological studies, primarily the Framingham study.  Because
those studies included very few nonwhites, we restricted our
estimates to white males aged 40 to 59.

     We valued reductions in heart attacks and strokes based on
the cost of medical care and lost wages for nonfatal cases (the
fatalities from heart attacks and strokes were included in the
estimate of deaths from all causes).  That procedure yielded
benefits of $60,000 per heart attack and $44,000 per stroke
avoided.  It is important to note that these estimates do not
account for any reductions in the quality of life for the victims
of heart attacks and strokes (e.g., the partial paralysis that
afflicts many stroke victims).

     Valuing reductions in the risk of death is difficult and
controversial, with a wide range of estimates in the literature.
EPA's RIA guidelines, for example, suggest a range of $400,000 to
$7 million per statistical life saved.  For our point estimate,
we used a value from the lower end of that range, $1 million per
case.  The benefits of reduced mortality dominate our estimates
of total blood-pressure-related benefits, which range from almost
$6 billion in 1986 to about $4.5 billion in 1992.


     Benefits of reducing pollutants other than lead.  Reducing
the amount of lead in gasoline should decrease emissions of
several pollutants in addition to lead.  Most of these reductions

-------
                                E-8


will result from less misuse of leaded gasoline in vehicles that
should use unleaded to protect the effectiveness of their pollu-
tion-control catalysts.  EPA expects the rule to significantly
reduce misfueling because it will be more expensive to produce
leaded regular gasoline (at 89 octane) with 0.10 gplg than to make
unleaded regular (at 87 octane).  As a result, the gap between the
retail prices of unleaded and leaded regular gasolines should
narrow, and may well reverse.  (At present, leaded regular is
slightly cheaper to make than unleaded regular, but the retail
price differential is much larger than the manufacturing cost
differential, apparently because of marketing strategies by
retailers.)

     Eliminating misfueling would substantially reduce emissions
of HC, NOX, and CO.  All three of these pollutants have been
associated with damages to health and welfare, and contribute to
ambient air pollution problems covered by National Ambient Air
Quality Standards (NAAQS).  To predict the emission reductions
that would be associated with eliminating misfueling, we used
survey data on the extent of misfueling, tests showing the effect
of misfueling on emissions per mile traveled, and estimates of
the numbers of miles traveled by vehicles of different ages and
types.  The rule also should reduce emissions of benzene (a
hydrocarbon that has been associated with leukemia) and ethylene
dibromide (a suspected human carcinogen, which is added to leaded
gasoline to control excess lead deposits in engines).

     We valued these reduced emissions in two ways.  The first
involved direct estimation of some of the health and welfare
effects associated with these pollutants.  This method is
conceptually correct, but suffers from various uncertainties and
the inability to generate estimates for some potentially
important categories.  Virtually all of the benefits that we
could quantify were associated with projected declines in ozone
as a result of reductions in HC and NOX emissions.

     Our second method valued the emission reductions based on
the implict cost per ton controlled of the emission control
equipment destroyed by misfueling.  The final estimate, used in
computing total and net benefits, was the simple average of the
two different methods.  Assuming that the 0.10 gplg rule will
eliminate 80 percent of misfueling, those estimates range from
$222 million in 1986 to $248 million in 1992.  (The estimates for
this category increase over time because the total amount of
misfueling is projected to increase in the absence of the rule.)

-------
                             E-9


     Maintenance and fuel economy benefits.  Reducing lead in
gasoline will decrease vehicle maintenance expenses associated
with the corrosive effects of lead and  its scavengers on engines
and exhaust systems.  The rule also is  likely to increase fuel
economy, both because additional reforming to replace the octane
now provided by lead will increase the  energy content of gasoline
and because leaded gasoline fouls oxygen sensors in newer
misfueled vehicles.

     Three categories of maintenance benefits were estimated:
exhaust systems, spark plugs, and oil changes.  Estimates for the
first two categories were based on fleet studies of vehicles in
use, which showed that exhaust systems  and spark plugs last
longer with unleaded than with leaded gasoline.  Estimates of oil
change benefits were based on studies showing that oil maintains
its quality longer with unleaded than with leaded.  Summing
together these three categories, we estimate that reducing lead
in gasoline from 1.10 gplg to 0.10 gplg will yield benefits of
$0.0017 per vehicle mile, or about $17  per year for a vehicle
driven 10,000 miles.  Because of the large number of vehicles
affected, the aggregate benefits are large, ranging from about
$900 million in 1986 to roughly $750 million in 1992.

     The fuel economy estimate, as noted above, has two
components.  To estimate the gain in fuel economy due to higher
energy content, we used the change in fuel density predicted by
the DOE refining model and applied it to a fuel economy formula
developed by the Society of Automotive  Engineers.  To estimate
the portion due to reduced fouling of oxygen sensors, we
estimated the change in the number of misfueled sensor-equipped
vehicles and used experimental data on  how much extra fuel is
consumed by vehicles with fouled sensors.  Total estimated fuel
economy benefits exceed $100 million in most years.


Costs and Benefits of Alternatives

     Table E-l summarizes several important non-monetary measures
of the benefits of the 0.10 gplg standard for a single year,
1986.  These estimates assume that the  rule will eliminate all
misfueling.  We also examined a wide range of alternative
standards, however, and considered various assumptions about the
impacts of those rules on misfueling, ranging from the rule
eliminating all misfueling to the rule  having no impact on
current misfueling rates.  In addition, we computed net benefits
with and without the preliminary estimates of blood-pressure-
related benefits.

     Regardless of the assumption about misfueling, and whether
or not the blood-pressure-related benefits were included, we
found that net benefits were maximized  at the tightest of the
alternative standards considered, 0.50  gplg for the second half

-------
                             E-10
TABLE E-l.  Nonmonetary Measures of Health and Environmental
            Benefits of 0.10 gplg in 1986, Assuming No Mis-
	fueling	

Reductions in thousands of
  children above selected
  blood lead levels	

    30 ug/dl                                         52

    25 ug/dl                                        172

    20 ug/dl                                        563

    15 ug/dl                                      1,726


Reductions in thousands of
  tons of pollutants	

    Hydrocarbons                                    305

    Nitrogen oxides                                  94

    Carbon monoxide                               2,116


Reductions in adult male
  health effects	

    Thousands of
      hypertensives                               1,804

    Myocardial
      infarctions                                 5,350

    Strokes                                       1,115

    Deaths                                        5,160

-------
                             E-ll


of 1985 and 0.10 gplg for 1986 and subsequent years.  Table E-2
presents year-by-year estimates of the costs and benefits of the
final rule under our "partial misfueling" case, which assumes
that misfueling will continue unabated under the 0.50 gplg
standard in 1985, and then will be reduced by 80 percent under
the 0.10 gplg standard in 1986 and subsequent years.  The net
benefits without blood-pressure-related effects exceed $1 billion
per year in 1986 and later years.  If the blood-pressure-related
benefits are included, net benefits exceed $5 billion per year.
Although many of the individual components of these estimates are
subject to uncertainty, the magnitude of the estimated monetized
benefits relative to the costs, together with the many
potentially important unmonetized benefits, indicate that rapid
reductions of lead in gasoline are amply justified.

-------
                                       E-12


TABLE E-2.  Year-by-Year Costs and Monetized Benefits of Final Rule, Assuming
	Partial Misfueling (millions of 1983 dollars)	

	1985    1986    1987    1988    1989    1990    1991    1992

MONETIZED BENEFITS

  Children1s
    health effects    223     600     547     502     453     414     369     358


  Adult blood
    pressure        1,724   5,897   5,675   5,447   5,187   4,966   4,682   4,691


  Conventional
    pollutants          0     222     222     224     226     230     239     248


  Maintenance         102     914     859     818     788     767     754     749


  Fuel economy         35     187     170     113     134     139     172     164


  TOTAL MONETIZED
    BENEFITS        2,084   7,821   7,474   7,105   6,788   6,517   6,216   6,211
TOTAL REFINING
  COSTS                96     608     558     532     504     471     444     441


NET BENEFITS        1,988   7,213   6,916   6,573   6,284   6,045   5,772   5,770


NET BENEFITS
  EXCLUDING BLOOD
  PRESSURE            264   1,316   1,241   1,125   1,096   1,079   1,090   1,079

-------
                            CHAPTER I




                           INTRODUCTION






     The Environmental Protection Agency  (EPA) is promulgating a



rule to sharply reduce the lead content of gasoline from its




current limit of 1.10 grams per leaded gallon (gplg) to 0.10 gplg.



The phasedown will take effect in two steps:  effective July 1,



1985, the limit will be 0.50 gplg; effective January 1, 1986, the



limit will be 0.10 gplg.  Although EPA believes that these



standards are feasible for the industry as a whole, to increase



individual refineries' flexibility in meeting that schedule,



EPA has proposed a rule change to allow refineries to reduce



lead use below allowable limits in 1985 and "bank" those credits



for use in 1986 and 1987.



     EPA's goal in promulgating this phasedown schedule is to



minimize the adverse health and environmental impacts of lead in



gasoline.  To further aid in meeting that goal, the Agency is



considering the elimination of all lead in gasoline.



     The health and environmental consequences of lead in gasoline



include both the direct health effects of exposure to lead and the



effects of higher emissions of conventional pollutants (hydrocar-



bons, nitrogen oxides, and carbon monoxide) from vehicles whose



pollution-control catalysts have been poisoned by the misuse of



leaded gasoline.  In addition to these health and environmental



benefits, which form the basis for this rule, reducing lead in



gasoline will provide benefits to vehicle owners in the form of




increased fuel economy and reduced maintenance expenditures for



lead-induced corrosion of engines and exhaust systems.  On the

-------
                               1-2






other hand, lead is a low-cost octane booster, so the proposed




rule will raise the cost of producing gasoline.



     EPA has determined that both the final phasedown rule and



the possible ban are "major" regulatory actions under the criteria



of Executive Order 12291, because higher gasoline production



costs would exceed $100 million per year under each action.  For



major rules, the Executive Order requires a Regulatory Impact



Analysis (RIA); this document constitutes the final RIA for the



phasedown being promulgated.  Although EPA has not relied on



banking in analyzing the costs or establishing the feasibility



of the final phasedown, this RIA also examines the impact of the



banking proposal on the costs and benefits of the phasedown



rule.  A separate preliminary RIA has been prepared for the ban,



now under consideration for as early as 1988; it employs the



same methods used in this document, and summarizes the costs and



benefits of a ban.  That preliminary RIA incorporates by reference



large parts of this final RIA.



     The remainder of this chapter provides an overview of the



problem of lead in gasoline, a brief review of earlier rulemakings



and the proposal made in August 1984, an analysis of the rationale



for government intervention, and a discussion of alternative



regulatory approaches.






I.A.  Background Information




     The rules being promulgated and proposed are part of a



series of actions taken by EPA over the past eleven years to



address the health and environmental hazards posed by lead in



gasoline.  This section provides some background information

-------
                               1-3






for these  latest actions.






I.A.I.  Previous Rulemakings




    EPA has regulated lead  in gasoline since 1973 to meet two



goals:  assure the availability of unleaded gasoline for those




vehicles with pollution-control devices  (catalysts) that are



rendered ineffective by leaded fuel, and reduce the adverse



health effects associated with exposure  to lead.



      EPA's original rule for lead in gasoline limited the lead



content per gallon averaged over all gasoline (both leaded and



unleaded)  sold by each refinery (38 FR 33741; December 6, 1973).



EPA also temporarily set separate and less stringent interim



limits for small refiners.



      In 1982, EPA promulgated new rules  (47 FR 49331; October 29,



1982) that: (1) switched the basis of the standard from all



gasoline to leaded only, (2) set a limit of 1.10 gplg, (3) phased



in uniform treatment of all refiners regardless of size, and (4)



allowed "constructive averaging" or "trading" of lead use across



refineries (e.g., one refinery could produce a gallon with 1.20



grams of lead if it traded with another  refinery that produced a




gallon with only 1.00 gram).



     The purpose in switching from a standard based on the



overall pool average to one based only on leaded gasoline was to




reduce total lead usage as sales of leaded gasoline declined



with the retirement of the older vehicles allowed to use it.



(With the overall gasoline-pool standard, refiners would have




been allowed to increase the amount of lead per gallon of leaded




gasoline as the total demand for leaded  declined.)  When EPA

-------
                               1-4






promulgated the 1982 rule, it projected a steady and substantial




decline in total lead in gasoline over the next decade.






I.A.2.  Continuing Problems




     Since 1982, new studies and reanalyses have shown lead to be




a greater health risk than previously thought.  In addition,




total lead in gasoline has not declined as projected in 1982.




Figure 1-1 compares the 1982 lead projections with actual lead




use in 1983 and with current projections for later years.




     The major cause of higher lead levels than previously pro-




jected is misfueling, the use of leaded gasoline in vehicles




designed to use unleaded.  EPA's 1983 survey of vehicles in use




indicated that about 15.5 percent of vehicles required to use




unleaded gasoline were misfueled with leaded (U.S. EPA, 1984e).




By comparison, in the 1982 survey, the overall rate was 13.5




percent, with lower average rates in areas with Inspection and




Maintenance (I/M) mobile source enforcement programs (6.2




percent) than in areas without such programs (15.1 percent)




(U.S. EPA, 1983a).  The 1983 survey showed increased misfueling




in I/M areas, reducing the gap.  These estimates probably under-




state the true extent of misfueling because owners were not




required to submit their vehicles for testing.  Moreover, the




surveys show higher misfueling rates in older vehicles, so




misfueling rates are likely to grow in future years as the




average age of catalyst-equipped vehicles increases.




     Misfueling not only increases lead emissions, but by




poisoning pollution-control catalysts, it increases emissions




of hydrocarbons (HC), nitrogen oxides (NOX), and carbon monoxide

-------
                                      1-5
L.
o
0>
>x
E
2
o>

o
Q>
V)
O
0)
      70
      60
      50 -
      40 -
      30 -
      20 -
      10 -
                                           rrent Projection
                               1982  Proj«ctio
                 i       i        i       i        i       i       I        I
        1981   1982   1983   1984   1985   1986   1987   1988   1989    1990

                                         Year

         FIGURE 1-1.  Comparison of 1982 and. Current Projections of Lead Use in Gasoline

-------
                               1-6






(CO).  In vehicles equipped with three-way catalysts (virtually



all post-1981 cars), lead-induced poisoning of catalysts increases



emissions by 500 percent for HC, 100 percent for NOX, and 300 per-



cent for CO (U.S. EPA, 1983b).  Each of these pollutants is



regulated by vehicle emission standards.   Each is also covered



by a National Ambient Air Quality Standard (NAAQS) or contributes



to a pollution problem covered by a NAAQS (HC and NOX form ozone,




and NOX forms NO2).



     Continued use of lead in gasoline, whether to meet the



demands of misfuelers or of legal users,  poses a serious threat



to health.  Several  studies have shown a  strong relationship



between lead in gasoline and levels of lead in children's blood,



with blood lead levels following gasoline lead closely, tracking



seasonal fluctuations as well as long-term trends.  Analyses



using several different data sets show that this relationship



remains strong and statistically significant even when other



potentially confounding variables are controlled for using multiple



regression and other statistical techniques.  These analyses



show that gasoline lead is related to blood lead levels in both



adults and children.



     Analyses also show that gasoline lead not only increases



average blood lead levels, but also raises the number of children



with dangerously high blood-lead levels.   Statistical analyses




indicate that lead in gasoline accounts for about half of the



number of children above 30 raicrograms per deciliter (ug/dl),



until recently the level set by the Center for Disease Control



(CDC) for follow-up testing and possible treatment.  Lead paint,

-------
                               1-7






emissions from stationary sources, and other non-gasoline sources



of lead exposure account for the other children at high levels,



although some of these additional exposures, including lead in



food, partially reflect past emissions of lead from gasoline.



     The adverse health consequences of high levels of lead in



children are well accepted.  They include damage to the kidney,



the liver, the reproductive system, blood creation, basic cellu-



lar processes, and brain functions.  The CDC recommends that



children with blood lead levels above 25 ug/dl receive follow-up



testing and possible medical treatment; upon further testing,



about 70 percent of children with blood lead levels above 25



ug/dl are expected to be classified as "lead toxic" under CDC



criteria.



     Increasing evidence also points to health effects at blood



lead levels below 25-30 ug/dl.  These effects include inhibition



of certain enzyme activities, changes in EEC patterns, impairment



of heme synthesis, increases in levels of a neurotoxic chemical,



possible interference with neurotransmission processes, impairment



of vitamin D activity, and impairment of globin synthesis.



Several studies also have found indications of lead affecting



cognitive functions (as measured by IQ tests and other means), as



well as having other neurobehavioral effects, at levels well



below 30 ug/dl.  Although EPA has not reached definite conclusions



regarding the specific blood lead levels at which such effects



occur, the Clean Air Scientific Advisory Committee (CASAC) has



recommended that the Agency consider these studies in the



Criteria Document for the ambient air quality standard for lead.

-------
                               1-8





     Concerns about the health effects of ambient exposure to



lead traditionally have focused on children.   Although lead has



a variety of adverse effects on the health of adults, most of



the known effects appear not to be of substantial concern except



at blood-lead levels in excess of 30-40 ug/dl.  Recent analyses,



however, indicate a strong and robust relationship between blood



lead levels and blood pressure, with no apparent threshold.



Those findings have important implications for estimating the



benefits of reducing lead in gasoline, because high blood



pressure, in turn, is linked to a variety of  cardiovascular



diseases including hypertension, myocardial infarction and strokes,





I.A.3.  The August 1984 Proposal



     Concerns about the health risks posed by lead in gasoline



and about the growing misuse of leaded gasoline in catalyst-



equipped vehicles led EPA to consider additional restrictions on



the lead content of gasoline.  In August 1984, the Agency pro-



posed to reduce the allowable lead limit to 0.10 gplg, effective



January 1, 1986.  EPA also indicated that it  would consider



alternative phasedown schedules with more than one step that



would start earlier than 1986.  In addition,  it offered two



alternatives for the final elimination of lead in gasoline:  a



ban to take effect in the mid-1990s, or allowing market forces



to eliminate leaded gasoline as demand shrank.



     The proposal and the alternatives reflected EPA's desire-to



eliminate lead in gasoline quickly, tempered  by two concerns.



First, the Agency wanted to ensure that the phasedown schedule

-------
                               1-9





could be met without excessive costs or gasoline shortages due



to lack of time to build additional refining capacity.  The



Agency's analyses, based on the Department of Energy's (DOE)



refinery yield model, indicated that refineries could meet the



0.10 gplg standard with existing equipment, but in the proposal



EPA solicited comments and additional data on that issue.



     Second, EPA was concerned that eliminating lead in gasoline



altogether might pose a serious risk of premature valve-seat wear



for certain older engines.  These engines include those in older



automobiles (primarily those manufactured before 1971) and some



gasoline-powered trucks and off-road vehicles (including farm



equipment) that do not have hardened valve seats.  Dynamometer



and track tests have shown that such engines can suffer premature



erosion of valve seats ("valve recession") with unleaded gasoline



under severe conditions (sustained high speeds and heavy loads),



but studies of vehicles in normal use generally have failed to



find excessive wear.  However, based on the available studies,



the proposal allowed 0.10 gplg to provide a margin of safety for



those engines that might need lead to protect against undue



valve seat recession.





I.A.4.  Information Received After August 1984 Proposal



     Several pieces of information received after the August



1984 proposal have strengthened EPA's determination to reduce



lead in gasoline as quickly as possible.  The CDC has now lowered



its recommended cut-off level for follow-up testing (from 30 ug/dl



to 25 ug/dl) as well as its definition of lead toxicity.  The



CDC cautions, however, that even this new, lower level should

-------
                               1-10





not be regarded as "safe"; rather, it reflects tradeoffs between



protecting children against the adverse effects of lead and the



risks associated with treatment, as well as the limited ability



of screening tests to reliably measure blood-lead levels below



25 ug/dl.  The reduction in this cut-off level more than triples



the number of children at risk under CDC criteria.



     Additionally, comments and data from refiners, plus additional



sensitivity analyses using the DOE refining model, have reinforced



the Agency's conviction that the refining industry as a whole



can meet a 0.10 gplg standard within a year of promulgation



using existing equipment.  Moreover, these same sources indicate



that it is possible to achieve significant reductions even sooner.



Thus, the Agency has decided to add an interim standard of 0.50



gplg to take effect on July 1, 1985.



     Although EPA believes that both standards are feasible for



the industry as a whole, EPA's recent proposal to extend the



lead trading program to allow "banking" also will increase flexi-



bility and help to ensure that individual refineries can meet



the phasedown schedule in the most cost-effective manner.  Under



that proposed rule change, refineries using less lead than allowed



under the applicable limits in 1985 could "bank" those extra



lead rights for use in meeting the tighter limits in 1986 and



1987.  Banking should reduce significantly the overall costs



of meeting the phasedown limits and increase a refinery's



flexibility in meeting unexpected problems, without permitting



any additional use of lead over the three-year period, 1985



through 1987.

-------
                               1-11





     Three additional types of information have contributed to



EPA's decision to consider a ban on lead in gasoline by 1988.



First, the Agency has received data indicating that valve-seat



recession with unleaded gasoline may be a much less serious



problem than earlier feared.  The Agency has reviewed data from a



large test performed by the U.S. Army in the mid-1970s that in-



volved switching many types of vehicles — including heavy-duty



trucks, construction equipment, motorcycles, and stationary engines



(such as generators), as well as light-duty vehicles — from



leaded to unleaded gasoline.  The Army's study showed no detectable



increase in valve-seat problems and resulted in all of the armed



services switching to unleaded gasoline (where available) by 1976.



In addition, the U.S. Postal Service switched its heavy-duty



trucks to unleaded in 1980, and a review of their computerized



maintenance records shows no evidence of abnormal rates of



valve seat recession.  Moreover, reanalyses of other studies



suggest that lead may cause serious engine durability problems.



EPA is reviewing these and other studies relevant to this issue.



     Second, although EPA continues to believe that the phasedown



rule will have a significant impact on misfueling, it is less



confident that the problem will be eliminated so long as any



lead remains in gasoline.  Moreover, the 1983 EPA misfueling



survey, the results of which became available after the August



proposal, showed that misfueling continues to be a serious problem,



and that the rates appear to be increasing.



     Finally, two recently published studies (Pirkle et al., 1985;



Harlan et al., 1985) indicate a strong relationship between blood

-------
                                1-12





lead and blood pressure in adult males.  Because of the well-



established relationships between gasoline lead and blood



lead, and between blood pressure and cardiovascular disease,



these studies have potentially important implications for public



health.  Preliminary estimates based on Pirkle et al. indicate



that lead in gasoline may be responsible for well over one million



cases of hypertension per year and for over 5,000 deaths from



heart attacks, strokes, and other diseases related to blood pres-



sure.  Moreover, these estimates cover only males aged 40 to 59



and, in the case of heart attacks, strokes, and deaths, only white



males in that age group.  (The Pirkle et al. study did not find a



statistically significant association between lead and blood pres-



sure in females, and available studies of the effects of blood



pressure on cardiovascular risks provide the most reliable dose-



response estimates for whites in that age range.)  The benefits



may extend to older males and to nonwhites as well.



     The Pirkle et al. and Harlan et al. studies have b'een published



in peer-reviewed journals.  A summary of the Pirkle et al. study



also was placed in the docket of the phasedown rulemaking (Schwartz,



1984c).  Because the papers have only recently been published and



have not yet received widespread review by the scientific community,



EPA has not relied on estimates of health effects related to blood



pressure in deciding on the phasedown rule being promulgated now.



These effects will be considered, however, in reaching a decision



on a ban.  We have included a chapter in this RIA on the potential



impacts of the phasedown on blood pressure and cardiovascular



disease.

-------
                               1-13





I-B.  Need for Government Intervention



     Lead in gasoline is a classic example of what economists



call a negative externality; individual users of leaded gasoline



do not bear all of the costs it imposes on society as a whole.



Users of leaded fuel reap short-term savings in the form of lower



fuel costs; they also bear higher maintenance costs.  Individually,



however, they bear only an infinitesimal fraction of the costs of



the health and environmental damages caused by their vehicles'



emissions of lead and, in the case of misfuelers, other pollutants



from poisoned catalysts.  This disparity between private and social



costs generates an overuse of the good, in this case lead in gaso-



line, and increases the damages imposed on society.



     The need for government intervention to rectify significant



negative externalities is well recognized and provides the pri-



mary intellectual basis for virtually all EPA regulations.  The



predominant approach in the United States has been to impose



standards that limit the level of the externality (in this case,



the amount of lead permitted in gasoline).  Market-oriented



approaches — such as pollution taxes or charges on emissions and



marketable permits — are rarely used, although they may be highly



efficient means for reducing negative externalities.  These alter-



native approaches are discussed more fully in later sections of




this chapter.



     Section 211(c)(l) of the Clean Air Act [42 U.S.C. §7545(c)



(1)] gives the Administrator of EPA broad authority to "control or



prohibit the manufacture ... or sale of any fuel additive" if its



emission products (1) cause or contribute to "air pollution

-------
                               1-14






which may be reasonably anticipated to endanger the public




health or welfare", or (2) "will impair to a significant degree



the performance of any emission control device or system ...




in general use ..."



     EPA believes that further reductions in the lead content of



gasoline are justified by both of the tests under Section 211(c)(l)



Lead in gasoline has been shown to raise blood lead levels, which



endangers public health,  and the misuse of leaded fuel damages



pollution control devices, substantially increasing emissions of



HC, NOX, and CO.






I.C.  Alternatives to New Regulations



     Regulation of lead in gasoline is amply supported both by



statutory authority and the significant negative externalities



associated with leaded gasoline.  In this instance, however, EPA



already has regulations in place, so the issue is not whether



there should be any regulation at all, but whether the regulations



should be tightened.  Before proposing stricter rules, the Agency



carefully considered alternative approaches that would not re-



quire new rules.  Three such alternatives — no change in policy,



public education efforts, and stepped-up enforcement against



misfueling — are discussed in this section.






I.C.I.  No Change in Policy




     Under this approach, the Agency would not change existing



regulations and policies  regarding lead in gasoline.  Lead use



would decline over time,   as existing vehicles designed to use

-------
                               1-15





leaded gasoline retired and the number of vehicles designed to



use unleaded increased.  As discussed earlier, however, lead use



would decline less rapidly than expected at the time of the



previous rulemaking, and the number of misfuelers would grow.



Our best estimate is that leaded fuel would not decline below 20



percent of total gasoline fuel use under this scenario.



     All of the benefit-cost calculations in later chapters



implicitly evaluate this alternative, because it is the baseline



from which costs and benefits are measured; by definition, it has



zero costs and benefits, and thus zero net benefits in each year.



Policies with positive net benefits yield higher net benefits than



this alternative.  Conversely, policies with negative net benefits



are less efficient, in economic terms, than the status quo.





I.C.2.  Public Education



     Efforts to educate the driving public about the problems



caused by misfueling offer potentially useful supplements to



current policy.  As discussed more fully in Chapter VII, the use



of leaded gasoline increases vehicle maintenance costs.  Indeed,



EPA estimates that the maintenance savings of reducing lead in



gasoline would exceed the higher costs of manufacturing unleaded



fuel.  EPA is undertaking a variety of initiatives to inform



vehicle owners of these maintenance costs, as well as the



adverse environmental effects of misfueling.



     Public education efforts, while useful, are unlikely to



significantly reduce the use of leaded gasoline, in large part



because retail price differentials between leaded and unleaded

-------
                               1-16






gasoline are high, roughly three to four times higher than the



manufacturing cost differentials.   (See Chapter II for further



discussion of this issue.)  In addition, the health and environ-



mental costs of using leaded gasoline are externalities that are



not borne by individual users of leaded gasoline.  Thus, even if



the social benefits of reducing lead in gasoline exceed the social



costs by a large margin, the strictly private benefits (reduced



maintenance costs) to most individual purchasers may be less than



their private cost (the retail price differential).





I.C.3.  Stepped-Up Enforcement



     Federal law makes it illegal  for service stations and other



gasoline dispensing outlets to put leaded fuel in vehicles



requiring unleaded.  Federal law does not apply to individual



vehicle owners who misfuel, although many states have such laws.



In theory, therefore, the misfueling problem could be mitigated



by stepped-up enforcement of existing state and federal laws.



     Unfortunately, however, massive enforcement efforts would



be very expensive, and only partially successful in eliminating



misfueling.  Moreover, even if it  were possible to eliminate



misfueling, the serious health effects associated with legal use



of leaded gasoline would continue  unabated.



     Misfueling presents a serious enforcement problem.



Currently, almost 100 million vehicles require unleaded fuel, and



there are over 100,000 retail gasoline outlets, plus even more



private outlets (e.g., private tanks used to fill commercial

-------
                               1-17





fleets).  Inspection and Maintenance (I/M) mobile source enforce-



ment programs test the emissions of individual cars, requiring



owners to make repairs if their cars exceed emission standards.



I/M programs appear to offer the most practical means of enforce-



ment, but less than one-fifth of all vehicles are in areas with



such programs.




     I/M programs are used in areas to help achieve compliance



with National Ambient Air Quality Standards, primarily by encour-



aging the improved maintenance of vehicles and their emission



control equipment.  They are an important part of strategies to



attain air quality standards, and they produce substantial emis-



sion reductions  independent of their effects on misfueling.



Extending I/M programs to areas already attaining standards would



be expensive; a  typical program costs about $6.50 per vehicle



inspection (U.S. EPA, 1981).  By the mid-to-late 1980s, more than



100 million vehicles requiring unleaded gasoline will be registered



in areas that do not currently have I/M programs.  If we assume



annual inspections, the cost of extending I/M programs to cover



all unleaded vehicles would be about $650 million per year, which



is higher than the estimated cost of reducing lead to 0.10 gplg




in all leaded gasoline.



     This estimate understates the cost of an I/M approach because



it does not include the costs of required repairs, the time spent



by owners to have their vehicles inspected, or the higher costs



of manufacturing additional unleaded gasoline to meet the demand



of existing and potential misfuelers deterred by inspections.



These categories of costs are likely to be substantial.

-------
                               1-18






     A national I/M program also is likely to miss many misfuelers.



Surveys show that while misfueling rates are lower in areas with




I/M programs than in those without, they are still significant;



the average misfueling rate in I/M areas appears to be about two



thirds of the overall national average,  but the difference has



been shrinking (see Chapter VI).  An I/M program aimed speci-



fically at misfueling (e.g., including tests for lead in tailpipes)



probably would be more successful, but even an extensive I/M



program is unlikely to solve the misfueling problem.



     More importantly, I/M programs and  other aids to enforcement



have no impact on lead emissions from vehicles legally permitted



to use leaded gasoline.   EPA estimates that in 1983,  legal users



accounted for about 85 percent of all leaded gasoline sales.  Over



time, with a decrease in the number of vehicles allowed to use



leaded and a rise in the number of vehicles designed  for unleaded,



that proportion will fall.  Even in 1988, however, EPA projects



that, in the absence of new rules, roughly two-thirds of the



demand for leaded gasoline will be legal.  Thus, we estimate that



even a fully effective enforcement program targeted on misfueling



could not reduce lead emissions from vehicles by more than 30



percent for the period from 1986 to 1990.






I.D.  Market-Oriented Alternatives




     Market-oriented approaches to environmental protection offer



the advantage of allocating control efforts on the basis of mar-



ginal control costs, yielding minimum costs for any given level




of overall protection.  The two major alternatives are marketable



permits and emission charges.

-------
                               1-19





I.D.I.  Marketable Permits



     EPA's current lead-in-gasoline regulations, by allowing



"constructive averaging," in effect set up a system of marketable



permits for lead in gasoline.  Refineries with lower-than-average



costs for producing octane without lead are allowed to reduce



their lead content below the limit of 1.10 gplg and to sell the



excess lead rights to refiners with higher-than-average costs, who



may then produce leaded gasoline with lead content above 1.10 gplg.



This approach offers clear advantages over a uniform standard of



1.10 gplg.  Lead is reduced by the same amount, but costs are



lower.  For those refineries that lower the lead content of their



gasoline, the sale of excess lead rights more than offsets higher



manufacturing costs.  Conversely, for those refineries that buy



lead permits, the savings in manufacturing costs more than offset



the purchase price of the permits.  Consumers presumably also



benefit when these lower costs translate into lower gasoline



prices.



     The current system is working well.  Based on confidential



reports filed with EPA by the industry, about 73 percent of



refiners participated in lead trading during the second quarter of



1984.  During that same period, about 14 percent of the total



amount of lead was traded.  Thus, the lead trading system initi-



ated in 1982 appears to be a major success story for introducing



market principles into environmental regulation.



     Under the new rule being promulgated, refineries will be



allowed to continue constructive averaging through 1985, until



the 0.10 gplg standard takes effect.  Once the 0.10 gplg standard

-------
                             1-20






applies, however, inter-refinery averaging no longer will be




allowed, because that standard is intended to protect those few



engines that may need a small amount of lead (or some other



additive) to protect against premature valve-seat recession.



Although EPA believes that the risk is small and that minor



fluctuations around this level over time will not damage engines



(hence, the rule permits quarterly averaging by refiners), con-



stant use of fuel with a lower lead content might damage a few



engines.  For a trading system to work, some refiners must produce




gasoline with a lower lead content than permitted by the rules,



but that could cause damage to some engines because many vehicle



owners consistently purchase fuel produced by the same refinery.



Moreover, the refinery cost savings from allowing trading at 0.10



gplg would be trivial, as the amount of lead involved will be so



small.




     EPA has proposed a rule change that will allow refineries



greater flexibility in meeting lead-phasedown goals without any



detrimental effects on health, the environment,  or those engines



that may need protection against valve seat recession.  Under



this proposal, refiners would be able to save some of the lead



that they are allowed to use during 1985 for use in 1986 or



1987.  Thus, for example, if a refinery produced five million



gallons of leaded gasoline at 0.90 gplg during the second quarter



of 1985 (when the limit will still be 1.10 gplg), and did not



sell the extra lead rights to other refineries,  it could "bank"



one million grams of lead ([1.10 - 0.90] x [5 million] = 1



million) for use in 1986 or 1987.  Similarly, a refinery that

-------
                               1-21





produced five million gallons at 0.40 gplg during each of the



last two quarters of 1985 (when the limit will be 0.50 gplg)



also would have one million grams available for supplemental use



in 1986 or 1987.  These banked lead rights could be used as



desired by the refinery in 1986 or 1987; it could, for example,



produce 10 million gallons at 0.20 gplg (rather than the 0.10



gplg standard that will apply then), or sell the million grams



to another refinery with higher marginal production costs.  This



change will increase refineries' flexibility (thus lowering



production costs) without increasing the amount of lead released



to the environment.





I.D.2.  Pollution Charges



     The second market-oriented alternative to mandatory lead



content reductions would be to impose a charge on gasoline based



on its lead content.  Ideally, the charge per gram of lead should



equal the marginal external damage caused by a gram of lead in



gasoline, thereby internalizing the health and environmental



damages caused by lead use.  With such a charge, the price of



leaded gasoline would rise, encouraging consumers to switch to



unleaded.  Refiners would have an incentive to reduce the lead



content of gasoline to the point where the marginal cost of



additional reductions would equal the charge.  As with marketable



permits, refiners would end up producing leaded gasoline with



different lead levels, depending on the refiners' marginal costs.



     Despite these attractive features, a charge has several



drawbacks and faces some obstacles.  The primary drawback is the



same as identified earlier with regard to permits:  leaded gaso-

-------
                               1-22






line that contains much less than 0.10 gplg will not protect those



few engines that may need valve lubrication.  The charge might be



modified in one of two ways to address this problem:  it could



be levied only on lead in excess of 0.10 gplg, or EPA could set a



minimum lead-content rule for leaded gasoline.  In either case,



EPA believes that an appropriate charge would be high enough to



drive virtually all leaded gasoline to the minimum level of 0.10



gplg.  If that happened,  the cost-minimizing advantage of this



approach would be lost; the effect would not be materially



different from a uniform standard of 0.10 gplg.



     Even with these limitations, a charge would still offer some



advantages.  It would give refiners additional flexibility over



the short run while they made needed adjustments in equipment and



operating practices.  If the charge were levied on all lead in



gasoline (not just that above 0.10 gplg), it also would provide



an additional disincentive for misfueling with 0.10 gplg gasoline



by raising its price.  In addition, it would discourage the use



of leaded gasoline in other vehicles unless required to prevent



valve-seat recession.



     One potential obstacle to levying a charge to reduce lead in



gasoline is that EPA may not have the necessary statutory author-



ity.  If new legislation, were needed, it almost certainly would



be accompanied by extensive debate about both the general principle



of using charges or taxes to control environmental hazards and the



specific tax rate to be set for lead, which could cause significant



delays in solving a serious environmental problem.  EPA believes



that the foregone benefits caused by such a delay would outweigh



any potential efficiency gains.

-------
                               1-23





I«E«  Alternative Standards



     EPA has considered a wide range of alternative phasedown



schedules.  These alternatives included both different lead levels



and different effective dates.  Chapters II through VII present



the methods used to estimate the costs and benefits of these



alternatives.  In each chapter, we survey the effects in question



and discuss the methods used to derive monetized estimates.  Each



chapter contains year-by-year estimates of the relevant effects



for three sets of rules, as summarized in Table 1-1:  (1) the



primary Proposal made in August (0.10 gplg in 1986), (2) the



sample Alternative schedule discussed in the August Notice of



Proposed Rulemaking, and (3) the Final Rule.  (Because the 1985



standard only applies to the second half of the year, our



estimates for that year are for half a year.)  In all cases, the



costs and benefits depend on the amount of misfueling eliminated.



The sample benefit estimates presented in Chapters III-VII assume



that all misfueling is eliminated.  The cost estimates in Chapter



II assume that the amount of misfueling declines linearly with



the level of the standard, from 100 percent of current levels at



0.50 gplg or above, to zero with a ban.  A range of alternative



assumptions about misfueling, and their impacts on the costs and



benefits, is explored in Chapter VIII.



     Throughout this document, we use a 10 percent real discount



rate to compute present values, and costs and benefits are expressed



in 1983 dollars.  The choice of a discount rate is controversial,



the subject of much debate among economists.  Generally, the



higher the discount rate, the lower the net benefits, because



costs usually are incurred sooner than benefits.

-------
                               1-24
TABLE 1-1.  Alternative Phasedown Schedules
Phasedown Schedules
Proposed
Alternative
Final
Lead Limit
1985* 1986
N.A. 0.10
0.50 0.30
0.50 0.10
(gpig)
1987
0.10
0.20
0.10

1988
0.10
0.10
0.10
* 1985 limit applies as of July 1.

-------
                               1-25





The rate we use, 10 percent, is at the high end of the ranges



typically discussed; many economists would regard a much lower



rate as more appropriate.  We have used 10 percent to be consist-



ent with guidelines from the Office of Management and Budget



(U.S. OMB, 1981).  We also considered presenting sensitivity



analyses using lower rates, but decided not to do so in the



interests of simplifying the presentation; lower discount rates



would increase our estimates of net benefits.



     Chapter VIII summarizes the costs and benefits of the final



phasedown rule and compares them to a wide range of alternative



schedules.  It also examines the impact of different levels of



misfueling under each of the alternatives.  Regardless of the



assumption about misfueling, the schedule contained in the final



rule appears to offer the highest net benefits of the options



considered.  Chapter VIII also examines the costs and benefits



of the recent proposal to allow banking of lead rights.

-------
                           CHAPTER II




               COSTS OF REDUCING LEAD IN GASOLINE






     Since the 1920s, petroleum refiners have added lead to gaso-



line as a relatively inexpensive way to boost octane.  To meet



octane demands with reduced amounts of lead, refiners have a



variety of options, the most  important of which is to perform



additional processing of gasoline components in reforming and




isomerization units.  In addition, refiners also may employ



additives other than lead, such as alcohols or MMT, or they may



purchase additional high-octane components, such as aromatics or



butane.  At the margin, however, all of these alternatives are



more expensive than lead for  making octane, so the cost of produc-



ing gasoline will  rise under  the rule being promulgated.



     These higher  refining costs will comprise virtually all of



the costs of the rule.  (The  one potential exception, damage to



the valve seats of some engines with unleaded gasoline, applies



only to the possible ban, not to the final phasedown rule; the



valve-seat issue is discussed in Chapter VII.)  To estimate those



costs, we used a model of the refining industry orignally developed




for the Department of Energy  (DOE).



     Section A of  this chapter explains why we estimated the social



cost of the rule using predicted changes in manufacturing costs,




rather than changes in retail prices.  Section B provides an over-



view of the DOE model and refinery processes generally.  Section C



discusses the input data and  assumptions used for our projections



and presents year-by-year cost estimates for reducing lead in



gasoline.  Section D reports  the results of numerous sensitivity

-------
                               II-2






analyses conducted to test the robustness of the cost projections.



Finally, Section E, for illustrative purposes, examines the impact



on costs of the proposal to allow "banking" of lead rights during




1985 for use in 1986 and 1987.





II.A.  Price versus Cost Differences



     The costs we are measuring are the social costs of reducing



lead in gasoline.  Our estimates of these costs are based on



estimates of changes in the costs of manufacturing gasoline (and



other petroleum products).  In the long run in a competitive



market, the change in manufacturing costs is likely to be fully



reflected in changes in the amounts paid by consumers.  In the



short run, however, the total amount paid by consumers may be



less than or greater than the change in manufacturing costs,



depending on supply and demand elasticities and other factors.



The divergence between the change in manufacturing costs and the



change in the amount paid by consumers, though, will be comprised



mostly of transfers from one segment of U.S. society to another,



rather than real social costs (or savings).



     By real social costs, we mean the costs of real resources



that are used to comply with the rule (i.e., the extra energy,



capital, labor, etc. that is needed to meet the tighter lead



standard).  If retail prices rise less than costs in the short



run, then the losses to refiners (and their shareholders) will




be offset by gains to consumers.  If prices rise more than costs,



then the losses to consumers will be offset by gains to producers.



In either case, the net cost to all citizens, including sharehold-



ers, is just the real resources used in meeting the rule.  Thus,

-------
                                II-3





so long as the rule does not affect the costs of distributing or



selling the product,  the change  in manufacturing costs is a good



approximation of the  social cost of the rule.



     The  DOE model estimates that at current lead levels (1.10



gplg) the marginal manufacturing cost differential between



unleaded  and leaded regular grades of gasoline is less than two



cents per gallon.  Retail prices, however, diverge by an average



of about  seven cents  per gallon  (Weekly Petroleum Status Report,



1984, various issues).  Some commenters suggested that the gap



between the model's estimate of  the cost differential and current



retail price differentials indicates either an error in the model,



or that it would be more appropriate to use the retail price



difference as the measure of social cost.  Both arguments have



some superficial plausibility, but fail to withstand careful



analysis.



     The much larger  retail price differential appears to reflect



marketing strategies  within the  industry rather than errors in the



model's cost estimates.  Price differentials at the wholesale level



are far smaller than  the retail  price spread, and discussions be-



tween EPA and petroleum refiners confirmed that manufacturing cost



differences between unleaded and leaded gasolines were about two



cents per gallon, and that inter-refinery trades occurred at a



differential of two to two-and-a-half cents per gallon.  Also, the



differential between  the spot prices of unleaded and leaded regular



gasolines in barge quantities has been between one and four cents



per gallon over the last several years.  At the end of 1983,



wholesale price differentials in the Gulf of Mexico were about

-------
                               II-4
three cents per gallon (Platt's Oilgram).



     The additional price gap between unleaded and leaded gasolines



at the retail level does not appear to reflect incremental distri-



bution or other costs.  Both types of gasoline are distributed



through the same network of pipelines, terminals, barges, and tank



trucks, and similarities in sales volumes  suggest that differential



inventory carrying costs do not apply-



     Two possible explanations of the price differential have been



offered.  The first is that service stations use their lowest cost



gasoline as the "fighting grade," prominently advertising it at a



low price to attract customers (including  those purchasing another



grade).  If this explanation is correct, unleaded regular may



become the new fighting grade,  as it will  be less expensive to



manufacture 87 octane unleaded than 89 octane leaded with 0.10



gplg.  (To produce 89 octane leaded gasoline at 0.10 gplg, the



octane level of the gasoline must be boosted to over 88 before



the lead is added.)  In that case, the price of unleaded gasoline



would be lower than leaded.  Service stations may continue to use



leaded gasoline as the fighting grade, however, because it is



the lower-volume product.




     The second possible explanation is that retailers are engaging



in price discrimination, taking a larger margin on the grade



(unleaded) with less elastic demand.  This explanation presupposes



that individual retail outlets do not face the flat demand curves



of a perfectly competitive market, but rather that gasoline sales



are characterized by monopolistic competition.  Under this scenario,

-------
                               II-5
the individual station owner  is a geographic monopolist, i.e., the



only retailer at that location.  Thus, the demand curve facing the



operator is not perfectly flat, hut  if the price is too high,



customers will drive to a competitor.




     Under such a market structure,  prices deviate from marginal



costs, roughly in inverse proportion to each product's demand



elasticity, but firms do not  reap economic profits (i.e., profits



beyond a normal return on capital) because of free entry of new



competitors.  If leaded customers are more price sensitive (per-



haps because they may have lower incomes and are more willing to



search out lower-price stations, or  because the leaded price is



displayed more prominently),  their demand elasticity (at the level



of the individual service station) will be higher than the elastic-



ity of demand for unleaded customers.  As a result, price discrim-



ination will cause the operating margin (price minus wholesale



cost) to be lower on leaded than on  unleaded.  If this explanation



is correct, the 0.10 gplg standard should greatly reduce the retail



price differential between unleaded  and leaded gasoline, though



it would not necessarily reverse it.



     Whichever of these explanations is correct, the difference in



manufacturing costs, rather than in  retail prices, wot'iv". represent



the social cost (as conventionally defined in cost-benefit



analysis) of reducing lead in gasoline.  This conclusion would




fail to hold only if it were  possible to show that reducing the



lead content of gasoline would increase the real costs of distrib-



uting gasoline.  No such evidence is available.

-------
                               II-6






II.B.  The Refinery Model




     This section discusses the DOE refinery model used to estimate




costs.  Section II.B.I provides a brief introduction to the model,




while Section II.B.2 gives an overview of refining processes and




how they are represented in the model.






II.B.I.  Introduction to the DOE Model




     The DOE refinery yield model uses a linear programming frame-




work to simulate refining operations (U.S.  DOE, 1984b).  The model




represents individual refinery units and their inter-relationships




using a series of about 350 equations.  Given a set of input




assumptions and constraints, it finds the least-cost method of




producing any specified set of final products.  (The model also




can be run to maximize profit given final-product prices.)




     In addition to estimating the total cost for a set of product




demands, inputs, and constraints, the model generates detailed and




useful information on important aspects of  industry operations,




including the rates at which costs change (the shadow prices of




the constraints) and the extent to which various types of equipment



are utilized under different scenarios.




     The model is based on many similar models developed and used




widely by the petroleum refining industry for its own planning




purposes.  The refining industry was among  the first to make




extensive use of linear programming.  The DOE model has been used




by EPA for several years in its analyses of the impacts of regu-




lations on the petroleum industry and on petroleum product pur-




chasers.  In the 1982 lead-in-gasoline rulemaking, EPA used the

-------
                               II-7






model to predict the costs of the rule and to predict the value




of lead rights under the trading program.  Although it is impos-



sible to verify directly the cost estimates made then, it is



important to note that the actual prices of traded lead rights



were slightly lower than predicted by the model, indicating that



it did not underestimate the value of lead to refiners as an



octane booster.




     The model also has been used by DOE for several purposes,



including:  evaluating crude mixes for the selection of storage



sites for the Strategic Petroleum Reserve, assessing the impacts



of petroleum disruptions on product supplies, and evaluating the



industry's capability to respond to changes in feedstock quality



or product demands.



     DOE subjected the model to two verification checks, and found



that for 1982 it generated a product slate very close to the actual



one.  It also closely predicted refinery-gate price differentials.



It is also important to note that these verifications covered a



period before EPA initiated interrefinery averaging of lead use,



indicating that the accuracy of the model's predictions is not



dependent on the trading of lead rights.  More recently, DOE



verified the model using 1983 data (U.S. DOE, 1985).






II.B.2.  Overview of Refining Processes



     So the reader can better understand how the model works, we




provide a brief description of refining processes.  Figure II-l



is a schematic diagram of a very simple refinery, often called a



topping plant, which processes low-sulfur crude oils.  A complex

-------
                               II-8






refinery contains distillation units and other types of processing




units;  Figures II-2, II-3,  and II-4 present schematics of such



refineries.  (The model contains considerably more detail than even



these exhibits indicate.)   In any given refinery, these different



process "units" are assembled into final structures that accomplish



different but related purposes.  The basic similarity of process



units makes it possible to model refineries.



     Basically, the model  is a system in which the various units



that make up all types of  refineries are represented by the boxes



in the schematics.  Each unit uses inputs (crude oil or an inter-



mediate product) to make one or more intermediate or final products.



The exact types and quantities of the product(s) made are functions



of the properties of the inputs of each unit and the process that



each performs.  Fuel and utilities (e.g., electricity and steam)



are consumed, and an operating cost is incurred for each operation.



A capital cost may or may  not be charged, as appropriate to the



particular analysis being  performed.  Table II-l provides a summary



of the basic types of refinery processes.  Appendix A contains a



more detailed description  of processing operations.



     Because all refineries are made up of these building blocks,



the smallest structure in  the model is a process unit rather than



a plant.  The individual functions that are modeled are the inputs



and outputs from each type of unit.  The model is made up of re-




finery units, each of which has an output (or a series of products),



the quantity of which is a function of the material that the unit



is "fed."  Some refinery process units incur some costs that vary



with how intensely they are run — called "severity".  In the case

-------
                                         II-9


                                    Figure  n-i


                   Schematic Diagram of a  Bimple  Oil Refinery


                                   (Topping  Plant)
     FLOW DIAGRAM OF TOPPING REFINERY

     PROCESSING- LOW SULFUR CRUDE OIL
  GAS


CRUDE
OIL
DISTILLATION

RESID



GASOLINE (LOW OCTANE)

NAPHTHA
LIGHT GAS
HEAVY GAS
UE

OIL ^
OIL H

«-
e
L
E
N
D
1
N
G

8
L
E
N
D
1
N
G
REFINERY
GAS FUEL
(CONSUMED
 INTERNALLY)

PETCMEM FEED
B MILITARY
JET FUEL
                                                                                  KEROSENE,

                                                                                  DISTILLATE

                                                                                  'FUEL OIL ft

                                                                                  DIESEL FUEL
                YIELD. VOLUME
                PERCENT OF
               RAW MATERIALS



                   3.1





                  33.1
             25.5
                                                                                  RESIDUAL
J
                                                                                  REFINERY
                                                                                  LIQUID FUEL
                                                                                  (CONSUMED
                                                                                   INTERNALLY)
             37.2


              *
^''Included with gas  fuel

-------
                                          11-10


                                    Figure II-2
                            Schematic Diagram of a Hydroskimming  Refinery
  FLOW DIAGRAM  OF HYDROSKIMMING  REFINERY
      PROCESSING LOW SULFUR CRUDE OIL
 GAS
                                                    BUTANE (HIGH OCTANE)
           GASOLINE (LOW OCTANE)
                                       GASOLINE (HIGH OCTANE)
 RESIDUE
                                                                                     REFINERY GAS
                                                                                     FUEL (CONSUMED
                                                                                      INTERNALLY)
                                                                                     GASOLINES
YIELD. VOLUME
 PERCENT OF
RAW MATERIALS


   4.9


   2.2
   29.0
                                                                                     KEROSENE,
                                                                                     JET FUEL.
                                                                                     DISTILLATE
                                                                                     FUEL OIL 8
                                                                                     DIESEL FUEL
                                                                                     RESIDUAL
                                                                                     FUEL OIL

                                                                                     REFINERY LIQUID
                                                                                     FUEL(CONSUMED
                                                                                      INTERNALLY)
  25.5
  38.3

    *
* Included with  gas  fuel

-------
                                                    11-11
                                                 Figure II-3
                             Schematic Diagram of  a Fuels Refinery
              FLOW DIAGRAM OF FUELS REFINERY
              PROCESSING HIGH SULFUR CRUDE OIL
                                                               r
NATURAL CAS LIQUIDS
 HYDROGEN
 RECOVERY 8
MANUFACTURE
            HYDROGEN
                    GASOLINE (LOW OCTANE)
                                                 GASOLINE (HIGH OCTANE
                  ((OPTIONAL!
                  I
                  L_ — _> (ASPHALT)
  SULFUR
  FOR HYDROGEN
  TREATING
  REFINERY GAS
  FUEL (CONSUMED
   INTERNALLY)
•+LPG
                                                                                               GASOLINES
                                                                                               KEROSENE,
                                                                                               JET FUEL,
                                                                                               DISTILLATE
                                                                                               FUEL OIL 8
                                                                                               DIESEL FUEL
                            RESIDUAL
                            FUEL OIL


                            REFINERY
                            LIQUID FUEL
                            (CONSUMED
                             INTERNALLY)
                                         YIELD. VOLUME
                                          PERCENT OF
                                         RAW MATERIALS

                                             1.5*
11.0


 2.3
                                            53.8
                                            27.4
                                                                                                                8.1
         •'   Percent by weight
         "«  Included  with  pas  fuel

-------
                                                       11-12
                                                Figure  II-4
                                        Schematic  Diagram of a  High  Conversion Refinerv
         FLOW DIAGRAM OF HIGH CONVERSION REFINERY                        	     *
              PROCESSING  HJGH  SULFUR CRUDE OIL
NATURAL GAS LIQUIDS
                                                                            (HIGH OCTANE)
                    GASOLINE (LOW OCTANE)
                                                  GASOLINE (HIGH OCTANE)
              LIGHT GAS OIL
    DISTILLATION
              HEAVY GAS
                  OIL
                                                 GASOLINE (MEDIUM OCTANE)
    DISTILLATION
         VACUUM
         PITCH
                                          I  I GASOLINE (LOW OCTANE)
                                          !—(NAPHTHA
                                          GAS OIL
 SULFUR

 FOR HYDROGEN
 TREATING
 REFINERY
 GAS FUEL
 (CONSUMED
   INTERNALLY)
                                                                                                 GASOLINES
                                                                                                 KEROSENE.
                                                                                                 JET FUEL.
                                                                                                 DISTILLATE
                                                                                                 FUEL OIL 8
                                                                                                 DIESEL FUEL
REFINERY
LIQUID FUEL
(CONSUMED
  INTERNALLY)
                                                                                                 HIGH SULFUR
                                                                                                 COKE
YIELD. VOLUME
 PERCENT OF
RAW MATERIALS

     1.4*
                                                                                                                 13.0
                 77.5
                 10.8
                  4.6*
        *   Percent  by weight
        ** Included with gas  fuel

-------
                                11-13
TABLE II-l.  Functional Characterization of Refinery Processes
SEPARATION
Separation on the Basis of
 molecular Weight

Distillation (atmospheric and
 vacuum fractionation of crude
 oil, naphtha splitting,
 depropanizing, stabilization)

Absorption (recovery of olefins
 frcm catalytic cracked gas,
 recovery of propane from
 natural gas or hydrocracked
 gas)

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

Separation on the Basis of
 molecular Structure

Extraction (recovery of
 benzene, toluene, and zylenes
 from catalytic reformate,
 removal of aromatics from
 lubricating oil feedstock)

Crystallization (dewaxing of
 lubricating oils, recovery of
 paraxylene from mixed xylenes)
ALTERATION (CONVERSION)	

A. Conversion on the Basis of
   Molecular Weight

   Thermal cracking (visbreaking,
    coking)

   Catalytic cracking

   Hydrocracking

   Alkylation

   Polymerization
B. Conversion on the Basis of
    Molecular Structure

   Catalytic reforming (benzene,
    toluene, and xylene manufac-
    ture and octane improvement)
   Isomerization (normal butane to
    iso for alkylation, normal
    pentane and hexane to iso
    for octane improvement)
                  TREATMENT TO REMOVE IMPURITIES

                  Hydrogen treatment (hydrotreating)

                  Caustic treatment (Merox, Bender)

                  Clay treatment  (of lubricating oils)

                  Acid treatment

-------
                              11-14






of some important octane-improvement processes (particularly



reformers), the higher the severity, the higher the octane pro-




duced, but also the lower the "yield" (the amount of gasoline



material produced per unit of input).  In general, the more the



process unit is used, the higher its marginal cost of processing



the next increment of feed will be.



     In the model, all processes consist of a series of linear



relationships that describe the process output and operating cost,



given specified inputs and a set of operating conditions.  The



relationships are stored in the model in the form of a process



data table.  (Table II-2 shows an example of such a table.)  Each



column in this process table represents the processing of a



specific type of crude oil and each row represents a specific input



or output stream, fuel,  utility consumption, etc.  For example, the



model specifies that as  one barrel of Saudi light crude is pro-



cessed, a mix of 15 intermediate streams is created.  The operation



consumes fuel, power, steam, and capacity, and incurs variable



operating costs of 9.3 cents per barrel.



     For producing high-octane components for low-lead or unleaded



gasoline, four types of  "downstream" units (i.e., units that pro-



cess outputs from the crude distillation process) are particularly



important:  fluid catalytic cracking (FCC) units, isomerization



units, alkylation units, and reformers.   Each of these units




enhances the octane of a different stream of intermediate products.



     Finally, after all  processing is complete, the refinery ends



up with numerous process output streams that are blended together



to produce final, salable refined products.  This activity is

-------
                             11-15
TABLE II-2.
Sample Process Data Table  from  Refinery Model:
Yields and Operating Cost  Coefficients for Crude
Distillation Unit



Saudi Arab light
Saudi Arab heavy
Mexican maya
Capacity factor

Still gas
Propane
Isobutane
Normal butane
Lt. St. run (C5-175) low oct.
Lt. st. run (C5-175) int. oct.
Lt. naph. (175-250) parf.
Lt. naph. (175-250) intm.
Naph. (250-325) parf.
Naph. (250-325) intm.
Hvy naph. (325-375) parf.
Hvy naph. (325-375) intm.
Kero. (375-500) jet fuel quality
Kero. (375-500) other
Dist. (500-620) hi sulfur
Hy gas oil (800-BTMS) (2.0% S)
Asph. , very hi sul.(4.3% S)
Crude Oil
Saudi Arab
Light Heavy
-1
-1

1.0 1.0
Yields (Fraction
0.001 0.001
0.003 0.003
0.002 0.002
0.013 0.015
0.051
0.040 0.035
0.070 0.060

0.050 0.044
0.020 0.011
0.020 0.020
0.020 0.014
0.115 0.090
0.015 0.005
0.130 0.090
0.180 0.180
0.143 0.300
Type
Mexican
Maya


-1
1.0
of Intake)
0.001
0.003
0.002
0.009


0.025
0.025
0.010
0.050
0.005
0.030
0.070
0.040
0.100
0.105
0.350
Operating Cost Coefficients

Fuel, fuel oil equivalent
Power, KWH
Steam, LB
Other var. op. cost, $
Capital charge
(Per Barrel of
-0.021 -0.022
-0.6 -0.6
-60.7 -63.4
-0.092 -0.093
varies
Throughput)
-0.020
-0.6
-57.9
-0.092

      i I1^ i i v^ ^ *•* w * » **  •**• "•^•~"~ »  '
      fuel oil,  power,  steam, etc.

-------
                              11-16






represented in the model by product blending units.  The blending



units contain quality data for all refinery streams and quality



specifications for final products.  The components are then com-



bined by the model such that the qualities of the blended mixes



meet the minimum requirements of product specifications.



     The refinery model can be operated in either of two modes —



minimum cost or maximum profit.   It can constrain product quanti-



ties and compute a minimum cost  solution.  (This is useful for



analyzing the country as a whole or large refining regions in



which aggregate demands can be forecast.)  Alternatively, the



simulation can vary product quantities to maximize profits at



preselected prices.



     The principal reason to use computer models to simulate petro-



leum operations is to measure differences between alternative



scenarios, and thereby estimate  the changes in petroleum activities



when some conditions change.



    It is important to note that the procedure used does not simply



find the optimal way to make up  the difference between the current



lead level and the lower level.   Rather, subject to constraints



discussed later, we first run the model at the current lead



standard and find the costs of meeting that standard; then we



repeat the process, finding the  cost of making gasoline at the



0.10 gplg standard.  The difference between the cost of producing



gasoline under the current standard and the new standard is the



cost estimate of the rule change.




     This approach minimizes the risk of error in the cost estimate,



because if the model mistakenly  assumes some flexibility that the

-------
                               11-17






industry does not have,  it would  reduce  the estimated cost of



meeting both the current  and  the  new  standard.  The difference



between the costs is only affected if  the overoptimistic assump-



tion generates greater savings  in one  case than in the other case,



and even then the size of the  error is only the difference in



overoptimization, not the full  impact  of the overoptimistic assump-



tion.  EPA believes that  this  approach to using the model adds



considerable confidence  to the  results.



     The refinery linear  programming model assumes (when uncon-



strained) that the least  expensive refinery equipment is used to



meet both the current and the  proposed gasoline lead standard.  In



a competitive market, this is  a reasonable assumption, because if



one refiner's marginal costs  are  lower than competitors', that



refiner would be expected to  increase market share.  As its market



share  increased, its utilization  of refinery equipment would



increase, particularly of octane-making equipment, and therefore



the refiner's marginal cost of  octane  would increase to the level



of its competitors.  A new equilibrium would be reached with some



market share having shifted,  but  not  necessarily a large share.



The fact that a linear program  gives  the same results as a competi-



tive industry had long been known and  used in economics (see, for



example, Baumol, Economic Theories and Operations Analysis,




Prentice Hall, 1961).



     The gasoline marketing system, which already moves millions



of barrels of gasoline per day  in pipelines around the country,



helps assure that the more efficient producers can get their



product to market.  In addition,  the  industry has a long tradition

-------
                              11-18






of trading components and products, which also results in a more



efficient allocation of feedstock-to-equipment.  A competitive



market and an increasing marginal cost of octane for all refiners



means that the model simulates the industry reasonably accurately.



The verification tests for 1982 showed the model accurately predic-




ting refinery products and prices.






II.C.  Base-Case Assumptions and Cost Estimates



     Refineries are complex operations with multiple inputs and



outputs.  The costs of producing any product depend in part on



the quantities of other products produced and their prices.  For



example, the cost of producing gasoline depends partially on how



much residual oil is sold and its selling price.  The more residual



oil is sold and the higher its price, the lower the cost of



producing gasoline.  As a result of such joint costs, cost esti-



mates for a change in any one product must consider the full



range of refinery operations.



     To estimate the costs of reducing lead in each year, we first



ran the DOE model in its cost-minimization mode with the constraint



that lead use not exceed 1.10 gplg (the current limit), and calcu-



lated the costs of the resulting solution.  We then reran the



model at lower lead limits (varying from 0.50 gplg to zero,



depending on the year) and recalculated the costs; our cost esti-



mates are the differences between the two runs.  At both 1.10



gplg and the lower levels, the model had to meet specified product



demands.




     In using the refining model, it is necessary to supply it



with the values of many parameters.  Section II.C.I describes the

-------
                              11-19


key input assumptions used in making our base-case estimates.

Section II.C.2 presents the resulting year-by-year cost estimates

based on those inputs.  Section II.D describes the results of

multiple sensitivity analyses conducted to test the robustness

of our conclusions.


II.C.I.  Base-Case Parameter Values

     The key assumptions input to the model can be divided into

two general categories:  gasoline demand and quality constraints,

and refinery operations.  Each category is discussed below.


II.C.I.a.  Gasoline Demand

     Demand for gasoline in 1983 was about 6.6 million barrels

per day, or a little over 100 billion gallons per year.  To project

gasoline demand in future years, we examined a 1983 study by Data

Resources Incorporated  (DRI) and projections from DOE's Energy

Information Administration (1983), and adjusted them upwards to

reflect smaller projected gains in fuel economy.*  We project that

gasoline demand will fall to about 6.5 million barrels per day in

1988, because of expected improvements in the average fuel effi-

ciency of vehicles in use, but then will level off in 1990 as

increases in vehicle miles travelled balance improvements in

miles per gallon.  Since we made our original projections, the

DOE projections have been revised upward to about our level

(U.S. DOE, 1984 Annual  Report to Congress).
* Compliance with the rule is harder and more costly with
  higher gasoline demand.

-------
                              11-20






     We projected the split in demand between leaded and unleaded




gasoline in two ways.  First,  we fit linear and logistic regres-



sions to the monthly leaded-unleaded split documented in the



Monthly Energy Review (from the Department of Energy), using time



as the explanatory variable.  We also regressed the thirteen-month



moving average, to remove seasonal and random variation.  These



all suggested an unleaded share of 67.5 percent (plus or minus 0.6



percent at a 95 percent confidence level) of the gasoline market



in 1988.  Our vehicle fleet model (described in Appendix B), using



historic scrappage rates from DRI, predicted essentially the same



unleaded share.



     We also had to estimate what portion of leaded demand is due



to "misfueling", as we expect the rule to have a major impact on



the misuse of leaded gasoline in vehicles equipped with catalysts.



Deliberate misfueling (i.e., the use of leaded gasoline in vehicles



designed for unleaded) appears to occur because leaded gasoline is



both cheaper and higher in octane than unleaded gasoline.  Several



EPA and private studies have showed that widespread misfueling has



slowed the decline in lead emissions significantly, and challenged



the assumption that leaded gasoline would soon be eliminated



because of lack of demand.  According to a 1983 survey by EPA



(U.S. EPA, 1984e), the current misfueling rate of light-duty



vehicles designed to use unleaded gasoline is about 15.5 percent.




(This figure was obtained after making certain adjustments; see



Chapter VI for more details.)   The 1982 survey found a 13.5 percent



misfueling rate (U.S. EPA, 1983a).  A panel study (Energy and



Environmental Analysis, 1984)  in which consumers kept diaries of

-------
                              11-21
their gasoline purchases indicates that raisfueling is even more



widespread, though many people misfuel only occasionally.



     We used the 1983 survey data to estimate misfueling rates by



age of vehicle (the older the vehicle, the higher the misfueling



rate) and by whether or not the vehicle was in an area with an



Inspection and Maintenance (I/M) program  (vehicles in I/M areas



misfuel less).  We also adjusted for the  fact that older vehicles



travel fewer miles per year, on average.  These adjusted rates were



then applied to our fleet model (described in Appendix B) to esti-



mate the demand for leaded gasoline by misfuelers.  Our estimates



of misfueling did not influence our estimate of total demand for



leaded gasoline (which was based on a statistical analysis of



historical trends, as discussed earlier), only the split in



leaded demand between misfuelers and legitimate users.



     The model requires further disaggregation of unleaded demand



into the demand for premium and regular.  In our base case we



assumed that the demand for premium unleaded would comprise 25



percent of the total demand for unleaded  gasoline.  This is up




slightly from the current level of 24 percent.



     Table II-3 presents our year-by-year estimates of gasoline



demand, both with and without misfueling.  Note that we project




that total demand for leaded gasoline will decrease, but misfuel-



ers1 demand (in the absence of further regulation) will increase




in both absolute and proportional terms.




     Refining costs depend on the octane  level of gasoline, as well



as the split between leaded and unleaded.  For each of the three

-------
                              11-22






grades of gasoline (leaded, unleaded regular, and unleaded pre-



mium) , we assumed that octane levels would remain at their current



averages, which are slightly higher than 87 octane for unleaded



regular, slightly higher than 91 octane for unleaded premium,



and a little above 89 octane for leaded gasoline.  These octane



ratings reflect the average of two measures, "motor octane" and



"research octane." We constrained the model to meet or exceed



all three octane measures in current gasoline.



     For 1986 and subsequent years, we assumed that the regulations



would reduce misfueling.  Tn these runs we assumed that half of



the misfuelers do so strictly for price, but that the others



also want the higher octane provided by leaded regular.  That



is, we assumed that of the potential misfuelers deterred by the



price change resulting from the rule, 50 percent would switch



to regular unleaded and the other 50 percent would require unleaded



gasoline averaging 89 octane.  (The latter could be either an



intermediate grade of unleaded, or a mixture of regular and



premium unleaded.)  Thus, we assumed that deterred misfuelers



would demand 88 octane unleaded, on average.  Because 0.10 grams



of lead only adds about one octane number to a gallon of gasoline,



89 octane leaded gasoline must start with an unleaded gasoline



with a "clear-pool" octane rating rating slightly above 88.



Therefore, our cost estimates at 0.10 gplg are not very sensitive



to assumptions about the amount of misfueling eliminated;



indeed, the costs rise slightly with the misfueling rate.



      We also had to specify the split between domestic and im-



ported gasoline supply.  Over 4 percent of gasoline is now imported

-------
                              11-23
TABLE II-3.  Year-by-Year Estimates of Gasoline Demand
Rule
With Misfuelina
Leaded
Unleaded
Without Misfuelinq
Leaded
Unleaded
1985
40
60
32
68
.2
.3
.1
.5
1986
37.5
62.8
28.8
71.5
1987
34
65
25
74
.9
.1
.6
.4
1988
32
67
22
77
.4
.3
.4
.2
1989
29.8
69.5
19.2
80.1
1990
27.6
71.4
16.4
82.6
1991
25.2
73.8
14.9
84.1
1992
24.3
74.7
13.4
85.6
Total               100.6  100.3  100.0  99.6  99.3  99.0  99.0  99.0

-------
                              11-24


and over 60 percent of imports are unleaded, but our base case

constrained the model to meet all demand for gasoline using

domestic refineries.  (For modeling purposes, "domestic refineries"

include those in Puerto Rico and the Virgin Islands.)

     We also constrained the model to meet other important charac-

teristics of current gasoline, including its Reid Vapor Pressure

(RVP).  Because certain high-octane blending components (such as

butane) and additives (such as alcohols) have high RVPs, this

constraint raised the estimated cost of meeting the rulo.

      In addition to gasoline demand, we also had to specify

demands for other products.  These were based on a study done

for DOE's Energy Information Administration (Decision Analysis

Corporation, 1983).  In most cases, we constrained the model

to produce these quantities domestically.  In a few cases, where

there are wellestablished and large international markets (e.g.,

for residual oil), we allowed the model to slightly increase
                                                  \
or decrease imports at fixed prices if that was the lowest-cost

solution.


Il.C.l.b.  Assumptions About Refinery Operations

     The DOE model requires that many individual parameters be

specified for refinery operations.  Here we discuss only the most

important of those assumptions.

     We based our estimates of available capacity on published

tabulations of existing equipment and announced projects that will

be completed in the relevant years.  Table II-4 presents the

estimates of available "stream-day" capacity used in the model.

For such capacity, we did not include a capital charge for its use

-------
                              11-25






in meeting tightened lead rules because the cost of that equipment




is already "sunk"; it does not vary with the standard.  In our



estimates for 1985, 1986, and 1987, we constrained the model to



use this existing capacity because it takes two to three years for



refiners to plan, obtain permits for, and complete major construc-



tion projects.  (There are "debottlenecking" and upgrading actions



that refiners can take to improve their ability to make octane in



as little as six months, other steps that might take a year, etc.



We have ignored such changes and, therefore, somewhat underesti-



mated industry capacity.)  For the later years, however, we



allowed the model to determine whether construction of additional




capital equipment was economical.  In those later years, the



model predicted that costs would be slightly lower if new equip-



ment (primariy isomerization units) were added.  This new equip-



ment, added in response to the rule, was included in the cost



estimates using a real capital charge rate of 15 percent (which



corresponds to a discount rate of 10 percent and an equipment



lifetime of just under 10 years).



     Because of maintenance and breakdowns, equipment cannot be



used to its full daily capacity, called "stream-day" capacity in



the refining industry.  To account for this fact, we limited capac-



ity utilization to 90 percent.  This figure reflects tight, but



feasible, operating conditions, based on industry practice and the



judgment of EPA's expert consultants on the refining industry



(Sobotka and Company).  This judgment was also reinforced in a



letter to the National Petroleum Refiners Association from one of




its consultants, who suggested that reformers  (a key piece of

-------
                            11-26


TABLE II-4.  Estimated U.S. Refinery Processing Unit Capacities
	for 1988 (thousands of barrels per day)	

Processing Unit	Capacity	

Crude distillation                              15,900

Vacuum distillation                              6,880

Coking                                           1,340

Visbreaking                                        170

Low pressure reforming (150 psig)                  940

Medium pressure reforming (250 psig)             2,280

High pressure reforming (450 psig)                 750

Catalytic cracking                               5,340

Hydrocracking                                      980

Alkylation                                         960

Polymerization                                      78

Butane isomerization                                56

Pen./Hex. Isomerization                            206

-------
                              11-27   ^






downstream equipment in meeting low-lead requirements) could be



operated at 92 percent of capacity (Soloman and Associates, 1984,




in Docket EN-84-05).  Although no data are publicly available on



the utilization of  individual pieces of equipment, it is important



to note that during the last four years of the 1960s, the refining




industry used more  than 90 percent of its crude capacity; in 1966,



the utilization rate was 91.8 percent.



     In addition to specifying maximum utilization rates for all



equipment, we also  specified minimum utilization rates for certain



relatively inefficient reformers.  Earlier versions of the DOE



model included two  types of reformers.  In response to criticisms



of that assumption, we modified the model to distinguish among



three types of reformers:  low-pressure (the most modern and



efficient), medium-pressure, and high-pressure (the oldest and



least efficient).  When unconstrained, the model uses the available



capacity of the low-pressure reformers first, then the medium-



pressure reformers, and then, only if more reforming is needed,




does it use any high-pressure reformers.



     Some refineries, however, do not have any medium- or low-



pressure reformers.  To account for this fact, we determined how



many refineries have only high-pressure reformers and required



that the model use 75 percent of their capacity.  This is a very



strong assumption, as it is tantamount to saying that those




refiners neither lose market share nor purchase such components



from other refiners.  Together with that requirement, we also



assumed that refineries with both high-pressure and better-quality




reformers would use the less-efficient reformers to some limited

-------
                              11-28






extent, and constrained the model to direct at least 270,000



barrels per day to high-pressure reformers; that works out to a



minimum capacity utilization rate for such reformers of 36 percent.



Because the model is more tightly constrained by the availability



of reformer feed than by reformer capacity, this minimum utiliza-



tion constraint increases the costs and makes the low lead levels



more difficult to achieve.  In most runs, for both 1.10 gplg and



lower levels, the model used the maximum amount of low-pressure



capacity, the minimum amount of high-pressure capacity, and all



of the swing took place in the utilization of medium-pressure



reformers.



     Several more detailed input assumptions also are worthy of



note.  In the Preliminary RIA (U.S. EPA, 1984f) issued at proposal,



we assumed that refiners would continue to use the current mix of



catalyst grades in their FCC units.  During the comment period, EPA



received information from commenters about some new, more efficient



catalysts now in use at some refineries.  After checking with some



catalyst manufacturers (W.R. Grace, 1984, and Union Carbide,



personal communication) to assure that capacity is available to



produce sufficient quantities of these new catalysts, we assumed



that 25 percent of FCC capacity could use the new catalysts by



the second half of 1985, and that 50 percent could use them by



1986; to be conservative, we doubled the selling price to account



for the cost of replacing existing catalysts early.




     The DOE model distinguishes among three types of naphtha



produced by the crude distillation process; different types of



crude produce different proportions of these naphthas.  In the

-------
                              11-29






Preliminary RIA, we allowed the model to allocate those naphthas



optimally for further processing  in isomerization units or for




direct blending  into gasoline.  In practice, such optimal alloca-



tion  is likely to be difficult, although it can be approached by



segregating the  storage of different naphthas, and through careful



purchases of different types of crude by refineries with different



isomerization units.  To limit this optimization, our base case



now forces the model to process a preselected mix of these naphthas



that  achieves only 20 percent of  the gain possible with optimal



segregation.  In response to several comments, we also reduced the



reformer yields  with parafinic naphthas at higher severity levels.



      In some runs, the model found that the most economical method



of meeting demand with less lead  included altering the mix of



crude inputs, switching from heavy crudes (such as those from



Saudi Arabia) to lighter, high-quality crudes (such as those from



Nigeria), despite the fact that we include a price differential



between such crudes.  To account  for possible rigidities in long-




term contracts and the like, we constrained the model not to



reduce its use of heavy crude in going from 1.10 gplg to a lower




level.



     The model also found it more economical to increase the use



of alcohols in gasoline in some cases, despite the fact that we




forced the model to increase gasoline production when alcohol was



used  (alcohol has a lower BTU content than other gasoline compo-



nents and, therefore, delivers fewer miles per gallon).  Gasoline



containing alcohol, however, faces uncertain consumer acceptance,




and many pipelines will not accept it because of possible moisture

-------
                              11-30






problems.  Thus, we constrained the model to use no more alcohol



at low lead levels than it used at 1.10 gplg.  Because of scheduled




expansion in ethanol capacity we allowed the model to use somewhat



more ethanol at either level than is currently used, but oxinol



levels were kept at current usage, and no MTBE use was allowed,



despite current usage of about 13,000 barrels per day for octane



enhancement of unleaded gasoline.



     In addition to limiting the model's use of alcohol to boost



octane at low lead levels, we did not allow it to use any MMT,



another octane enhancer.  During the mid-1970s, refiners used MMT



to increase octane in unleaded gasoline.  After tests showed that



MMT harmed catalysts, such use was banned.  MMT is still legal in



leaded gasoline, however,  and some unknown amount currently is



used.  Nonetheless, to be  conservative, our base case did not



include any possible use of MMT to increase the octane of low-lead



gasoline.



     One additional change has been made in our cost estimates.



When EPA began analyzing potential changes in its lead regulation



in late 1983, it assumed that oil prices would remain constant in



1983 dollars.  That is, they were assumed to increase with infla-



tion.  These price assumptions were kept in the March 1984 Cost-



Benefit Study, Prelimary Regulatory Impact Analysis, and subse-



quent docket submissions.   In fact, however, oil prices, far from



increasing by a few dollars per barrel since 1983, have fallen by



several dollars per barrel, even in nominal terms.  Table II-5



shows the original price estimates for Nigerian light and Saudi



heavy crude used in our model runs, the prices at the end of

-------
                              11-31
TABLE II-5.  Prices of Crude Oil and Petroleum Products in 1983
             and 1985 (dollars per barrel)
Product
Crude Oil
Nigerian Crude
Saudi Heavy
Residuel Oil
0.3% Sulfur
0.5% Sulfur
1% Sulfur
2% Suffur
Other Products
Isobutane
Normal Butane
LPG
1983
Price
31.50
28.00
28.00
27.00
26.00
24.75
26.50
25.70
18.20
1984
Current
Dollars
29.18
28.60
28.00
27.43
26.76
26.18
24.50
23.75
17.78
Price
1983
Dollars
27.40
26.88
26.29
25.76
25.13
24.58
23.00
22.30
16.70

-------
                              11-32






1984, and what those prices are equivalent to in constant 1983



dollars.  It also shows those same prices for residual oil,




butanes, and propane.



     Product prices are based on about a six-month period in



1984 to avoid seasonal distortions.  The differences between the



first and last columns show that this assumption has seriously



distorted our cost estimates.



     Rather than rerun all of the refinery analyses to find the



costs under these lower prices, we have adopted the following



approach.  We have assumed that oil prices stopped falling in



December 1984 and will henceforth increase with inflation.  We



have repriced the cost estimates of our model runs using the new



prices, assuming (to be conservative) that the costs of crude



fell by the same percentage drop as the heavier crudes, not the



lighter crudes.  To test the reliability of this approach, we



reran the model for 1986 and 1988 with the new prices in, and



compared it to the results of our repricing.   These costs should



be somewhat lower, because, in response to the change in the



relative prices of crudes and products, refiners may find it less



expensive to change some of their operating procedures.  As Table



II-6 shows, the costs using our procedure are quite close, but



slightly higher than the costs given by completely rerunning the



model.  We also checked the marginal cost estimates for some of



the more highly constrained sensitivity analyses, and achieved



close agreement as well.




     Because the British coal strike may have been keeping up the




price of high sulfur residual oil on the world market, we also

-------
                              11-33
TABLE II-6.  Cost of 0.10 gplg Standard with New Oil Prices
             New Model Run versus Repricing, Assuming No
             Misfueling (millions of dollars)
Year
1986
1988
New Model Run
573
502
Repricing
608
531

-------
                              11-34






recalculated the costs assuming that high sulfur residual oil was



selling for $1 per barrel less than current prices.  This only



increased the cost of our 1986 case by $14 million, so these esti-



mates are not very sensitive to that potential factor.






II.C.2.  Base-Case Results



     Table II-7 presents information on the model's estimates at



1.10 gplg and at lower lead levels in 1985, 1986, 1987, and 1988.



(The figures for 1985 apply only to the second half of the year.)




These estimates assume that 0.50 gplg in 1985 does not eliminate



any misfueling and that 0.10 gplg in the later years reduces mis-



fueling by 80 percent.  (See Chapter VITI for a discussion of



alternative assumptions about misfueling and the costs associated



with those assumptions.)  In addition to the cost estimates, the



table reports several other important pieces of information:  the



capacity utilization rates for various types of reformers (low-,



medium-, and high-pressure, plus the aggregate figure) and the



marginal cost of producing an octane-barrel of gasoline.  (An



octane-barrel is defined as raising the octane of a barrel of



gasoline by 1 point.)  Note that the 0.50 gplg standard for the



second half of 1985 has relatively little impact.  The cost is



only $96 million; the overall utilization of reformers rises from



50 percent at 1.10 gplg to 59 percent at 0.50 gplg.  The marginal




cost of producing an octane-barrel rises from 15.8 cents at 1.10



gplg to 20.4 cents at 0.50 gplg.  Thus, refiners should be able  to



meet the 0.50 gplg standard with relative ease and at a relatively



moderate marginal cost by using excess capacity.

-------
                               11-35





      With regard to these estimates  of  marginal  costs,  it  is



 interesting to note the comment  of one  refiner  (Amoco submission



 to Public Docket EN-84-05)  critical  of  EPA's  cost  estimates:



 "Current wholesale spot market price differentials between  leaded



 and unleaded imply a marginal cost of about  20  cents/BON  [barrel



 of octane number].  The trading  value of  lead rights reflects a



 similar cost of octane..."   As shown in Table II-7, the model



 with EPA's base-case assumptions projects a marginal cost of 15.8



 cents per octane-barrel at  1.10  gplg in 1983  dollars, or 16.7



 cents in 1984 dollars,  very similar  to  the Amoco estimate for



 current (late 1984) conditions.



      The base-case results  for 0.10  gplg  in 1986 and 1987 show



 somewhat tighter operating  conditions,  but no feasibility problems.



.In both years, overall  reformer  utilization rises  to 66 or  67



 percent, and the marginal cost of an octane-barrel rises to a



 little more than 29 cents.



      The results for 1988 are of interest because  that  is the



 first year in which we  assumed that  additional  capacity could be



 added.  The model  found it  most  economical to add  about 98,000



 barrels per stream-day  of isomerization capacity.  That allows



 slightly lower use of reforming  capacity  than in 1986 or 1987,



 and reduces the marginal cost of octane to 28.4  cents per barrel



 at 0.10 gplg.   The cost of  that  extra isomerization capacity



 ($322 million), annualized  at 15 percent, is  included in the



 $535 million cost  estimate  for 1988,  and  also in the cost estimates



 for subsequent years.

-------
                            11-36
TABLE II-7. Base-Case Results for 1985-1988, with Partial Misfueling
Year
Reformer utilization
(%) by reformer type
Marginal
Cost of
 Octane
Total
 Cost
Lead
1985
1.10
0.50
1986
1.10
0.10
1987
1.10
0.10
1987
1.10
0.10
limit
gplg
gpig
gplg
gplg
gplg
gplg
gplg
gplg
Low
90
90
90
90
90
90
90
90
Med.
40
53
40
67
41
66
42
62
High
36
36
36
36
36
36
36
36
Agg.
50
59
50
67
52
66
52
64
(^/barrel)
15.
20.
16.
29.
16.
30.
17.
28.
8
4
3
2
4
0
2
4
(million $)
N.A.
96*
N.A.
608
N.A.
558
N.A.
532
 rCost for quarters III and IV.

-------
                              11-37






     The 1988 results also are interesting because they reflect



the long-run marginal cost of producing octane (including capital



charges).  The reason that the marginal costs for 1.10 gplg are



lower is that the refining industry has excess octane-producing



equipment, and we have not included the costs of that sunk capital



in estimating either total or marginal costs.  Note also that



the marginal cost of 0.50 gplg in 1985 (20.4 cents per octane-



barrel)  is far below the long-run marginal cost, which suggests



that refiners would not find it cost-effective to build new capa-



city to meet that standard even if it were possible; the minimum-



cost solution for 1985 would not include any new equipment.  In



1986 and 1987, however, at 0.10 gplg, the marginal costs slightly



exceed the long-run figure, indicating that it would be cheaper



to build new capacity, if it were possible (although it is still



possible to meet the 0.10 gplg standard without new capacity).



     Table II-8 presents the year-by-year cost estimates under



the base-case assumptions.  Estimates are shown for three differ-



ent rules:  the original proposal (0.10 starting 1/1/86), the



illustrative alternative discussed in the Notice of Proposed




Rulemaking (0.50 gplg on 7/1/85, 0.30 on 1/1/86, 0.20 on 1/1/87,



and 0.10 on 1/1/88), and the Final Rule (0.50 gplg on 7/1/85 and



0.10 on 1/1/86).  Again, these estimates assume full misfueling



in 1985 and 80 percent reductions in misfueling in subsequent



years, when the 0.10 gplg standard applies.  The costs fall over



time because of projected declines in the demand for leaded




gasoline.

-------
                             11-38
TABLE II-8.  Year-by-Year Estimates of Costs of Meeting
             Alternative Rules, with Partial Misfueling
             (millions of 1983 dollars)
Rule
Proposed
Alternative
Final
1985
0
96
96
1986
608
364
608
1987
558
448
558
1988
532
532
532
1989
504
504
504
1990
471
471
471
1991
444
444
444
1992
441
441
441

-------
                               11-39





JI.D.  Sensitivity Analyses




     The base-case results shown  in Table  II-8 represent EPA's



"best estimates"  of  the  cost  of complying  with the phasedown



rule; i.e., they  are based on  what the Agency believes are the



most realistic  assumptions about  refinery  capabilities and gaso-



line demands.   The Agency also conducted numerous sensitivity



analyses to test  the robustness of the results.  Our efforts



focused on 1986,  because that  is  the  first year in which the 0.10



gplg standard will apply.  As  shown in Table II-7, the 0.50 gplg



standard in 1985  is much easier to meet, as measured either by



reformer utilization or  by the marginal cost of an octane-barrel.



The 1987 results  are similar  to those for  1986, but in practice



the 0.10 gplg standard should  be  easier to achieve in that year



because refiners  will have two years  to adjust their operations.



     Most of the  sensitivity  runs focused  on changes unfavorable



to the rule; i.e., on alternative parameter values that would



increase the estimated costs.  We did so not because we believe



that such values  are more likely  to occur  than those that would



reduce the cost of the rule,  but  rather to probe the hypothetical



conditions under  which the rule would become extremely costly or,



possibly, infeasible with existing refinery equipment.  Many of



the sensitivity analyses, particularly those varying several




parameters simultaneously, responded  to comments received on the




August 1984 proposal.



     We performed two types of sensitivity analyses.  The first




dealt with the  issue of  aggregation:  to what extent does a national




model underestimate costs by  failing  to account for regional

-------
                              11-40






differences? We concluded that the base-case cost estimates are




not artificially low because we used a national model.  The



disaggregated runs also indicated, however, that the rule would



cause tighter operating conditions in Petroleum Allocation for



Defense Districts (PADDs) I-IV and VI than in PADD V (the West



Coast, Alaska, and Hawaii), so our second set of sensitivity



analyses, designed to probe the limits of feasibility, focused



on PADDs I-IV/VI (the rest of the country).






II.D.I.  Level of Aggregation



     Several comments received before and after the August 1984



proposal criticized EPA's cost analysis for relying on a national



model.  To examine that issue, we conducted two sets of sensitivity



analyses.



     The first compared EPA's estimates to those based on an



analysis performed by Turner-Mason Associates (TMA) (1984) for



the Lead Industries Association.  TMA disaggregated the refining



industry into six groups, based on geographic location, size and



type of refinery, and other factors.  It estimated the cost of



meeting the 0.10 gplg in 1988 to be $995 million, compared to



EPA's estimate at the time of proposal of $503 million.  In



addition to disaggregating, however, TMA changed several other



assumptions.  To see to what extent those changes, rather than



the level of aggregation, explained the difference in results,



we incrementally made each of those changes in the inputs to the



national DOE model.  The end result was that our national model



with all of the TMA assumptions predicted slightly higher costs

-------
                              11-41






than the TMA analysis, $1,016 million vs. $995 million (September




19 r "Supplemental Analysis of Refining Costs," submission to Public



Docket EN-84-05).  This result indicated that different levels of



aggregation were not the source of discrepancy between EPA's and



TMA's estimates, and that EPA's model accurately reproduced the



results obtained with TMA's substantially finer level of disaggre-



gation.




     Note that the base-case estimates in this Final RIA incor-



porate several of the more pessimistic assumptions used by TMA,



in particular a reduction in reformer yields at high severities.



Other assumptions made in the TMA analysis, however, do not



appear reasonable, in particular the capital charge rate (27



percent real) and a reduction in butane use of 50 percent from



its current level.



     The second approach to the aggregation issue divided the



refining industry into two units, PADD V (the West Coast) and



PADDs I-IV (the rest of the continental U.S.) plus PADD VI (the



Virgin Islands and Puerto Rico).  The rationale for this analysis



was that, while PADDs I-IV/VI are tightly interconnected by pipe-



lines and water shipping, PADD V is relatively isolated; only a




few small pipelines connect PADD V to the rest of the country,



and transport by sea between the Gulf and the West Coast is




relatively expensive, though by no means impossible.



     Table II-9 compares the national base-case estimates with



the separate estimates for the two parts of the country for 0.10



gplg in 1986.  Note that the sum of the two regional estimates




is insignificantly different than the national estimate, $611

-------
                              11-42






million vs. $608 million, which indicates that using the national



model did not bias downward our base-case estimates.  Table II-9



also reveals, however, that the rule generates somewhat higher



operating rates and marginal costs for producing octane in



refineries in PADDs I-IV/VI, which suggests that any problems



from tighter operating constraints would show up in that part of



the country before they occurred in PADD V.



     It is also important to note that very fine levels of dis-



aggregation can lend a strong upward bias to the cost estimates



by failing to account for the flexibility in the refining industry



to trade products and shift production in the face of changes in



market conditions.  Suppose, for example, that each refinery were



modeled separately, with the constraint that it produce exactly



the same slate of products that it produces now.  That approach



would overestimate total costs, because it would not allow for



adjustments in product mixes across refineries.  Such adjustments



may be quite substantial, as refineries with better octane-making



equipment would increase their total production of gasoline or



increase the proportion of production that is premium gasoline,



while other refineries would reduce their total gasoline output



(or premium gasoline share), possibly shifting to the production



of other petroleum products that place less burden on octane



enhancement.






II.D.2.  Other Parameters




     We conducted extensive sensitivity analyses of other key



parameter values  in PADDs I-IV/VI.  Table 11-10 summarizes the

-------
                             11-43
TABLE II-9.  Costs of Meeting the 0.10 gplg Standard:
	Comparison of National and Regional Results for 1986
Model
National Model

  Base case
Reformer Utilization
(%) by Reformer Type
Low  Med. High  Agg.
 Marginal
 Cost  of
  Octane
(j/barrel)
   Total
    Cost
(million $)
 90   67   36   67
   29.2
    608
Regional Models

  PADDs I-IV/VI

  PADD V

    Total
 90   71   36   69

 90   60   36   60
   31.6

   21.0
    531

    _8_0

    611

-------
                              11-44






base-case and alternative assumptions explored.  As noted earlier,



we focused on adverse changes; i.e., those that would make it more



difficult to meet the standard.  We also ran some sensitivity



analyses of changes that would make it easier to meet the rule



than our base case predicts.






II.D.2.a.  Assumptions Varied



     The first item listed in Table 11-10 is the demand for un-



leaded premium gasoline.  In our base case, we assumed that the



demand for unleaded regular would be 25 percent of the total demand



for unleaded gasoline, up slightly from its current level of 24



percent.  As shown in Table 11-10, we also ran sensitivity analyses



that assumed that the fraction of premium would grow.  Case Al



assumes that the demand for premium would grow by about 1.5 per-



centage points per year, reaching 27 percent of total unleaded



demand in 1986 and 30 percent of the total in 1988.  Case A2



assumes that premium demand grows at an even higher rate, about



5 percentage points a year, reaching 34 percent of total unleaded



demand in 1986.  This high rate of growth is out of line with



historical trends and appears implausible, particularly if the



marginal cost of producing octane rises, as it will with the



phasedown.




     Case B tightens the constraint on capacity utilization of



all downstream processing units (including, most importantly, FCC



and reforming units) from 90 percent of stream-day capacity to 85



percent.  Several commenters suggested that EPA's assumption of



90 percent was too optimistic.  Although the Agency continues to

-------
                           11-45
Case
Al
A2
B
C
D
El
E2
E3
Parameter/Assumption
Premium share of unleaded demand
Premium share of unleaded demand
Maximum capacity utilization of
downstream processing units
Share of FCC units using new
catalyst
Maximum ethanol use, '000 barrels/day
Summer RVP limit
Summer RVP limit and 5% higher demand
Summer RVP, summer gasoline demand, and
Base-Case
Value
25%
25%
90%
50%
60
No
No

Alternative
Value
27%
34%
85%
0%
30
Yes
Yes

     summer distillate demand                   No

F    Substitution of light for heavy crudes
     permitted                                  No

G    MMT use allowed in leaded gasoline         No

H    Premium share of unleaded demand          25%

I    Percent of naphtha segregated
     for gasoline processing                   20%
 Yes


 Yes

 Yes

 23%


100%

-------
                              11-46






believe that 90 percent utilization can be achieved, this sensi-




tivity analysis examined the impact of a lower utilization rate.



     Case C assumes that all FCC units continue to use the same



catalysts that they do now; the base case, as discussed earlier,



assumes that by 1986 half of the FCC capacity would have switched




to one of the newer, more efficient catalysts now available.



     Case D examines the impact of reduced alcohol use.  In the



base case, we constrained the model to use no more alcohol



(ethanol) at 0.10 gplg than it used at 1.10 gplg, which was



60,000 barrels per day-  For this sensitivity analysis, we limited



ethanol use to 30,000 barrels per day.



     Case E analyzes only the summer quarter, when vapor pressure



constraints are tighter.  This case makes one extreme assumption



about the summer quarter: that the refining industry's normal



reliance on seasonal storage of gasoline components to meet



summer demand does not occur, and that all summer demand must be



met by current production.  This seems unlikely, however, because



the industry has 300 million barrels of gasoline storage capacity,



in part because it normally produces extra gasoline in the winter



and spring and stores it to meet summer demand peaks (Schwartz,



memo in Docket EN-84-05, September 11, 1984e).  Case El looks at



the RVP change in isolation, while case E2 combines it with 5



percent higher daily demand for gasoline, which is typical of the



summer quarter.  In both cases, distillate production is kept at



the annual average  (whereas it normally is reduced in the summer



to allow for greater gasoline production), and no account is taken



of imports.  Case E3 combines case E2 with a 50,000 bpd reduction

-------
                              11-47






in distillate.  In all three cases, the cost estimates presented



later are only for one quarter, and are not comparable to cost



estimates for the other (annual-cost) runs.  The other measures



of "tightness," however, can be compared.




     Case F allowed the model more flexibility in choosing the



crude slate.  The base case constrained the solution to use as



much heavy crude at 0.10 gplg as it did at 1.10 gplg.  This sensi-



tivity analysis allowed the solution to include some substitution



of high-quality light crudes for heavy crudes if that was the



most economical approach; but even in this sensitivity analysis,



the swing was not allowed to exceed 250,000 barrels per day.



     Case G allowed the model to use an alternative octane booster,



MMT, at up to 0.05 grams per gallon, but only in leaded gasoline.



That compares to the average of 0.0625 grams per gallon of MMT



used in unleaded gasoline until its use in unleaded was banned



because of its adverse effects on tailpipe hydrocarbon emissions.



The use of MMT is still permitted in leaded gasoline, although



data on current levels of use are not publicly available.  The



base case makes the pessimistic assumption that no MMT will be




used.



     Case H assumed lower octane demand, with the premium share



of demand for unleaded gasoline falling to 23 percent.  This




could reflect several possibilities.  First, if the price of



leaded rises above that of unleaded, some current legal users of



leaded gasoline (as well as misfuelers) may switch to unleaded



regular; as discussed in Chapter VII, the experiences of the




U.S. Armed Services and the U.S. Postal Service suggest that

-------
                              11-48






many vehicles designed to operate on leaded regular  (89 octane)



will perform satisfactorily on 87 octane unleaded.   During the



public hearings, several refiners also mentioned that regular



leaded gasoline in California has been reduced to 88 octane.



Second, this drop in octane could occur if sharp increases in



the marginal cost of producing octane led some refiners to slightly



reduce the octane levels of their products.



     Case I altered the assumption about the ability of refineries



to segregate different types of naphthas for further processing.



This sensitivity analysis allows for greater segregation of



naphthas, as was assumed in the estimates made for the Preliminary



RIA.






II.D.2.b.  Results of Sensitivity Analyses



     Table 11-11 presents the results of varying the parameter



values individually.  In all cases, the estimates are for 0.10



gplg in 1986.  Note that these results cover only PADDs I-IV/VI,



the part of the country where difficulties, if they occur, are



likely to show up first.  The first line shows the results for



PADDs I-IV/VI under the base case assumptions.



     As the other lines in the table show, the impacts of varying



the individual assumptions are modest.  The maximum  cost increase



is less than 19 percent, and that occurs only under  the higher of



the two high-octane scenarios.  That scenario assumes that the



premium share of unleaded demand increases at 5 percentage points



each year (from a base of 24 percent in 1984), which seems highly



unlikely.  Indeed, that rate of growth in the premium share  is so

-------
                                 11-49
TABLE ll-ll.  Effects of Varying Individual Parameters/Assumptions:
Case
Base
Al
A2
B
C
D
El
E2
E3
F
G
H
I
I
1
Changes I
None
27% premium
35% premium
85% utilization
Old catalysts
Reduced alcohol
Summer RVP
El , summer demand
E2, summer distillate
Crude flexibility
Use of MMT
23% premium
Naphtha segregation
Reformer utilization
[%) by reformer type
jOW
90
90
90
85
90
90
90
90
90
90
90
90
90
Med.
71
72
82
84
87
77
77
90
90
70
69
70
66
High
36
36
36
36
36
36
36
36
36
36
36
36
36
Agg.
69
69
74
74
77
72
72
78
78
68
68
68
66
Marginal
Cost of Total
Octane Cost
(jzf/barrel) ($ million)
31.6
32.8
33.0
29.2
32.5
33.0
33.2
34.4
31.6
31.6
29.0
30.3
25.2
531
541
629
486
617
554
135*
136*
127*
513
489
496
441
  *Cost for summer quarter only

-------
                              11-50






high that it implies that the absolute amount of unleaded regular



gasoline would fall 6 percent from 1984 to 1986, a highly improb-



able occurence.



     Looking at the other measures of tightness, the most severe



impacts occur under case E2, which assumes summer RVP and 5



percent higher demand with no distillate adjustment, seasonal



storage, or imports.  Even in that case, however, the marginal



cost of producing octane rises less than 9 percent (to 34 cents



per octane-barrel), and utilization of high-pressure reformers



does not rise above the minimum level that we forced the model



to use.  Overall reformer utilization was only 78 percent of



capacity, up from 68 percent in the base case.  These increases,



of course, would be balanced by lower-than-average operating




costs, marginal costs, and reformer severities in the other three



quarters when RVP standards are higher than the annual average.



Case E3 shows that adding even small distillate flexibility brings



the marginal cost of octane back down to the base case value.



     Case B, lowering maximum downstream utilization from 90



percent to 85 percent, yielded counter-intuitive results; the cost



is slightly lower than in the base case, as is the marginal cost



of octane.  The utilization of the medium-pressure reformers



increased in this case, from 71 percent of stream-day capacity to



84 percent.  Closer examination of the output of the model revealed



that 85 percent utilization raised the cost of both the 1.10 gplg



and the 0.10 gplg cases, as expected, but reduced the difference



between the two.  Tightening that constraint appears to have raised



the total cost of producing octane, but lowered its incremental

-------
                              11-51






cost over the relevant range.




     On the positive  side of  the  ledger, the most  important



parameter change was  case I,  which allowed more optimal segrega-



tion of naphthas, as  in  the runs  made for the Preliminary RIA;



that reduced the costs in PADDs I-IV/VI from the base case by 17



percent, to $441 million.  It also reduced overall reformer



utilization from 68 to 66 percent of capacity, and had a more



dramatic impact on the marginal cost of producing octane, reduc-



ing it by 20 percent, to 25.2 cents/barrel.  Allowing the use of



MMT in leaded gasoline,  case G, also had a significant impact on



costs.




     In addition, we  ran sensitivity analyses that varied several



parameter values simultaneously to probe the limits of feasibility;



the results are shown in Table 11-12.  They should be interpreted



carefully, for while  each parameter change may be plausible alone,



increasing the number of simultaneous negative changes generates



increasingly implausible circumstances.  Furthermore, market



forces of supply and  demand create feedback loops that make some



of the extreme cases  exceedingly  unlikely.  For example, if the



marginal cost of producing octane rises (due, say, to lower



utilization of downstream capacity), the price of high-octane



unleaded premium is likely to rise, reducing the demand for that




product (or at least  forestalling the increases assumed in cases



Al and A2)  and the demand for high-octane alcohol is likely to




increase, not decrease.



     Lowering the utilization rate for downstream equipment and



simultaneously reducing  the allowable amount of alcohol, run Ml

-------
                              11-52
TABLE 11-12.  Effects of Varying Multiple Parameters/Assumptions:
              PADDs I-IV/VI
Run
Base
Ml
M2
M2a
M3
M3a
M3b
M4
M4a
M5
M5a
M6
M6a
Changes**
None
Utilization (B) ,
less alcohol (D)
(B), (D), plus
old catalysts (C)
(B),(D),(C), plus
lower premium (I)
(B) ,(C) ,(D), plus
higher premium (Al
(B),(C),(D),(A1),
plus MMT (G)
(B)f(C),(D),(Al),
plus MTBE and
oxinol
Old catalysts (C) ,
higher premium (Al
less alcohol (D)
(C),(A1) ,(D),
plus MMT (G)
Old catalysts (C) ,
less alcohol (D) ,
summer RVP (El)
(C) ,(D) ,(E1), plus
lower premium (I)
(C) ,(D) ,(E1), plus
higher premium (Al
(C) ,(D),(E1) ,(A1),
plus MMT (H)
Reformer Utilization
(%) by Reformer Type
Low
90
85
85
85
) 85
85
85
),
90
90
90
90
) 90
90
Med.
71
85
85
85
85
85
85
90
90
90
90
90
90
High
36
38
78
70
78
68
61
40
36
58
58
62
64
Agg.
69
75
83
82
83
81
80
79
78
83
83
84
83
Marginal
Cost of
Octane
(jd/barrel)
31.6
33.4
59.4
52.0
61.0
49.4
48.3
49.4
41.4
54.2
51.3
85.4
48.3
Total
Cost
($ million)
531
524
775
754
830
717
773
763
655
217*
202*
231*
197*
                (Table 11-12 continues on next page)

-------
                             11-53
TABLE 11-12.   fP»nfrin..o,l)
                         Reformer Utilization
                         (%) by Reformer Type
Marginal
Cost of    Total
Octane     Cost
Run Chanaes** T.^w M-3^ - Hinh AqqT (//barrel)
M7
M7a

M7b

M7c

M7d

(C) ,(D) ,(A1), plus
summer RVP and
demand (E2) 90 90 89 90 87.0
(C),
plus
(C) ,
plus
(C),
(G),
(C),
MTBE
(D),(A1),(E2),
MMT (G) 90 90 86 89 62.8
(D) ,(A1) ,(E2),
storage 90 90 52 82 54.5
(D),(A1) ,(E2),
plus imports 90 90 55 83 48.2
(D),(A1),(E2),
, oxinol, and
summer distillate 90 90 76 87 48.2
($ million)
267*

220*

215*

193*


218*
    *Cost  for  summer  quarter  only

   **Letters refer  to parameters listed  in Table 11-10

-------
                              11-54






in Table 11-12, has little impact on any of the measures.  Also



eliminating the use of any new FCC catalysts, as in run M2, makes



for substantially tighter, though still feasible, operating



conditions, raising the total cost to $775 million and the mar-



ginal cost to 59.4  cents/octane-barrel.  Run M2a made the same



negative changes as run M2, but lowered the demand for premium



unleaded to 23 percent (2 percentage points below the base case,



but only 1 percentage point below current demand); compared to



run M2, that cut the cost about 3 percent, and reduced the marginal




cost of octane by about 12 percent.



     Run M3 added higher demand for premium to the changes made



in run M2.  That further tightened the predicted operating condi-



tions, but did not make the attainment of the 0.10 gplg standard



infeasible.  Run M3a made those same changes, but allowed the



use of MMT, which cut total and marginal costs significantly.



Run M3b also duplicated M3, but allowed additional purchases of



oxinol (10,000 bpd) and MTBE (27,000 bpd) at prices that reflect-



ed both their cost and their adverse impact on fuel economy.



It also cut total and marginal costs substantially.  Both of



these  increases are within existing capacity for those products.



     Run M4 probed a slightly different combination of sensitivi-



ties,  combining low ethanol use, high octane demand, and old



catalysts  (as well as no imports, no MTBE, etc.).  Again, costs



increased  but the industry would still be able to comply.  Once



more,  adding some MMT (run M4a) substantially reduced costs.



     Runs M5-M7 were summer sensitivity runs.  Again, to probe  the

-------
                              11-55






limits of feasibility, we began by assuming no seasonal storage




was available.  Run M5 assumed that  no new catalysts would be



used, that alcohol use could not exceed 30,000 barrels per day,



and that all gasoline would have to  meet summertime RVP specifica-



tions with no distillate flexibility, imports, or MTBE use.



This is a very demanding, and unlikely, set of events, but 0.10



gplg remained feasible.  Lowering the demand for premium (run



M5a) brought the  total and marginal  costs down significantly;



storage of high octane components produced during the first half



of the year  (as is done by most refiners now) would be one way



to reduce the amount of premium gasoline that had to be produced



in the summer.  The cost estimates for these two runs (and the



subsequent ones in the table) are only for the summer quarter.



Again, the higher-than-average costs in the summer (when RVP



standards are tighter than average)  are balanced by lower-than-



average costs in  the winter (when RVP standards are looser than



average) .



     Run M6, which added higher demand for unleaded premium to



run M5, generated a model solution that must be characterized as




infeasible.  The marginal cost of producing octane was extraordi-



narily high, and  the model ran so much extra crude oil that some



of the products of crude distillation had no outlet; they were




simply dumped.  The conditions assumed in that run, however, are



collectively highly implausible.  If the marginal cost of octane



were to rise to the level shown in run M6, the price of high-octane



unleaded premium would rise sharply, making the surge in unleaded



premium demand and lack of use of existing MTBE and oxinol capacity

-------
                              11-56






contemplated in run M6 impossible.  Moreover, because run M6



only applies to summer conditions, normal use of existing storage



capacity could alleviate the summertime crunch, since the same



case (M4) with annual average RVP levels was feasible.  It is



also interesting to note that modest use of MMT, as allowed in



run M6a, restored feasibility and brought the estimated marginal



cost down to a much more comfortable level.



     Run M7 added higher overall gasoline demand to the changes



in run M6; again, the result was infeasibility, though for an



even more implausible set of conditions.  Again, storage of



gasoline or high-octane gasoline components produced in the



winter or spring could alleviate this summertime problem since M4



is the same run for the full year.  Relaxing some of our least-



likely conservative assumptions in other areas also restored



feasibility.  For example, run M7a shows that MMT restored fea-



sibility, bringing the marginal cost of an octane-barrel back



down to 62.8 cents.  Run M7b shows that 400,000 barrels per day



of seasonal storage also restored feasibility, reducing the



marginal cost of octane to 54.5 cents, and run M7c shows that



taking account of imports and MMT (but not storage) reduced



costs even further.  Run M7d shows the results of allowing the



normal seasonal distillate flexibility, use of existing MTBE



capacity, and use of 10,000 bpd more of existing oxinol capacity,



but neither imports nor seasonal storage.  Again, feasibility is



restored.




     To summarize the summer sensitivity runs, it takes an



extremely unlikely combination of high-cost sensitivities (e.g.,

-------
                               11-57






higher premium demand,  no use  of  the  newer catalysts, lower use




of ethanol,  summer RVP  standards, high  total demand, no imports,



no MTBE,  no  MMT,  no  increase  in oxinol, no seasonal storage, no



distillate flexibility, and no trading  with PADD V) to produce



an infeasible result.   Relaxing the assumption about seasonal



storage,  within the  limits of  existing  storage capacity, always



restores  feasibility (case M4), as does relaxing many of the



other assumptions singly  (e.g., the assumption on MMT, on dis-



tillate flexibility,  or on using  existing MTBE capacity).  Since



the  industry currently manipulates to its advantage the use of



storage capacity, MMT,  MTBE,  swings in  distillate production,



and  imports  simultaneously, EPA believes that concern about



summer infeasibility problems  is  unwarranted.



     We also ran  the two most  extreme sensitivity analyses (runs



M3 and M7) on PADD V to check  our assumption that problems, if



they occured at all,  would show up first in PADDs I-IV/VI; Table



11-13 reports the results.  As a  comparison with the correspond-



ing runs  in  Table 11-12 shows, conditions in PADD V are projected



to be looser (as measured by  reformer utilization and the marginal




cost of producing an octane barrel) than in PADDs I-IV/VI.



     To assure ourselves that  the 0.50  gplg standard for the



second half  of 1985  would be  feasible,  even under worst-case



conditions,  we ran the  two most extreme sensitivity analyses,



shown in Table 11-14.  Even in the worst case examined (M7),



overall reformer utilization  did  not exceed 75 percent and the



marginal cost of producing an  octane-barrel did not rise above



24 cents.  The results show that  0.50 gplg in 1985 remains quite

-------
                           11-58
TABLE 11-13.  Sensitivity Analyses for 1986:  PADD V
Run
Changes
              **
Reformer Utilization
(%) by Reformer Type
Low  Med.  High  Agg.
  Marginal
  Cost of     Total
  Octane      Cost
(^/barrel) ($ million)
Base   None                 90

M3     Utilization (B),
       less alcohol (D),
       catalysts (C),
       higher premium (Al)  85

M7     Old catalysts (C),
       less alcohol (D),
       higher premium (M),
       summer RVP and
       demand (E2)          90
                          58   36   59
                          70   34   68
                          77   36   76
                         21.5
                         29.8
                         32.2
      *Cost for summer quarter only

     **Letters refer to parameters listed in Table 11-10.
               80
              107
               31'

-------
                            11-59
TABLE 11-14.  Sensitivity Analyses for 1985;  PADDs I-IV/VI
Run    Charges**	

Base   None

M3     Utilization  (B),
       higher premium  (Al)
       less alcohol  (D)
       old catalysts (C)

M7     Old catalysts (C),
       less alcohol  (D),
       higher premium  (Al)
       summer RVP and
       demand (E2)
                        Marginal
Reformer Utilization    Cost of    Total
(%) by Reformer Type    Octane     Cost
Low  Med.  High  Agg.  (^/barrel) ($million)
 90
 90
55
 85   78
84
36
     36
36
60
     71
75
  21
         22.3
23.4
79
           89
48'
      *Cost for summer quarter only

     **Letters refer to parameters listed in Table 11-10,

-------
                              11-60






feasible and not excessively costly under all circumstances.



     These sensitivity analyses clearly indicate that the phase-




down schedule contained in the rule is feasible with existing



equipment under expected conditions.  Even if several conditions




are more adverse, simultaneously, than in our base case, the



0.10 gplg standard remains feasible, although it may require



very efficient utilization of refineries and careful attention



to the full array of methods available for meeting octane demands



with reduced lead.  Only in the very worst cases, which combine



many adverse conditions simultaneously, does it appear that the




refining industry would experience great difficulty in complying



with the rule and then only for three months.  Such multiple



worst-case scenarios are useful for probing the limits of feasi-



bility, but they are too implausible to deserve much weight, and



certainly too implausible to affect the outcome of this rule-



making.  Moreover, as discussed in the next section, the proposed



rule to allow banking would render even those circumstances



feasible.






II.E.  Impact of Banking on Costs



     EPA has proposed to allow the "banking" of lead in 1985 for



use in 1986 and 1987 (50 FR 718; January 4, 1985), when the 0.10



gplg standard will apply.  Under this proposal, refiners will




have the option of reducing lead use before it is required, and



then applying those early reductions to increase the amounts of



lead they are allowed to use in the two succeeding years.  The



purpose of this provision is to give individual refiners lower

-------
                              11-61






costs and extra flexibility in reducing lead  (even though the



standards are both feasible and reasonable without it) without



increasing the total amount of lead used in gasoline between



1985 and 1987.  Although refiners would be under no obligation



to avail themselves of the right to bank, EPA expects that the



majority would, for at least three reasons:   (1) lead provides a



greater octane boost at low levels, so it is more valuable in



1986 and 1987 at 0.10 gplg than at the higher levels permitted



in 1985; (2) in the short run, the marginal cost of producing



octane rises with the amount of octane produced, again making



lead more valuable in 1986 and 1987 than in 1985; and (3) banking



lead rights in 1985 will give refiners extra  flexibility to deal



with unexpected problems (such as equipment breakdowns) in the



later years.






II.E.I.  Base-Case Banking Results



     It is difficult to predict precisely how banking will be



used.  Some refiners are likely to cut their  lead use vory quickly



so as to bank a large amount, either for later use at that refinery



or for sale to other refineries less well-equipped to produce



octane without lead, while others will bank relatively little.



Whatever the specific pattern followed, however, the experience



with interrefinery averaging ("trading") over the last few years




suggests that the refining industry will make effective use of



this mechanism to reduce costs by reallocating lead use to those



refineries (and, with banking, to those times) where lead has



the highest marginal value in reducing production costs.  (As

-------
                              11-62






discussed in Chapter I, refiners representing about three-fourths



of all refining capacity now trade in any given quarter.)



     To explore the potential impacts of banking, we examined two



possible phasedown patterns.  For this analysis, we assumed (as



above) that misfueling would continue unabated during 1985, when




the standard will be 1.10 gplg during the first half of the year



and 0.50 gplg during the second half of the year.  For 1986 and



1987, when the limit will be 0.10 gplg, we assumed that misfueling



would fall to 20 percent of its current level.  We also assumed



that misfueling would follow this same pattern with banking.  If



refiners use less than 1.10 gplg in the first half of 1985, or



less than 0.50 gplg in the second half, the marginal cost of



producing leaded gasoline will increase.  Each gallon of leaded



gasoline produced below the limit will produce banked rights



that have value, so the net marginal cost of producing a gallon



of leaded gasoline will be little affected.  Similarly, in 1986



and 1987, refiners who use banked rights to produce leaded gaso-



line with more than 0.10 gplg will face lower marginal costs of



production than they would have otherwise, but they also will



consume rights that have a market value, so the net marginal



cost will be higher, roughly the same as it would have been at



0.10 gplg.  As a result, it seems reasonable to make the same



misfueling assumptions with and without banking.




     Table 11-15 compares the schedule without banking to two



possible alternatives with banking.  Alternative 1 assumes that



refiners do not start banking until the second quarter of 1985,



at which point they use an average of 0.60 gplg, thus banking

-------
                            11-63


TABLE 11-15.  Alternative Phasedown Patterns with Banking (gplg)

                     1985 (by quarter)       1986        1987
Alternative	I    II  111-IV	

Without Banking      1.10  1.10  0.50        0.10        0.10

With Banking

  Alternative 1      1.10  0.60  0.40        0.25        0.19

  Alternative 2      0.80  0.60  0.45        0.30        0.21

-------
                              11-64






0.50 grams, on average, for each gallon of leaded produced.



(Note that this industry average could reflect wide variations



across refiners; some might not bank at all, while others saved



large amounts.)  Under alternative 1, we also assume some banking



in the last half of the year, with leaded gasoline averaging 0.40



gplg, slightly below the limit of 0.50 gplg.  A total of 7.0



billion grams of lead (about 22 percent of the total allowed in



1985) would be banked during 1985, allowing refiners to average



0.25 gplg in 1986 and 0.19 gplg in 1987.  Shaving 0.10 gplg from



the annual average in 1985 translates into a larger per-gallon



increase in 1986 or 1987, because the amount of leaded gasoline



produced in the later years is smaller; the total amount of lead



use over the three years, however, is the same as without banking.



     Alternative 2 assumes that some refiners are able to reduce



lead more quickly, so that banking begins in the first quarter of



1985.  Those extra banked rights from the first quarter are then



used to reduce banking slightly in the last half of 1985 and to



achieve slightly higher lead levels in 1986 and 1987.  The amount



banked is 9.1 billion grams.  Again, the total amount of lead



used is the same as without banking.



     Table 11-16 compares the year-by-year costs and the present



values of the costs with banking to those without.  The estimated




savings are substantial; $173 million for alternative 1 and $226



million for alternative 2.  These estimates probably understate



the actual savings that will be realized with banking, because




they do not account for the extra flexibility it allows in meeting



unexpected problems (e.g., equipment breakdowns or a sudden

-------
                            11-65






TABLE 11-16.  Refining Costs Under Alternative Phasedown Patterns,
with
Alternative
Without Banking
With Banking
Alternative 1
Alternative 2
Partial
1985
96

176
170
Misfueling (m
1986
608

420
378
illions or 1
1987
558

463
452
983 dollars)
Present
Value
1,105

932
879

-------
                              11-66






surge in summertime demand) and in reallocating lead use to



those refiners with higher marginal costs of producing octane.



     Table 11-17 reports the marginal costs of an octane-barrel



for 1985 (by quarter), 1986, and 1987 under the alternative



banking scenarios.  Under either alternative, the marginal cost



remains under 22 cents per octane-barrel in 1985 and rises to



only 26.7 cents in 1986.  These estimates should be compared to



the marginal costs without banking, shown in Table II-7.  It is



particularly interesting to note that with banking, the marginal



cost of octane never rises above the level shown for 1988, when



new equipment will first be available.  This result has several



important implications.  First, it indicates that, with banking,




the phasedown schedule is no more expensive at the margin than



it would be if the Agency delayed the phasedown until 1988, when



refiners can build new equipment.  Second, it suggests that



refiners would not find it cost-effective to build new capacity



before 1988, even if it were possible.  (It would be cheaper to



buy banked lead rights.)  Finally, it suggests that the 1985-to-



1987 phasedown (again, with banking) should not cause financial



difficulties for refiners who could profitably operate at 0.10



gplg once they had time to add equipment.  Indeed, to the extent



that some refiners have trouble obtaining loans for capital equip-




ment, these results suggest that they will have an easier time



from 1985 to 1987 than they will in later years, because during



the earlier years they can buy lead rights, the cost of which is



an operating expense, not a large capital outlay requiring loan



or equity financing.

-------
                            11-67
TABLE 11-17.  Impact of Banking on Marginal Costs of Octane
            	(cents per barrel)
Year
Quarter
1985
I
II
III-IV
1986
I-IV
1987
I-IV
Without
Banking
N.A.
N.A.
20.4
29.2
30.0
With
Alt. 1
N.A.
21.4
21.1
26.7
25.0
Banking
Alt. 2
17.6
21.4
21.1
22.5
25.0

-------
                              11-68






II.E.2.  Sensitivity Analyses with Banking



     We also reran several of the sensitivity analyses for 1986



with banking, with the results shown in Tables 11-18 (for alter-



native 1) and 11-19 (for alternative 2).  As a comparison of those



tables with Table 11-12 shows, banking greatly reduces the diffi-



culty of meeting the rule, even in the unlikely scenarios that



make several adverse changes simultaneously.  For example, without



banking, as shown in Table 11-12, cases M6 and M7 both drive the



marginal cost of an octane-barrel in PADDs I-IV/VI in 1986 over



85 cents and are essentially infeasible.  If refiners follow



alternative 2 with banking, however, they can use 0.30 gplg in



1986.  As Table 11-19 shows, reformer utilization can then fall



and the marginal cost of octane decreases by more than 50 percent,



to about 40 cents per octane-barrel.  This greater comfort in



1986, of course, is partly offset by higher costs in 1985, but



the reductions needed in 1985 should not strain the capacity of



the refining industry.  Moreover, this analysis of PADDs I-IV/VI



understates the benefits of banking because rights banked in



PADD V could be sold to refineries in PADDs I-IV/VI.  (Based



on a separate analysis of PADD V, we expect such transfers to



occur, as the model shows a lower marginal value of lead in PADD



V than in PADDs I-IV/VI, at any given level of lead use.)



     These analyses suggest that while banking is not necessary



to meet the phasedown schedule, it does yield significant cost



savings and, perhaps more importantly, provides an extra margin



of safety against unexpectedly adverse conditions (e.g., higher



octane demands or lower-than-expected ability to utilize down-

-------
                            11-69
TABLE 11-18.  Sensitivity Analyses for 1986 with Banking:
	Alternative 1, PADDs I-IV/VI          	
                           Reformer Utilization
                           (%) by Reformer Type
                  Marginal
                  Cost of     Total
                  Octane     Cost
Run
Base
Base
Ml
M2
M3
M4
Changes**
None (no banking)
None (with banking)
Utilization (B),
less alcohol (D)
(B), (E), plus
old catalysts (C)
(B),(C),(E), plus
higher premium (Al)
Old catalysts (C) ,
Low
90
90
85
85
85

Med.
71
60
81
85
85

High
36
36
36
49
55

Agg.
69
62
72
77
79

(jz< /barrel)
31.6
22.7
23.7
40.3
40.5

($ million)
531
420
365
489
531

       higher premium  (Al),
       less alcohol  (D)     90

M5     Old catalysts (C),
       less alcohol  (D),
       summer RVP  (El)      90

M6     (C),(D),(E1), plus
       higher premium  (Al)  90

M7     (C),(D),(A1), plus
       summer RVP  and       90
       demand (E2)
87



90


90


90
36



40


46


72
77



79


81


86
32.1
39.7
46.6
51.5
      *Cost for summer quarter only

     **Letters refer to parameters listed in Table 11-10
514



140*


149*


169*

-------
                            11-70
TABLE 11-19.  Sensitivity Analyses for 1986 with Banking:
	Alternative 2, PADDs I-IV/VI        	
                           Reformer Utilization
                           (%) by Reformer Type
                  Marginal
                  Cost of    Total
                  Octane     Cost
Run
Base
Bank
Ml

M2
M3
M4
Changes**
None (no banking)
None (with banking)
Utilization (B) ,
less alcohol (D)
(B) ,(E), plus
old catalysts (C)
(B) ,(C) ,(E), plus
higher premium (Al)
Old catalysts (C) ,
Low
90
90

85
85
85

Med.
71
58

78
85
85

High
36
36

36
43
46

Agg.
69
61

71
76
77

(i /barrel)
31.6
22.7

23.3
34.2
36.2

($ million)
531
378

324
411
440

       higher premium (Al),
       less alcohol (D)      90

M5     Old catalysts (C),
       less alcohol (D),
       summer RVP (El)      90

M6     (C),(D)f(El). plus
       higher premium (Al)  90

M7     (C),(D),(A1), plus
       summer RVP and
       demand (E2)          90
84
89
90
90
36



36


36



61
75
78
78
84
29.8



30.5


40.0



42.6
      *Cost for summer quarter only

     **Letters refer to parameters listed in Table 11-10.
453
121'
127'

-------
                              11-71






stream refinery units, such as FCC units and reformers).   Moreover,



these gains are achieved without any increase in the amount of



lead allowed.

-------
                           CHAPTER III



               HUMAN EXPOSURE TO LEAD FROM GASOLINE





     Estimating the health benefits of an environmental regulation



requires predicting how the regulation will affect human exposure



levels,  in most cases, exposure estimates require extensive model-



ing of emissions, dispersion patterns, population distributions,



and the amounts of inhaled or ingested material that are absorbed



by the human body.  Such modeling requires that many parameters be



estimated, often on the basis of very limited information.



     In the case of lead in gasoline, however, exposure can be



assessed directly using several large data sets that make it



possible to relate lead in gasoline directly to lead in individ-



uals' blood, without taking the intermediate steps of dispersion



modeling, etc.  Analyses of these different data sets have



shown a strong and consistent relationship between the amount of



lead in gasoline and the amount of lead in blood, a relationship



confirmed by experimental data as well.



     This chapter presents the methods used to estimate the impact



of reducing lead in gasoline upon levels of lead in the blood of



children and adults.  These projections are used in Chapters IV



and V to estimate health benefits for children and adults, respe-



ctively.  Section A of this chapter addressess the basic issues,



including the data and the statistical methods used.  Section B



discusses the question of causality.  Section C presents estimates



of the numbers of children whose blood lead levels would be reduced



below various levels as a result of reducing lead in gasoline.

-------
                              III-2






III.A.  The Relationship Between Lead in Gasoline and Lead in Blood




     Individuals are exposed to lead from gasoline through many



pathways.  When leaded gasoline is burned in an engine, small



amounts are deposited in the engine and exhaust system, but most



of it is emitted from the tailpipe to the air, where it remains



suspended for a time before settling to the ground.  Some



exposure occurs through direct inhalation of the emitted lead.



Additional exposure occurs from ingestion of lead-contaminated



dust, or inhalation of such dust that has been stirred up.   Lead



from gasoline also deposits on food crops and is then ingested.



These multiple routes make it very difficult to model individual



exposure pathways.  It is possible, however, to estimate the



total amount of lead exposure from gasoline using statistical



methods, as discussed below.






III.A.I.  Recent Studies



     Several recent articles have shown that blood lead levels



for all age groups will fall as gasoline lead content falls.



The first important statistical studies were done by Billick et



al. (1979), who showed a strong relationship between the blood



lead levels of several hundred thousand children screened in New



York City's lead screening program and local gasoline lead use.



Figure III-l on the next page shows this relationship graphically.




     In 1982,  Billick et al. presented additional regression



analyses on data from New York City's lead screening program (with



data on several additional years); a Chicago screening program



(800,000 children over more than ten years); and a Louisville,



Kentucky program, all of which confirmed the earlier results.

-------
                                III-3
                             FIGURE III-l
                    Relationship Between Gasoline
                 Lead and Blood Lead in New York City
   35  -
^^M

8
IU
                              BLACK
                            HISPANIC
                       GASOLINE LEAD
   30  -
O
III
HA , y,\
•u I/ \ .X\ / \ / \
* K \ / \ •' v \ -
1 7 v \/
1 ,5 - \y
IU
o

10 -
. . t 1 1 . 1 1 1 1 1 1 t 1 t 1 t
1 Vv-V
V ^ A
/\ v>-
vv
V

i _i_ J. j i j ! • j i
z
m
5-° s
AD (Billion olgr
o
<*
K
3.0 2.

                                                              6.0
    1970
1971     1972    1973     1974     1975
         QUARTERLY SAMPLING DATE
                                                   1976

-------
                              III-4






     An EPA report (Janney, 1982) found a strong association be-




tween gasoline lead and blood lead in the data from the second



National Health and Nutrition Examination Survey (NHANES II),



after controlling for age, race, sex, degree of urbanization, and



income.  Analyses of the NHANES data by the Centers for Disease



Control (CDC) and the National Center for Health Statistics perform-



ed concurrently showed similar results.



     Annest et al. (1983) published a paper analyzing the nation-



wide downward trend of blood lead levels from 1976 to 1980 that



was demonstrated in the NHANES II data.  Blind quality-control



data were used to ensure that there was no drift in laboratory



measurements.  This downward trend was present after controlling



for age, race, sex, region of the country, season, income, and



degree of urbanization, and was present in each age-sex-race sub-



group.  Gasoline lead was a significant predictor (p < 0.001) of



blood lead after controlling for age, race, sex, degree of urbani-



zation, income, season, and region of the country, both in all



groups and separately for blacks, whites, white males, white



females, white children, white teenagers, and white adults.



     A recent paper by Schwartz et al. (1984b) of EPA's Office of



Policy Analysis presented the results of a study of the relation-



ship between blood lead levels and gasoline lead.  Several data



sets were employed for this analysis, including the NHANES II



and the CDC lead poisoning screening program.  The statistical



results indicated a highly significant regression coefficient



for gasoline lead levels, which was consistent across all of




the data sets.  Estimates of environmental lead from sources

-------
                              III-5

other that gasoline indicated that paint and other dietary lead
were not the primary sources of the observed decline in blood
lead levels during the 1970s.
     Gasoline lead was strongly associated with both the level
of lead in human blood and with the prevalence of elevated blood
lead levels.  The association appeared to be causal because other
factors, such as changes in dietary lead and paint lead, did not
account for the changes in blood lead levels that have been asso-
ciated with gasoline lead; the results suggested that more than
one-half of the lead in the average American's blood in the second
half of the 1970s was due to gasoline.  Since gasoline lead usage
in that period was restricted by regulation to about 60 percent
of what otherwise would have occurred, those regulations appeared
to have reduced substantially the average blood lead level in the
U.S. and the number of children with lead toxicity-

III.A.2.  Available Data Sets
     Because gasoline sales and the use of various gasoline
additives are regulated by federal law, information on them is
available from the Department of Energy and the Environmental
Protection Agency.  Several data sets contain information on
individual blood lead measurements.  Most of these data sets
target children in high-risk groups who have been screened for
lead poisoning.  One data set, however, NHANES II, is a large
representative sample of both adults and children in the U.S.
This section describes the data sets used in Section III.A.3. to
estimate the relationship between gasoline lead levels and
blood lead levels.

-------
                              III-6






III.A.2.a.  Gasoline-Use Data



     We combined monthly data on national leaded gasoline sales



(from the U.S. Department of Energy) with quarterly average lead



concentrations in grams per gallon (reported to EPA by refiners)



to compute national monthly gasoline lead usage.  For the Chicago,



New York, and Louisville areas,  we used gasoline sales data from



the Ethyl Corporation's monthly survey of area gasoline sales,



combined with national grams per gallon of lead, to obtain metro-



politan gasoline-lead usage.






III.A.2.b.  The NHANES II



     The data base for the regressions used to estimate the



coefficients in our prediction models was the health and demo-



graphic information collected in the NHANES II survey.  The U.S.



Bureau of the Census selected the NHANES II sample according to



rigorous specifications from the National Center for Health



Statistics so that the probability of selection for each person



in the sample could be determined.  The survey used subjects



chosen through a random multi-stage sampling scheme, designed



to utilize the variance minimization features of a stratified



random sample.  A total of 27,801 persons from 64 sampling areas



was chosen as representative of the U.S. non-institutionalized



civilian population, aged six months through 74 years.  Of those




27,801 persons, 16,563 were asked to provide blood samples,



including all children six months through six years and half of



those aged seven through 74 years.  The nonrespondent rate for



blood samples was 39 percent and did not correlate with race,

-------
                               III-7


sex, annual family  income, or  degree of urbanization.*  A study

of the potential  nonresponse bias  indicated that this was not

a significant problem  (Forthofer,  1983).

     Lead concentrations  in the blood of sampled persons and con-

trol groups were  determined by atomic absorption spectrophotometry

using a modified  Delves Cup micro-method.  Specimens were analyzed

in duplicate, with  the average of  the two measurements being used

for the statistical analysis.  Bench quality control samples were

inserted and measured  two to four  times in each analytical run to

calibrate the system.  In addition, at least one blind quality-

control sample was  incorporated with each 20 NHANES II blood

samples.  No temporal  trend was evident in the blind quality-

control measurements  (National Center for Health Statistics, 1981).

     The NHANES II  data did, however, display a marked relation-

ship between blood  lead and gasoline lead.  Figure III-2 plots

gasoline lead and blood lead over  time.  Note how closely the

changes in blood  lead  track the changes in gasoline lead, follow-

ing both seasonal variations and the long-term downward trend.

Figure III-3 plots  blood  lead  as a function of gasoline lead after

controlling for age, race, sex, income, degree of urbanization,

region of the country, educational level, smoking, alcohol consump-

tion, occupational  exposure, dietary factors, and interactions

among those factors; again, note the strong relationship.
* Because children were less likely to respond (51 percent of
  the children did not provide blood for lead determinations in
  the NHANES II data set), they were double-sampled.  The weights
  used to adjust the data to the national population accounted
  for both the oversampling and under-response of the children.

-------
FIGURE III-2
    110
   I100
     90
   G
   OT
   w
   a



   S
i
2 70
   •O
   w
   O

   O

   O
   Ul
     50 H
        LEAD  USED  IN  GASOLINE  PRODUCTION AND

        AVERAGE NHANES II  BLOOD LEAD LEVELS

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

                     GASOUNE
       AVERAGE

       BLOOD

       LEAD LEVELS
'/
          1976
                 1977
                1978

                YEAR
1979
1980
                                                        16
                                                        n
                                                        w


                                                     15 1
                                                      14
                                             » *
                                             Up
                                             in
                                                           9
                                                       -11
                                                     -10
                                                           a
                                                           ID
                                                           O
                                                  I
                                                  ao

-------
    19


    18


    17

T3

3  16
o

3  .
Q


m  14
o
UJ


I  13


    12


    11


    10
          III-9

       FIGURE III-3


NHANES II Data: Blood Lead versus Gasoline Lead
         10  2.2  2.4   2.6  Z8  aO   32  3.4   a6  3.8   4.0  4.2  4.4  4.6  4.8  5.0  5.2   5.4  56  5.8   &0

                               ADJUSTED GASOLINE LEAD (100 METRIC TONS PER DAY)

            Note-  Each point represents the average of approximately 310 observations.  However
                  the  regression line shown is the true regression line for all 9987 individuals.

-------
                              ni-io





III.A.2.c.  The CDC Screening Program for Lead Poisoning



     During the 1970s, CDC provided funds for community-wide



screening programs in many cities.  Approximately 125,000 children



were screened each quarter of the year.  Results reported back



to CDC included the number of children screened, the number with



lead toxicity (defined as free erythrocyte protoporphyrin levels



above 50 ug/dl and blood lead levels over 30 ug/dl), and the



number with severe lead toxicity-  Figure III-4 shows a plot of



gasoline lead versus the percent of children with lead toxicity.





III.A.2.d.  Chicago, New York, and Louisville, Kentucky Data



     Billick analyzed detailed data from the screening programs



in New York, Chicago, and Louisville, including the average



blood lead levels as well as the percent of children with blood



lead levels over 30 ug/dl.  We have analyzed these data further,



focusing particularly on the Chicago data,  which include a



l-in-30 sample of the over 200,000 blood lead screening tests



performed in Chicago between 1976 and 1980.   Figures III-5 and



III-6 show the relationship between gasoline lead and blood lead



in Chicago and New York, respectively.





III.A.3.  Statistical Analyses of Exposure



     We used multiple regression analysis to examine both the



relationship between gasoline lead and  individual blood lead,



and the relationship between gasoline lead  and the probability



of undue lead exposure (above 30 ug/dl)  or  lead toxicity (in the



CDC data).  For data from the NHANES II  survey,  we performed the




analyses on individual blood lead measurements,  with the explanatory

-------
    8
~  7
#
i.
i
Q  5
    4
£
2  o
1
5  2

    1
                                               in-ii

                                            FIGURE  III-4
                             CDC Data:  Gasoline Lead versus Percent of Children
                                                with Lead Toxiclty
        10  12   14   16   18   20  22  24   26  28  30   32   34  36   38   40  42   44   46  48   50

                         GASOLINE LEAD (BILLIONS OF GRAMS PER CALENDAR-QUARTER)

-------
   25

   24
                                               111-12
                                            FIGURE III-5
Chicago  Data-. Gasoline Lead versus  Blood Lead
   23
 O)
^D
Q
S
i20
m
   19
   18

   17
   16
         0.15   0.18   0.21    0.24    0.27   0.30   0.33    0.36   0.39   0.42    0.45   0.48   0.51    0.54
                           GASOLINE LEAD  (BILLIONS OF GRAMS PER CALENDAR-QUARTER)

-------
   24



   23



   22



   21


I
-2 20
Q


n 19



   18



   17



   16



   15
CO
                                                 111-13


                                               FIGURE  III-6


                                 New  York City Data:  Gasoline  Lead  versus  Blood Lead
         0.20    0.25     0.30    0.35    0.40    0.45    0.50     0.55    0.60    0.65     0.70     0.75     0.80

                           GASOLINE LEAD (BILLIONS OF GRAMS PER CALENDAR-QUARTER)

-------
                              111-14






variables being gasoline lead, age, race, sex, degree of urbaniza-



tion, alcohol consumption, smoking, occupational exposure, dietary



factors, lead in canned food, region of the country, educational




attainment, income, season of the year, and interactions among



these variables.  We also ran separate regressions for various



age, race, and geographic groups.  We weighted the results by the



the inverse of the probability of selection to produce nationally



representative estimates.  The regressions were run using a pro-



gram (SURREGR) designed to account for the stratified, clustered



sampling procedure in the NHANES II survey, and many different



specifications were considered.



     Table III-l presents the results for the linear regression



of gasoline lead and other explanatory variables on blood lead



levels.  (See Appendix C.I for definitions of the other variables.)



Table III-2 presents the results for the logistic regression on



the probability of blood lead level over 30 ug/dl for children



aged six months to 7 years.   Table III-l shows that gasoline lead



was a highly statistically significant predictor of blood lead



levels (p < 0.0001; t-statistic  > 60)  even after accounting for



potential confounding variables.  The average value of gasoline



lead usage was 426 metric tons per day, suggesting that an average



of about 9 ug/dl of blood lead was due to gasoline lead during



that period.   Additional regression results on the NHANES II




data, including results for specific subgroups used in later



analyses, are in Appendix C.




     For the CDC data, only quarterly data on the percent of



children screened with and without lead toxicity were available.

-------
                                    I11-15
Effect
Intercept
Gasoline Lead
Low Income
Moderate Income
Child (under 8)
Number of Cigarettes
Occupational ly Exposed
Vitamin C
Teenager
Male
Male Teenager
Male Adult
Small City
Rural
Phosphorus
Drinker
Heavy Drinker
Northeast
South
Midwest
Educational Level
Riboflavin
Vitamin A
^ 	 ~ —
Coefficient
6.15
2.14
0.79
0.32
3.47
0.08
1.74
-0.04
-0.30
0.50
1.67
3.40
-0.91
-1.29
-0.001
0.67
1.53
-1.09
-1.44
-1.35
-0.60
-.188
0.018
Standard Error

0.037
0.059
0.034
0.125
0.000
0.063
0.000
0.05
0.19
0.26
0.26
0.085
0.10
0.00
0.03
0.10
0.11
0.14
0.25
0.02
0.005
0.000
P-Value

0.0000
0.0025
0.0897
0.0000
0.0000
0.0000
0.0010
0.1841
0.2538
0.0026
0.0000
0.0039
0.0003
0.0009
0.0007
0.0000
0.0028
0.0005
0.0115
0.0000
0.0186
0.0355
* The coefficients of the dummy variables show how much blood lead (in ug/dl)
  is, on average, attributable to a specific effect.  The coefficient of gaso-
  line lead shows the number of ug/dl of blood lead attributable to each 100
  metric tons per day of gasoline lead use.

-------
                               111-16


TABLE  III-2.   Logistic  Regression on  Probability of Blood Lead
	>  30  ug/dl  for  Children 6  months to 7 years	


                    Logistic Regression Results*


Black children = under 8 years old, 479 observations

Dependent variable:  1 if blood lead  is over 30 ug/dl; 0 otherwise

  Model Chi square = 39.63 with 5 D. F.

Variable       Beta       Std. Error       Chi square         P

                                              30.13         U.OOOO

                                              12.40         0.0004

                                              12.26         0.0005

                                               3.33         0.0679

                                               4.39         0.0361

                                               0.90         0.3433

Fraction of concordant pairs of predicted probabilities
and responses =  0.718


White children = under 8 years old, 2225 observations

Dependent variable:  1 if blood lead  is over 30 ug/dl; 0 otherwise

  Model Chi square  = 33.58 with 5 D.F.

Variable       Beta       Std. Error       Chi square         P

                                              43.93         0.0000

                                               8.59         0.0034

                                              17.21         0.0000

                                               3.23         0.0724

                                               5.36         0.0206

                                               2.31         0.1285

Fraction of concordant pairs of predicted probabilities and
responses = 0.637
Intercept
Gaslead
Poor
Age 1
Age 2
Age 3
-6.9468
0.8633
0.9815
1.1404
1.1938
0.5428
1.2656
0.2452
0.2803
0.6246
0.5696
0.5728
Intercept
Gaslead
Poor
Age 1
Age 2
Age 3
-8.1667
0.6331
1.2174
1.4332
1.7168
1.1405
1.2322
0.2160
0.2935
0.7978
0.7415
0.7503
*All  logistic  regression  results were  run  using PROC LOGISTIC
 within  the Statistical Analysis System  (SAS).  This procedure
 uses  individual  data where  the dependent  variable  is one  if
 the  individual  is  above  the threshold,  and  ze*-o otherwise.

-------
                               111-17

We performed both weighted  linear and logistic regressions on
the probability of  lead  toxicity as a function of gasoline
lead.
     Tables III-3a  and III-3b  show the results of regressing the
percent of children with  lead  toxicity against gasoline lead for
the 20 quarters between  1977 and 1981.   "Dummy" represents the
period before CDC modified  its definition of lead toxicity in
1978.  The correlation coefficients between gasoline lead and the
percent of children with  lead  toxicity (0.8027) and between the
change in gasoline  lead  and the change in the percent with lead
toxicity  (0.817) were both  very large, and the regression
coefficient was highly significant (p <  0.0001).  (The magnitude
of the coefficient  is not directly comparable to the NHANES II
both because of the difference in outcome variable -- percent
toxic versus blood  lead level — and the difference in the units
of gasoline lead.)  The  regressions predicted that if there had
been no lead in gasoline at all, there would have been approxi-
mately 80 percent fewer cases of lead toxicity.  This does not
mean that 80 percent of the cases were due solely to gasoline
lead, but rather that in 80 percent of the cases both gasoline
lead and other exposures were required to bring children above
the CDC definition  of lead toxicity.
     For the Chicago, New York, and Louisville data, we performed
logistic and linear regressions of the probability of blood lead
levels being over 30 ug/dl for each race separately, with
explanatory variables being age, season, and gasoline lead.
Gasoline lead was always significant, and explained the seasonal

-------
                                       111-18
    TABLE III-3a.  Regression of CDC Screening Data:   Percent of Children
                   with Lead Toxicity on Gasoline Lead
Variable
Constant
Gas Lead*
Dummy
Coefficient
1.363
0.1601
-1.036
T-Statistic
1.722
5.322
-1.524
Incremental R?_
0
0.5215
0.0427
P-Value
0.05
<0.001
0.07
    R2 = 0.687

    Durbin Watson statistic = 1.786

    Simple correlation coefficients;

                     Gas Lead          Dummy

    Percent Toxic     0.8027          0.4068

    Gas Lead            -             0.6927
* Gasoline lead is in billions of grams per quarter.   In the last quarter
  of 1978, gasoline lead use was 40 billion grams.  By the second quarter
  of 1980, regulatory action had reduced this to 20 billion grams.
    TABLE III-3b.   Regression of CDC Screening Data:   Change  in Lead Toxicity
                   on Change in Gasoline Lead
Variable
Constant
Delta Gas Lead
Coefficient
0.3675
0.3938
T-Statistic
1.43
5.84
P-Value
0.09
<0.0001
    R2 = .6670

    Durbin Watson statistic = 1.887

    Simple correlation coefficient;    0.8167

-------
                               111-19

variation  in elevated blood  lead levels.  We also regressed quar-
terly average  blood  lead  levels in Chicago and New York against
gasoline lead,  race, age, and  season.  Table III-4 shows the
results of  regressing the Chicago data for black children for the
18 quarters from  1976 until  mid-1980.  (Our results differ from
Billick's  in that  he only had  lead concentration values for two
quarters of the year and had to interpolate the others.  We had
lead concentrations  for all  four quarters and, not surprisingly,
using this better  gasoline-lead data gave a stronger relationship.)
Again, gasoline lead was an  excellent predictor of children's
blood lead levels  (p < 0.0001).  Autocorrelation corrections were
run, and showed no significant correlation, nor did they produce
a noticeable change  in gasoline lead coefficients.  Since the
gasoline lead  coefficient in Table III-4 was in billions of grams
per quarter for the Chicago  area, we had to adjust from local to
national units  to  compare it to the NHANES II results.  Scaling
by the ratio of Chicago's gasoline use to the nation's, and
converting to  the  units used in the NHANES II regressions (100
metric tons per day), the Chicago coefficient would correspond
to a national  coefficient of 2.08, which is essentially identical
to the NHANES  II results (a  coefficient of 2.14, as shown in
Table III-l).
     Regressions on all of the data sets showed that gasoline
lead was an extremely significant explanatory variable both for
individual blood lead levels and for the percent of children with
undue lead exposure or lead  toxicity.  Gasoline lead appeared to
have accounted  for 60 percent of the lead in Americans in the

-------
                                111-20
TABLE III-4. Black Children in Chicago:  Regression of Average Blood~Lead
             Levels on Gasoline Lead Levels
Variable
Constant
Gas Lead*
Age 1
Age 2
Age 3
Age 4
R2 = 0.6194
Durbin Watson
Coefficient T-Statistic Incremental P.2
15.48 22.33 0
17.02 11.27 0.5757
0.85 1.53 0.0106
1.68 3.03 0.0418
1.01 1.84 0.0153
0.66 1.20 0.0065

statistic =2.01
P-Value
-
<0.0001
0.07
0.005
0.05
>0.10


Simple correlation coefficients:

Average Blood
Gas Lead
Gas Lead Age 1 Age 2 Age 3
Lead 0.7587 0.0009 0.01615 0.0337
000
Age 4
0.0342
0
*Gas Lead is lagged one month behind quarterly gasoline lead in billions
 of grams.  The average for this period in Chicago was 0.379 billion grams.

-------
                              111-21


second half of the 1970s, and to have explained both the seasonal

increases in blood lead levels from winter to summer and the

long-term drop in blood lead levels during the late 1970s.  EPA

refinery reports show that the rate of decline in gasoline lead

accelerated after late 1978, and this was paralleled by an accel-

erated decline in blood lead levels.

     Gasoline lead had the same coefficient in rural areas, in

urban areas, and in urban areas with populations over one million,

In the NHANES II regressions, a dummy variable for residence in

central cities versus suburban areas was not statistically signi-

ficant.  The gasoline lead variable that best correlated with

blood lead was gasoline sales for the preceding month.


III.B.  The Question of Causality

     While not crucial to a rule designed to be precautionary in

nature, we did examine whether there is a causal relationship

between gasoline lead and blood lead.  In epidemiology there

are several general criteria for determining whether association

represents causality (Kleinbaum et al., 1982; Lilienfeld and

Lilienfeld, 1980).  The most useful criteria are:

     Is there experimental evidence to support the findings?

     Do several studies replicate the results?

     Does a dose-response relationship exist?

     Are there consistent effects in different types of studies?

     Does cause precede effect?

     Is the model biologically plausible?

     Is it unlikely that other factors not included in the
     analysis would change the results?

-------
                              111-22






We will address these issues in turn, with particular emphasis on



the last one, as no study can ever measure all possible confounding




factors.






III.B.I.  Experimental Evidence



     Facchetti and Geiss (1982) investigated the contribution of



gasoline lead to blood lead in Turin, Italy during the late 1970s



by changing the isotopic composition of the lead added to gasoline,



and monitoring the isotopic composition of blood lead.  This iso-



topic lead experiment indicated that changes in the isotopic com-



position of air lead followed closely and rapidly changes in the



isotopic composition of gasoline lead.  Changes in the isotopic



composition of blood lead also paralleled changes in gasoline lead.



Based on modeling of the results,  Facchetti and Geiss estimated



that at least 25 percent of the (high) blood lead levels in Turin



were due to gasoline lead;  this was at least 6 ug/dl.  Since



blood lead isotopic ratios  were still changing when the gasoline



isotopes were switched back, the actual impact of gasoline



lead is probably higher.



     Manton (1977) analyzed isotopic changes in blood lead in the



United States and found the contribution of airborne lead (predomi-



nantly gasoline) was between 5 and 10 ug/dl in most of his subjects.



     Tera et al. (1985) recently analyzed the isotopic ratios



of blood lead in children in Washington, B.C. as the isotopic



ratio of air lead changed.   Their data showed that, as late as



1983, at least 38 percent of the lead in children's blood still



came from gasoline lead, despite the 50 percent reduction in




gasoline lead since 1978.

-------
                              111-23






     Our analysis of the impact of gasoline lead indicated that



in the late 1970s, about 9 ug/dl of blood lead resulted from the



lead in gasoline.  This magnitude of effect is similar to that



found in the  isotope studies.





11I.E.2.  Does Cause Precede Effect?



     One way  to determine whether the trend in blood lead was



caused by gasoline lead or was due to another variable is to



examine the lag structure in light of our biomedical knowledge.



The half-life of lead levels in the blood is about 30 days



(Rabinowitz et al., 1976).  Because the average blood test oc-



curred on the 15th of the month, the current month's gasoline lead



would have had only 15 days to affect blood lead levels, and so,



though significant, should have a lesser impact.  The previous



month's gasoline lead, on the other hand, represents emissions



on average 15 to 45 days prior to examination, and since direct



inhalation and even dust exposure shows up rapidly in the blood,



we would expect this one-month lagged gas lead to be more signifi-



cant, with a  noticeably higher coefficient.  Similarly, we expected



gasoline sold two months previously to be less significant and



of lesser magnitude.  If gasoline sales were merely a proxy for



time, however, all three months should be equally good predictors,



since t, t-1, and t-2 (where t is the month since commencing the



survey) equally represent the passage of time.



     We regressed the individual blood lead levels in the NHANES



II survey against current, one-month lagged, and two-month lagged



gasoline lead simultaneously, and found one-month lagged gasoline



lead was most significant and two-month lagged gasoline lead was

-------
                              111-24






least significant.  This suggested that the causal model was



correct.  The coefficient of two-month lagged gasoline lead was



one-half that of one-month lagged gasoline lead, which matches



the one month half-life of lead in the blood.






III.B.3.  Replicability and Consistency



     We replicated our analysis of the national NHANES II data



with our analysis of the site-specific Chicago and New York



screening data and with Billick's analysis of the screening data



from Chicago, New York, and Louisville.  In addition, Rabinowitz



and Needleman (1983) have examined umbilical cord blood from over



11,000 consecutive births at Boston Women's Hospital between 1979



and 1981.  They found a strong association (p < 0.001) with gaso-



line lead used in the Boston area, and also that one-month lagged



gasoline lead had the highest correlation.  No significant monthly



variation was noted in the mothers' education levels, smoking, or



drinking, and water lead levels increased somewhat over the period,



while blood lead and gasoline lead levels fell.  Thus, both local



and national data from different studies, collected by different



investigators, show the same pattern of gasoline lead being signi-



ficantly related to blood lead.



     The results of studies of both individual blood lead levels



and average blood lead levels were consistent with analyses of



elevated blood lead levels.  Billick examined the probability of



blood lead levels above 30 ug/dl in Chicago, New York, and




Louisville, and found a strong relationship to gasoline lead.



We repeated that analysis using logistic regressions, with the




same results.  Our investigation of the national CDC screening

-------
                              111-25





program data, using both linear and logistic regressions, found



the probability of lead toxicity was strongly dependent on the



amount of lead in gasoline.  Finally, we performed logistic



regressions on the NHANES II data of the probability of both black



and white children (under 8 years) and preteenagers (ages 8-14)



being over 30 ug/dl of blood lead, and again found a strong



relationship to gasoline lead levels.





III.E.4.  Does a Dose-Response Relationship Exist?



     To assure ourselves that the linear relationship we found



was the true form of the dose-response relationship, we divided



the NHANES sample in half.  We repeated the regression for whites



in the second half of the survey period, when the average gasoline



lead levels were roughly 50 percent of those in the first half.



The gasoline lead coefficient was essentially unchanged at



1.94 (compared to 2.14 for the full sample).  This indicated a



stable linear relationship.



     For blacks, the sample size was too small to divide the



sample, so we used an alternate procedure.  We regressed log



(blood lead) against demographic variables and log(gasoline



lead), for a range of zero gasoline intercepts.  This model



chose the best exponent for the relationship blood lead = (gasoline



lead)B.  For the intercept (8 ug/dl) with the highest R2, the



exponent was 0.98, indicating a linear dose-response relationship.



A square-root regression was also tried, and gave an inferior



fit to the linear regression.  The Chicago data, as noted before,



gave the same magnitude of gasoline lead's effect on blood lead,

-------
                              111-26






and visual examination of the regression plots in Figures  III-3



to III-6 confirm the linearity  in the relationship.






III.B.5.  Biological Plausibility



     Gasoline lead produced 90 percent of the emissions of lead



into the air in the 1970s, and it was the major source of  lead



contamination of the environment.  Lead is emitted as predominantly



respirable particulates (less than 1 micron) from auto exhausts.



Respirable particulates reach into the lung and show a high absorp-



tion rate.  Gasoline lead is a major source of lead in street and



household dust and in soil contamination and, therefore, in addition



to direct inhalation, results in secondary exposure through the



inhalation and absorption of dust.  Roels (1976)  has shown a high



correlation between blood lead and lead on the hands, presumably



from air and dust contamination.  Lead is known to be absorbed



from both the lung and gut.   Thus, the fact that gasoline lead



is related to blood lead is biologically plausible.






III.B.6.  Control for Confounding Factors



     Billick's analysis of children controlled for age and race,



and our reanalysis controlled for season as well.   The Annest



et al. analysis of the NHANES II data controlled for age, race,



sex, income, degree of urbanization, region of the country, and



season.




     We used several approaches to control for confounding factors.



Where we had sufficient information, we included potential con-



founding factors,  or surrogates for them,  in our regressions.



In addition, we examined external data to check for changes in

-------
                              111-27





confounding factors that might bias our results, and performed



several  internal statistical checks to examine the likelihood of



misspecification error.





III.B.6.a.  External Validation



     A great deal  is known about the sources of lead exposure.



The major general  sources of body lead are food, water, and paint.



While we have limited data on the specific sources of exposure



of the individuals in our data sets, if all other general exposures



were, on average,  constant during the period, the effects of these



will be part of the constant term in our regressions.  Bias would



occur only if these sources changed over the period.



     The Food and  Drug Administration's estimates of lead in the



diet (based on market basket surveys) during the NHANES II period



are shown on Table III-5.  As there was no downward trend in



dietary lead intake, this was unlikely to have been a potential



source of bias in  the model.  Lead in canned food did change



during the period, and we included a variable for this in our



regression model.  (See Appendix C.)  Dietary variables were used



to represent differences in food consumption between individuals,



which were also accounted for by age, sex, race, income, education,



and regional variables.

-------
                              111-28






TABLE III-5.  Lead in the Diet (micrograms/day)
Fiscal Year
1976
1977
1978
1979
1980
Infants
21
22
25
36
— —
Toddlers
30
28
35
46
—
Males aqed 15-20
71.1
79.3
95.1
81.7
82.9
     The exposure to lead from water is predominantly a function



of pH levels.  The pH of drinking water supplies appeared to have



been constant over the four to five-year period of our data.



     Change in paint lead exposure was also an unlikely source



of bias for several reasons.  First, adult blood lead levels



decreased by almost as much (37 percent vs. 42 percent) as child-



ren's blood lead, and adults generally do not eat paint.   Second,



because ingestion of paint lead usually results in the absorption



of enough lead to produce large increases in blood lead,  we



would expect a drop in paint lead exposure to reduce blood lead



levels only in people whose levels are above the mean.  However,



the drop in blood lead recorded in the NHANES II data shifted the



entire distribution dramatically.   Indeed, even low blood-lead



groups showed major declines in blood lead.  This would not have



occurred if ingestion of paint lead were the determinant.



    Third,  the drop occurred across geographical boundaries, in



suburbs as  well as central cities.  (Suburbs have a lower fraction



of pre-1950 housing stock, and, therefore, inherently less expo-

-------
                              111-29


sure to lead paint, yet they showed the same drop and the same

gas lead coefficient.*)  Finally, lead paint removal programs

during this period reached only 50,000 of the 30 million housing

units with lead paint (less than 0.2 percent), so exposure was

unlikely to have changed in this period (Morbidity and Mortality

Weekly Reports, 1976-1980).

     To lay these issues to rest, however, we did an additional

analysis of the Chicago data.   For each year of the screening

program, the CDC reported the percent of lead toxic children

with a lead paint hazard in their home or the home of a close

relative.  In 1978, however, the homes of all screened children

in Chicago (no matter what their blood lead level) were checked

for lead paint hazard.  This survey of over 80,000 housing units

established the general prevalence of lead paint exposure in

the screened population.  With the probability of paint lead in

the house given lead toxicity, and the probability of paint

lead in the house in general,  we used Bayes Theorem to compute

the probability of lead toxicity given paint lead in the house,

and the probability of lead toxicity when lead paint was not in

the house.   We then regressed these quarterly probabilities on

gasoline lead.  The results are shown in Table III-6.
* Shier and Hall (1977) analyzed over 2,500 housing units in
  Pittsburgh and found that the fraction that had lead paint
  concentrations above 2 mg/cm^ decreased from over 70 percent
  in pre-1940 housing to about 43 percent in 1940-59 housing,
  and to 13 percent in post-1960 housing.

-------
                              111-30
TABLE III-6.  Chicago Probability of Blood Lead > 30 ug/dl
              With and Without Paint Lead Hazard in the Home
Variable
Without Lead
Paint Hazard
Constant
Gas Lead
With Paint
Lead Hazard
Constant
Gas Lead
Coefficient t-Statistic
-2.087 -0.6583
35.398 4.423
12.80 0.579
80.16 1.436
P-Value
0.26
0.0005
0.30
0.07
     As we expected, gasoline lead was a more significant

explanatory variable for children not exposed to lead paint

than for the exposed group.  Over 80 percent of the children

were not exposed to lead paint at home, and among them gasoline

lead was highly significant and explained most of the elevated

blood lead levels.  It was also striking that even among children

with an identifiable lead paint hazard in their homes, gasoline

lead was still strongly related to the probability of lead

toxicity-


III.B.6.b.  Seasonality

     U.S. blood lead levels were strongly seasonal, with summer

levels substantially higher than winter levels.  However, when

the NHANES II data were tested with seasonal variables,  none of

them was statistically significant, or even close to significant,

when gasoline was included; the highest F-statistic for any

season was 0.82.   This indicated that the same gasoline lead

-------
                               111-31

coefficient  successfully explained both  the short-term  increases
in  blood  lead  levels  from winter  to summer and  the  long-term
decrease  in  blood  lead levels  over the four-year period.  The
six regressions  on the Chicago, New York, and Louisville data,
as  mentioned before,  also indicated that seasonal!ty was insigni-
ficant  once  gasoline  lead was  in  the model.  This suggested that
other long-term  trends in lead exposure cannot have biased the
gasoline  coefficient, as the short-term and long-term gasoline
coefficients were  the same.

III.B.6.C.   Other  Time Trends
     To test the hypothesis that  there was another unknown lead
factor  that  was  decreasing over the period, and whose effects
might be  attributed to gasoline in our regressions, we repeated
our analysis with  time as a variable; time was entered as the
number  of days after  February  1,  1976 that a blood sample was
taken.
     The  results indicated that time was not significant when
gasoline  lead was  in the regression.   Moreover, the effect of
gasoline  lead on blood lead was reduced by only 23 percent if
we  kept the  insignificant time variable in.  In addition, the
Chicago data for average blood lead were analyzed with time as a
variable.   Here, again, gasoline  lead was significant (p < 0.02)
while time was insignificant (p > 0.30).   We also tested time-
squared on the CDC data,  and it was insignificant.

III.B.6.d.   Geographic Sampling Pattern
     To be certain that the pattern of geographic sampling over
time in the NHANES II period did  not produce changes in blood lead

-------
                              111-32






that the regression falsely attributed to gasoline, we included



dummy variables for four regions of the country in our analysis.



To check further, we inserted dummy variables for all 48 locations



identified by the National Center for Health Statistics (NCHS).



NCHS did not release locational data for 16 counties with popu-



lations below 100,000;  these were all represented by one additional



dummy variable.  Rerunning the regressions with these identifiers



changed the gasoline lead coefficient by only about 5 percent,



which was insignificant.



     This result was extraordinary because the NHANES II sampled



different cities at different times, and the regression allowed



the differences in blood lead levels over time to be attributed



to changes in city location rather than gasoline lead.   This meant



that the coefficient of gasoline lead was heavily determined by



the month-to-month changes in blood lead and gasoline lead within



each city during the two months or so spent at each site.   Yet



the results did not change.



     In addition, we performed a regression that had an interac-



tion term between gasoline lead and the city identifiers.   This



procedure allowed a different slope in the gasoline lead/blood



lead relationship for each county.  The regression still yielded



a coefficient for national gasoline lead of 1.83, with p < 0.0001.



     Finally, the fact  that analyses of many individual cities



across the country (Boston, Chicago, New York, and Louisville)



yielded similar results suggested geographic location was  not an




important source of bias.

-------
                               111-33





III.B.6.e.  Subgroup Analysis



     To verify the  robustness  and stability of the relationship



between gasoline  lead and blood lead, we ran regressions of the



NHANES II data for  several demographic groups separately.  This was



done both to verify that we properly controlled for these vari-



ables, and to investigate whether the observed drop in blood lead



levels was, in fact, due to a  universal exposure such as gasoline



lead, or could have been due to a source primarily affecting



certain subgroups.  The stability of the gasoline lead coefficient



across these subpopulations reduced the likelihood of a speci-



fication bias.



     This was because the bias introduced into the gasoline



coefficient due to  an omitted  variable would be proportional to



the regression coefficient that variable would have had if



included.  If the gasoline lead coefficient was insignificantly



different among the regressions run for different subgroups, the



omitted variable either must not have significantly biased the



gasoline lead coefficient, or must have been coincidentally



constant among all  the demographic groups.  The only known such



variables were gasoline lead and food lead.



     We performed separate regressions for males, females, adults,



and children, for cities over  one million, and for smaller urban



areas.  The maximum pair-wise  difference in the gasoline lead



coefficient among the six subgroups was less than 10 percent.  In



addition, we changed the definitions of large city (from over



1,000,000 to over 250,000), of rural (to include rural areas or



cities under 10,000), and of the age categories.  The coefficient

-------
                              111-34


of gasoline lead changed by only 2 percent when we performed this

regression.  The stability of our findings, given the many addi-

tional tests we conducted, testifies to the robustness of the

relationship between blood lead and gasoline lead.  In addition,

a study by Annest et al. (1983) of the U.S. Public Health Service,

also using data from the NHANES II, found that the only reasonable

explanation for the decline in blood lead levels was the decline in

the amount of lead in gasoline.  The Centers for Disease Control,

the epidemiological branch of the Public Health Service, has also

endorsed this relationship in its comments on this proposed rule

(Public Docket EN 84-05) and the 1982 rule.


III.C.  Impact of Rule on Numbers of Children
        Above Various Blood-Lead Levels

     In estimating the benefits of reducing lead in gasoline, we

relied on the regression coefficients estimated from the NHANES II

data set because it is the largest and most representative of those

available.  Estimates of adult health benefits use the continuous

functional forms because the relationship between blood lead and

blood pressure is a continuous function of blood lead levels;

their application is discussed in Chapter V.  The benefit esti-

mates for children, however, are functions of the numbers of

children brought below various blood lead levels.  Section III.C.I

describes how we used the results of the NHANES II logistic

regressions to estimate changes in the numbers of children above

these various levels.  Section III.C.2 discusses the interpreta-

tion of these results with respect to prevalence and incidence.

-------
                               111-35





III.C.I.   Estimation  Procedure



     To estimate  the  numbers  of  children  above  different  blood



lead levels, we relied  on  logistic  regressions  estimated  from



the NHANES II  data, of  the type  reported  in Table  III-2.   These



regressions were  estimated separately  for blacks and whites, and



for each  two-year age group from 6  months through  13 years.  In



each regression,  the  dependent variable was the natural logarithm



of the odds of being  above the level,  while the independent



variables  were various  demographic  factors and gasoline lead.



     To predict how the number of children above each level



would change as the amount of lead  in  gasoline was reduced, a



mechanism  was  needed  to forecast the distribution  of blood lead



levels as  a function  of gasoline lead.  In this analysis, we



assumed that the  distribution of blood lead would  remain  log-



normal as  gasoline lead levels declined.  Then, estimates of the



mean and variance of  the associated (transformed)  normal  distri-



bution could be used  to determine the  percentage of the popula-



tion above any specific blood lead  level.  The estimates  of the



mean and standard deviation of the  underlying normal distribution



were derived from logistic regression  estimates of the percentage



of children with  blood  lead levels  above  30 ug/dl  and SURREGR



estimates  of the  mean of the log-normal distribution using the



Statistical Analysis  System (SAS) procedure, SURREGR.



     If the -distribution 'X1 is  normal with mean 'u1 and  standard



deviation  's1   (X:N (u,s)),  then  Y = exp (X) is log-normal with a



mean of 'a1 and a standard deviation of 'b1, where

-------
                              111-36






(III-l)          a = exp (u -I- 1/2 s2) and






(III-2)          b = exp (2u + s2) (exp (s2) -1)






Further, if eg and vg are the same percentiles of the log-normal



and its corresponding normal distribution, respectively, we have






(III-3)                6g = eXp (U + Vg S).






We used the logistic regressions to estimate eg in eguation



(III-3) and the SURREGR regressions to estimate a in eguation



(III-l), which yielded






(III-4)                a = exp (u + 1/2 s2)






(III-5)                eg = exp (u + vg s)






Solving these equations for u and s produced a quadratic equation:






(III-6)        0 = (In (eq) - In (a)) - v  s + .5s2






which had the solution s = v  +  [v 2 - 2 (In (e ) - In (a))]  °-5,






Only the smaller root yielded sensible values for u and s.   Then



u = ln(a)-l/2 s2.  Using the estimated values for u and s,  we



determined percentages of the distribution above 10, 15, 20, and



30 ug/dl by looking up the results of (In (10) - u)/s, etc., in



the normal table.



     We used a logistic regression eguation to estimate the



percentage of children over 30 ug/dl to control for problems of



multiple sources of exposure.  If we had simply used the regres-




sions explaining the mean and assumed a constant standard devia-

-------
                              111-37





tion, we would have predicted that removing lead from gasoline



would have resulted in there being no children above 30 ug/dl.



This seemed unreasonable because paint and food are known



alternate sources of lead, and also are associated with high



blood lead levels.  The logistic regressions confirmed that the



geometric standard deviation changes as the mean falls.  Because



of the sensitivity of the blood lead distribution to age, we



estimated separate distributions for each two-year age interval.



The tabulated changes in the numbers of children above various



levels represented the sum of distributions for each age category.



The regression results are shown in Appendix C.



     For children from six months to seven years old, we used



logistic regressions for the percent above 30 ug/dl of blood



lead.  For children aged eight to thirteen, we used logistic



regressions for the percent above 20 ug/dl blood lead because



there were too few observations above 30 ug/dl for the logistic



procedure to yield accurate estimates.



     Table III-7 presents the estimated reductions in the numbers



of children over various blood lead levels from 15 ug/dl to 30



ug/dl.  The estimates are presented for three phasedown schedules:



the original proposal (0.10 gplg starting 1/1/86); the alternative



discussed in the NPRM (0.50 gplg on 7/1/85, 0.30 on 1/1/86, 0.20



on 1/1/87, and 0.10 on 1/1/88); and the final rule (0.50 gplg on



7/1/85 and 0.10 on 1/1/86).  All of the estimates assume that



misfueling is totally eliminated.  (The impacts of alternative



assumptions about misfueling are explored in Chapter VIII.)

-------
                             111-38
TABLE III-7.  Estimated Reductions in Numbers of Children Over
              Various Blood Lead Levels,  Assuming No Misfueling
              (thousands of children)
Blood Lead Level
Rule
30



25



20



15



10



ug/dl
Proposed
Alternative
Final
ug/dl
Proposed
Alternative
Final
ug/dl
Proposed
Alternative
Final
ug/dl
Proposed
Alternative
Final
ug/dl
Proposed
Alternative
Final
1985

0
22
22

0
72
72

0
232
232

0
696
696

0
1,972
1,972
1986

52
46
52

172
154
172

563
501
563

1,726
1,524
1,726

4,949
4,354
4,949
1987

47
45
47

157
149
157

518
491
518

1,597
1,508
1,597

4,595
4,333
4,595
1988

43
43
43

144
144
144

476
476
476

1,476
1,476
1,476

4,261
4,261
4,261
1989

39
39
39

130
130
130

434
434
434

1,353
1,353
1,353

3,918
3,918
3,918
1990

36
36
36

119
119
119

400
400
400

1,252
1,252
1,252

3,637
3,637
3,637
1991

32
32
32

106
106
106

357
357
357

1,125
1,125
1,125

3,283
3,283
3,283
1992

31
31
31

103
103
103

348
348
348

1,098
1,098
1,098

3,215
3,215
3,215

-------
                         111-39





III.C.2.  Incidence Versus Prevalence



     Our predicted decrease in the number of children above a



given threshold is for a specific point in time; our cost esti-



mates are for an entire year.  If children remain above 30 ug/dl



for less than a year, there will be more children above 30 ug/dl



in a year than we estimated and our benefits will be understated.



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



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



benefits.



     This raises the difficult epidemiological issue of prevalence



versus incidence.  Prevalence is the percent of people who have



the condition of interest at a particular time (e.g., the percent



of people with the flu on February 14).  Incidence is the percent



of people who develop new cases of the flu in a given time period



(e.g., the month of February).  Prevalence is incidence times



average duration.



     This issue is important because the NHANES II survey,



upon which we based our regressions, measured the prevalence of



cases above 30 ug/dl blood lead or other thresholds, rather than



the incidence.  Yet the benefits we want to estimate would, in



fact, be reduced numbers of cases in a time period, i.e.,



incidence.



     Clearly an excursion of a child's blood lead level above



30 ug/dl for a day or two will produce less damage than a pro-



longed elevation.  However, data indicate that such occurrences



are not very likely.   Odenbro et al. (1983) in Chicago found



fairly stable blood lead levels in individual children with high

-------
                              111-40


levels.  For these children, levels remained high for more than

a few days, usually for months or years.  However, if the average

duration of elevated blood lead was six months,  the actual number

of children affected in a year would be twice the average preva-

lence for the year.  This obviously would affect our benefit

estimates.

     Because, as discussed in Chapter IV, we only valued cognitive

losses for children in CDC categories III and IV,* and because

data from Odenbro et al. suggested that such children's blood lead

levels remain elevated for a long time unless treated, we believe

our prevalence estimate is reasonable for estimating cognitive

effects.  Medical management costs, on the other hand, seem more

reasonably associated with incidence.

     In any case, it was necessary to determine  the duration of

elevated lead levels.  To do this we evaluated several available

pieces of information.   They all suggested that  the average

duration was less than one year, so that our estimate of prevalence

(based on the NHANES data) understated annual incidence.
* CDC classifies children as "lead toxic"  if they have blood lead
  levels above 30 ug/dl and free erythrocyte protoporphyrin (FEP)
  levels above 50 ug/dl.  (FEP is a measure of the derangement of
  the heme synthesis process caused by lead.)   Children with blood
  lead levels between 30-49 ug/dl but with FEP under 50 are in
  CDC category Ib.   Children between 30-49 ug/dl blood lead
  and 50-109 ug/dl FEP are category II.   Category III children
  have blood lead levels between 30 and  49 ug/dl and FEP levels
  between 110 and 249 ug/dl or blood lead  levels between 50 and
  69 ug/dl and FEP levels between 50 and 249 ug/dl.   Category
  IV includes all children with blood lead greater than 70 ug/dl,
  and children with blood lead between 60  and 69 ug/dl and FEP
  levels above 250 ug/dl.

-------
                              111-41





     Our first source was the CDC screening program data.  This



program screened about 100,000 to 125,000 children per quarter



of the year to detect lead toxicity.  Approximately 6,000 to



7,000 cases were found each quarter; this established the general



prevalence of lead toxicity in the screening population.  However,



this prevalence rate showed strong within-year variation, with



levels much higher in the third quarter, summer, when gasoline



consumption was also highest.  This intra-year variation suggested



that the average duration was not so long that the effects of



quarterly changes in exposure were swamped by cases that origin-



ated in earlier quarters.



     We also have used the CDC lead screening data in another



way.  CDC reported, quarterly, the number of children under



pediatric management, which included all the new cases discovered



during that quarter plus the children remaining under pediatric



management who had been discovered with lead toxicity in the



previous quarters.  We compared that number to the sum of the



cases detected in the same quarter plus the previous two quarters



and found the results were quite close.  This suggested that



children remained under pediatric management for an average of



three quarters.  However, children generally had several medical



visits after their blood levels returned to normal to ensure that



the decline was real.  This implied that the average duration of



blood lead levels above 30 ug/dl was even shorter, closer to two



quarters.   If this is true, then it is possible that we have



underestimated the annual incidence of cases of children above




30 ug/dl by as much as a factor of two.

-------
                              111-42






     The amount of time it takes for lead toxicity percentages to




respond to fluctuations in gasoline lead levels also may help to



determine the duration of lead toxicity.  If this time is rela-



tively short (e.g., a few months or less), it is unlikely that



duration would extend beyond a year.  For lead toxicity to last



a year or more, one would expect lead toxicity levels to be



relatively insensitive to intra-annual variation in gasoline



lead.  The CDC data, which show such variation, suggest that the



average duration is substantially less than one year.



     Two other data sets supported the conclusion of a short lag



between gasoline lead and blood lead levels.   First, in the



NHANES II data, we examined both the lag structure of blood lead's



relationship to gasoline lead, and whether any seasonal dummy var-



iables were significant in explaining the large observed seasonal



variations in blood lead.   Schwartz et al. (1983)  found that the



lag structure of average blood-lead levels'  dependence on gasoline



lead could extend to three months.



     In addition, Billick (1982)  examined the results of the



screening programs for lead toxicity in Chicago (800,000 children



screened) and in New York City (450,000 children screened)  over a



10-year period and found a strong seasonal pattern in the number of



children with lead toxicity.  This pattern followed the seasonal



variation of gasoline use.  When Schwartz et  al. (1984b) analyzed



these data in a logistic regression, gasoline explained the



cyclical variation in blood lead levels, with no seasonal variable



obtaining a p-value of better than 0.38.

-------
                              111-43





     All of this suggested that the average time a child spent



above 30 ug/dl was short enough so that quarterly prevalence



rates corresponded well to quarterly exposure incidence.



Therefore, our estimate of the annual incidence of children



above 30 ug/dl is likely to be low, as is our estimate of



avoided medical expenses.

-------
                            CHAPTER IV



         BENEFITS OF REDUCING CHILDREN'S EXPOSURE TO LEAD





     The scientific literature presents evidence of a wide range



of physiological effects associated with exposure to lead.  These



range from relatively subtle changes in various biochemical



measurements at very low levels of exposure, with uncertain impli-



cations for health, to severe retardation and even death at very



high levels of exposure.  Although such effects are found in indi-



viduals of all ages, particular concern has focused on children,



because thay appear to be at greater risk.



     Because the body is a complex structure of interdependent



systems and processes, effects upon one component will have



cascading implications throughout the body.  This interdependence



is well illustrated by multi-organ impacts resulting from the



inhibition of heme by lead, and consequent reduction in the body



heme pool.  These effects are depicted graphically in Figure IV-1,



taken from EPA's most recent Draft Lead Criteria Document



(p. 13-31).



     This chapter summarizes the available evidence of the



effects of lead on children, and develops rough estimates of the



benefits of reducing exposure to lead by reducing lead in gaso-



line.  Section A deals with the pathophysiological effects of



lead, while Section B addresses the evidence on neuropsychologi-



cal effects (primarily reduced cognitive ability).  The final



section discusses the methods used to monetize the benefits of



reducing children's exposure to lead.  Although these estimates



in no way cover the complete range of potential benefits, they



total several hundred million dollars per year for the rule



being promulgated.

-------
                                     IV-2


                     FIGUHE IV-1  Multi-Organ Impacts  of  Lead's
                                      Effect  on  the  Heine Pool
•IOIKI
«MH
































IMOI
IT root

IDTTMIWOItTIC KDUCIO
«»<"' • 7r:r "



•IUML MOUCIOMIiOniOTIIIIS
iMicn IM tTiocmnwiti """"



4
i
MMiiHOcm.^ „„„„,, ,„„,,,„. tf
UIICTI vuuiii o f(
., . 1 \

V








IHICTt * Will •IlllltttO
iunie«MTI

iiricnMiiimns
MOM. MO
KM»« Clltl
*
•»*IMOIITILIII>IIO>
MOMUVI CKOUCIIO*

•TAIIIIIIOIVIIOWIIII
oiHuvoiniriTiii

•tMIHIOIMdlllL
t«UI MMfOITUIS

•rumoMLCiiw
•Oil UIICMO
MBIUCdl

•»>!•! OULCIW
MLC • CTCLIC
•uciiOTioi m»ioinii

•MIIIIOOI10IIIICATIO*
•1 l«VIIIOMII«T*l


•rumo
M10IIIIO1IOI




MTIIIIOMItlflW
Of TUmOMUII


•Mimo
•voioirimot
01 COH1OOI


UUtOVUCUlM
oiniimrroiiciiMcn








^ MPMIIIOIOIIMO
t*o»ioi«iiOHifiiT















-*" LIVUSOr TMV»TOM.M|

*
oniunoioiiioOLlAiniii
MMOKUUHIITIII
IMCTIW
           Multi-organ impact of reductions of heme body pool by lead. Impairment of heme
synthesis by lead (see Section 12.3) results in disruption of a wide variety of important physio-
logical processes in many organs and tissues. Particularly well documented are erythropoietic,
neural, renal-endocrine, and hepatic effects indicated above by solid arrows (—»*•). Plausible
further consequences of heme synthesis interference by lead which remain to be more conclu-
sively established are indicated by dashed arrows (	•*-).

-------
                               IV-3






IV.A.  Pathophysiological Effects



     Elevated blood lead levels have long been associated with



neurotoxicological effects and many other pathological phenomena:



an article on lead's neurotoxicity was published as early as 1839,



on anemia in the early 1930s, on kidney damage in 1862, and on



impaired reproductive function in 1860.  From an historical per-



spective, lead exposure levels considered acceptable for either



occupationally-exposed persons or the general population have been



revised downward steadily as more sophisticated biomedical tech-



niques have shown formerly unrecognized biological effects, and as



concern has increased regarding the medical and social signifi-



cance of such effects.  In the most recent downward revision of



maximum safe levels, the Centers for Disease Control (CDC)



lowered its definition of lead toxicity from 30 ug/dl blood lead



and 50 ug/dl of free erythrocyte protoporphyrin (FEP) to 25 ug/dl



blood lead and 35 ug/dl FEP-  The present literature shows biologi-



cal effects as low as 10 ug/dl (for heme biosynthesis) or 15 ug/dl



(for renal system effects and neurological alterations).



     There is no convincing evidence that lead has any beneficial



biological effect in humans (Expert Committee on Trace Metal



Essentiality, 1983).



     The finding of biological effects at blood lead levels as



low as 10 ug/dl potentially has important implications for public



health, as such levels are common in the U.S. population.  As



Table IV-1 shows, between 1976 and 1980 over three-quarters of



children under the age of 18 had blood lead levels in excess of



10 ug/dl, and 15 percent exceeded 20 ug/dl.  The rates among blacks

-------
                                IV-4
TABLE IV-1.  Blood Lead Levels Of Children in the United States
	1976-80 (percent in each cell; rows sum to 100 percent)
                     <10 ug/dl
10-19
ug/dl
20-29
ug/dl
30-39
ug/dl
40-69
ug/dl
     All Races

     all ages            22.1     62.9     13.0     1.6     0.3
     6 months-5 years    12.2     63.3     20.5     3.5     0.4
     6-17 years          27.6     64.8      7.1     0.5     0.0

     White

     all ages            23.3     62.8     12.2     1.5     0.3
     6 months-5 years    14.5     67.5     16.1     1.8     0.2
     6-17 years          30.4     63.4      5.8     0.4     0.0

     Black

     all ages             4.0     59.6     31.0     4.1     1.3
     6 months-5 years     2.7     48.8     35.1    11.1     2.4
     6-17 years           8.0     69.9     21.1     1.0     0.0
 Source:  Table 1, Advance Data #79, May 12, 1982, from Vital and
 Health Statistics, National Center for Health Statistics (Supple-
 mental Exhibit 4.)  NOTE: These results were produced after
 adjusting the data for age, race, sex, income, degree of urbani-
 zation, probability of selection, and non-response to the NHANES
 survey.

-------
                               IV-5


and among preschool children were even higher.

     Lead's diverse biological effects on humans and animals are

seen at the subcellular level of organellar structures and pro-

cesses, and at the overall level of general functioning that

encompasses all of the bodily systems operating in a coordinated,

interdependent way.  The biological basis of lead toxicity is its

ability, as a metallic cation, to bind to bio-molecular substances

crucial to normal physiological functions, thereby interfering

with these functions.  Some specific biochemical mechanisms

involve lead's competition with essential metals for binding for

sites, inhibition of enzyme activity, and inhibition or alteration

of essential ion transport.  The effects of lead on certain sub-  <

cellular organelles, which result in biochemical derangements

common to and affecting many tissues and organ systems (e.g., the

disruption of heme synthesis processes), are the origin of many of

the diverse types of lead-based functional disruptions of organ

systems.

     Lead is associated with a continuum of pathophysiological

effects across a broad range of exposures.  In addition to the

high level effects mentioned above, there is evidence that low

blood-lead levels result in:

       1. Inhibition of pyrimidine-5'-nucleotidase (Py-5-N)
          and delta-aminolevulinic acid dehydrase (ALA-D)
          activity, which appears to begin at 10 ug/dl of
          blood lead or below (Angle et al., 1982).
          Hernberg and Nikkanen (1970) found 50 percent of
          ALA-D inhibited at about 16 ug/dl.

       2. Elevated levels of zinc protoporphyrin (ZPP or FEP)
          in erythrocytes (red blood cells) at about 15 ug/dl.
          This probably indicates a general interference in
          heme synthesis throughout the body, including inter-

-------
                               IV-6
          ference in the functioning of mitochondria (Piomelli
          et al. , 1977) .

       3. Changes in the electrophysiological functioning of
          the nervous system.  This includes changes in slow-
          wave electroencephlogram (EEC)  patterns and increased
          latencies in brainstem and auditory evoked potentials
          (Otto et al.,  1981, 1982, 1984) which begin to occur
          at about 15 ug/dl.   The changes in slow-wave EEC
          patterns appear to  persist over a two-year period.
          Also, the relative  amplitude of synchronized EEC
          between left and right lobe shows effects starting at
          about 15 ug/dl (Benignus et al.,  1981).  Finally,
          there is a significant negative correlation between
          blood lead and nerve conduction velocity in children
          whose blood lead levels range from 15 ug/dl to about
          90 ug/dl (Landrigan et al., 1976).

       4. Inhibition of  globin synthesis, which begins to
          appear at approximately 20 ug/dl  (White and Harvey,
          1972; Dresner  et al., 1982).

       5. Increased levels of aminolevulinic acid (ALA)  in
          blood and soft tissue, which appear to occur at
          about 15 ug/dl and  may occur at lower levels
          (Meredith et al., 1978).  Several studies indicated
          that increases of ALA in the brain interfered  with
          the gamma-aminobutyric acid (GABA) neurotransmitter
          system in several ways (Draft Criteria Document,
          p. 12-128 ff).

       6. Inhibition of  vitamin D pathways, which has been
          detected as low as  10 to 15 ug/dl (Rosen et al.,
          1980a, 1980b;  Mahaffey et al.,  1982).  Further, as
          blood lead levels increased, the  inhibition became
          increasingly severe.

These data cite the lowest observed effect  levels to date, and

do not necessarily represent  affirmative findings of thresholds

below which exposures can be  considered safe.

     The types of specific effects listed above as occurring at

blood lead levels below  25 ug/dl indicate (a) a generalized lead

impact on erythrocytic pyrimidine metabolism, (b) a generalized

lead-induced inhibition  of heme synthesis,  (c) lead-induced

interference with vitamin D metabolism, and (d) lead-induced

perturbations in central and  possibly peripheral nervous system

-------
                                IV-7






functioning.  The medical significance of all these effects is




not yet fully understood.  However, current knowledge regarding



the deleterious and vital nature of the affected physiological



functions both individually and in the aggregate suffices to



warrant both public health concern and efforts to minimize



their occurrence due to lead exposure.



     As lead exposure increases, the effect on heme synthesis



becomes more pronounced and effects broaden to additional bio-



chemical and physiological mechanisms in various tissues, causing



more severe disruptions of the normal functioning of many organ




systems.  At very high lead exposures, the neurotoxicity and



other pathophysiological changes can result in death or, in some



cases of non-fatal lead poisoning, long-lasting sequelae such



as mental retardation and severe kidney disease.



     This chapter discusses the known pathophysiological effects



of lead that occur in children, particularly the hematological




and subcellular neurotoxic effects, and the expected change in



the number of children at potential risk of those effects under



EPA's regulation.  EPA is considering these effects in its



current National Ambient Air Quality Standard process.






IV.A.I.  Effects of Lead on Pyrimidine Metabolism



     The best-known effect of lead on erythrocytic pyrimidine




metabolism is the pronounced inhibition of Py-5-N activity.  This



enzyme plays a role in the maturation of erythrocytes, as well as



erythrocyte function and survival; it controls the degradation



and removal of nucleic acid from the maturing cell (reticulocyte),

-------
                                IV-8


Interference with this process can increase red cell membrane

fragility.  As noted earlier, the disruption of this function by

lead has been noted at exposure levels beginning from 10 ug/dl.

At blood lead levels of 30-40 ug/dl, this disturbance is suffi-

cient to materially contribute to red blood cell destruction and,

possibly, decreased hemoglobin production contributing to anemia

(World Health Organization, 1977; National Academy of Sciences,

1972; Lin-Fu, 1973; Betts et al., 1973).  The significance of

this interference with pyrimidine metabolism transcends the red

cell; the mechanism of this inhibition suggests a wide-spread

impact on all organs and tissues.


IV.A.2.  Effects on Heme Synthesis and Related Hematological
         Processes

IV.A.2.a. Mitochondrial Effects

     The mitochondrion is the critical target organelle for

lead toxicity in a variety of cell and tissue types, followed

probably by cellular and intracellular membranes.  The scientific

literature shows evidence of both structural injury to the mito-

chondrion (e.g., Goyer and Rhyne, 1973; Fowler, 1978; Fowler et

al. , 1980; Bull, 1980) and impairment of basic cellular energetics

and other mitochondrial functions (e.g., Bull et al., 1975; Bull,

1977, 1980; Holtzman et al., 1981; Silbergeld et al., 1980).


IV.A.2.b.  Heme Synthesis Effects

     The best-documented effects of lead are upon heme

biosynthesis.  Heme, in addition to being a constituent of hemo-

globin, is an obligatory constituent for diverse hemoproteins in

all tissues, both neural and non-neural.  Hemoproteins play

-------
                                IV-9






important roles in generalized functions, such as cellular ener-



getics, as well as in more specific functions such as oxygen



transport and detoxification of toxic foreign substances (e.g.,



the mixed-function oxidase system in the liver).  Available data



on elevated arainolevulinic acid (ALA) and free erythrocyte proto-



porphyrin (FEP) levels, inhibited ALA-D, and the like show inhi-



bition in the heme biosynthetic pathway at low blood-lead levels,



with statistically significant effects detectable at 10-15 ug/dl



(Meredith et al., 1978; Piomelli et al., 1982; Angle et al., 1982).



This heme biosynthetic disturbance may result in the impairment



of many normal physiological processes in a host of organ systems



and/or the reduced reserve capacity of many cells or organs to



deal with other types of stress (e.g., infectious diseases).



     The interference of lead with heme synthesis in liver



mitochondria appears to result in the reduced capacity of the



liver to break down tryptophan, which, in turn, appears to



increase levels of tryptophan and serotonin in the brain (Litman



and Correia, 1983).  Such elevation of neurotransmitter levels



may be responsible for some of the neurotoxic effects of lead,



since elevated tryptophan levels have been associated with encepha-




lopathy, and elevated serotonin levels produce neurologic symptoms



similar to acute porphyria attacks.



     The elevation of ALA levels is another indication of lead's



interference in heme synthesis and mitochondrial functioning.




Such elevations can have serious neurotoxic implications.  In



vitro studies have shown that ALA can interfere with several



physiological processes involved in the GABA-ergic neurotrans-

-------
                              IV-10






mitter system, including a possible role as a GABA-agonist



(Brennan and Cantrill , 1979).  There appears to be no threshold




concentration for ALA at the neuronal synapse below which



presynaptic inhibition of GABA release ceases.



     Since ALA passes the blood brain barrier and is taken up by



brain tissue, it seems likely that elevated ALA levels in the



blood correspond to elevated ALA levels in the brain (Moore and



Meredith, 1976).  Furthermore, lead in the brain is likely to



enhance brain ALA concentrations because neurons are rich in mito-



chondria, the subcellular site of ALA production.  As mentioned



earlier, blood ALA elevations begin to be detectable at 18 ug/dl



of blood lead (Meredith et al., 1978).






IV.A.2.C.  Impact of Lead on Red Blood Cell Abnormalities



     High levels of blood lead ( > 40 ug/dl) are known to produce



anemia.  To examine the association between lead and red cell



abnormalities in individuals below 30 ug/dl, we performed two



analyses.  First, we examined the relationship between blood



lead levels and various red cell indices.  Second, because FEP



is a more stable indicator of a person's lead exposure over



several months than a single blood lead measurement, as well as



their sensitivity to lead, we also analyzed the relationship



between elevated FEP levels and anemia.  We found that blood




lead levels below 30 ug/dl were associated with increased risks



of microcytosis and hypochromia, and that FEP levels were



associated with increased risks of anemia in children, even



below 35 ug/dl of FEP.

-------
                              IV-11


     For our analysis we used data from the NHANES II survey.

Among the hematological information collected were mean corpus-

cular volume (MCV), mean corpuscular hemoglobin (MCH), serum

iron, hematocrit, FEP, and percent transferrin saturation.  We

used regression analysis of these data for 1,967 children under

the age of eight to determine whether there was a relationship

between blood lead levels and the presence of hematological

abnormalities.


IV.A.2.c.l.  Effects of Lead Exposure on Blood Cell Volume
             and Hemoglobin Content

     We found that blood lead was inversely related to both mean

cell volume (MCV) and mean cell hemoglobin (MCH), even for blood

lead levels below the current CDC 25 ug/dl guideline for deter-

mining lead toxicity.  Linear regressions were performed of MCV

and MCH on blood lead levels in children, controlling for race,

age, family-income, iron status (i.e., the level of iron in their

blood), degree of urbanization, and other nutritional factors.

(As previous work led us to expect, percent transferrin saturation

was a superior control for iron status compared to serum iron

and was used throughout our analysis.)

     Table IV-2 lists the variables considered in the regressions.

Income was found not to be a significant confounding variable

once we controlled for iron status, and was dropped from the

analysis.  Race also had no bearing on MCV once iron status was

controlled for, although it was a significant explanatory variable

for MCH.  This suggested that there may be additional dietary or

biochemical factors predisposing black children to lower erythro-

cyte hemoglobin levels.

-------
                              IV-12

TABLE IV-2. Variables Considered in Regressions of FEP, MCV,
	MCH, and Anemia                  	
Age under 2

Age 2-4

Age 4-6

Race

Sex

Degree of Urbanization

Family Income

Serum Albumin

Dietary Calcium

Dietary Calories
Serum Copper

Dietary Carbohydrates

Dietary Fat

Serum Iron

Blood Lead

Dietary Phosphorus

Dietary Protein

Transferrin Saturation

Dietary Vitamin C

Serum Zinc

-------
                               IV-13





     The regressions for both MCV and MCH found blood lead to be



a significant explanatory variable (p < 0.0001 and 0.0033, respec-



tively) for the decreases in each.  "Hockey stick" regressions



on the MCV relationship indicated a threshold at 10 ug/dl, the



same level at which heme synthesis disturbance by lead has been



reported to begin.  Figures IV-2 and IV-3 show these relationships



for MCV and MCH respectively.



     We also analyzed the probability of children having



abnormally low MCV and MCH levels as a function of blood lead,



since this is a clearer sign of physiological derangement.  For



this analysis, we used logistic regressions.  Once again, blood



lead was a significant explanatory factor (p < 0.0001), both in



mean cell volume being low (MCV < 80 femptoliters [fl]) and in



mean cell volume being seriously low (MCV < 74 fl).  Blood lead



levels were also significantly associated (p < 0.023) with the



percent of children having MCH less than 25 picograms (pg), but



only for children under six.



     To test the hypothesis that the relationship with MCV held



at low blood-lead levels as well as high blood-lead levels, we



repeated the regression for MCV < 74 fl using only those child-



ren whose blood lead levels were less than 25 ug/dl.  The regres-



sion coefficient for blood lead was unchanged and significant



(p < 0.014).  Thus, blood lead levels under 25 ug/dl were also



associated with increased risks of microcytosis.



     Because blacks have a higher incidence of thalessemia, we



repeated the MCV regressions for white children only.  Lead was

-------
                              IV-lH
                           FIGURE IV-2
The Relationship Between MCV and Blood Lead After Adjusting
             for all Other Significant Variables
       82.5
        79.5 -
                               ADJUSTED BLOOD LEAD
The regression line is the true regression  for  all  1,967  children.
For ease of display, each point represents  the  average  of about
100 children with consecutive blood lead  levels.

-------
                               IV-15
                            FIGURE IV-3

The Relationship between MCH and Blood Lead After  Adjusting
            for  All  Other Significant Variables
   MCH (pg/cell)

     28.0
     27.5  -
     27.0 -
     26.5
                  15
20         25
   LEAD (jjg/dl)
30
                                                               35
The regression  line is the true regression  line  for all 1,967
children.   For  ease of display, the points  shown represent the
average of  about  100 children with consecutive  blood lead levels.

-------
                               IV-16






still a significant predictor of MCV (p < 0.0001), after control-




ling for all other significant variables.  We also used Mentzer's



index, which indicates that persons with thalessemia generally



have an MCV/red cell count of < 11.5.  Even after we eliminated



all such people, lead was still a significant predictor of MCV,



with no change in the coefficient.



     To further investigate the relationship between lead and



abnormal hematological variables, we used our regression to



predict the percentage of children with MCV < 74 fl as a function



of blood lead for two cases:  children with average transferrin



saturation levels (22.4 percent saturated for children in the



NHANES II survey) and children with transferrin saturation levels



one standard deviation below average (13.6 percent).  The results



are shown in Figure IV-4.  Note that at 25 ug/dl of blood lead,



almost 10 percent of the children with average iron levels and 17



percent of the children with below average iron levels had MCVs



of less than 74 fl.



     The relative risk of children having MCV levels less than



74 fl when their blood lead levels were 25 ug/dl compared to



10 ug/dl was 1.98 (the 95 percent confidence interval was 1.44-



2.71).  Using the same 10 ug/dl reference point, the relative



risk at 20 ug/dl was 1.53 (1.27-1.95 at 95 percent).  Since



logistic regressions gave the same results when we used only



children with blood lead levels under 25 ug/dl, and since the



95 percent confidence limits on the relative risk did not include



1.0, these results showed increased risks of hematological abnor-



malities in children at blood lead levels of 20 ug/dl and below.

-------
                           IV-17

                        FIGURE IV-4
PERCENT OF CHILDREN WITH MCV BELOW 74
                 (Age 6 Months to 8 Years)
   25
  20
   15
 u
 "o
   10
                   With Transferrin Saturation
               One Standard Deviation Below Average
                                     With Average
                                  Transferrin Saturation
                  10      15     20     25

                     Blood Lead Level (pg/dl)
30
35
               Prediction of Percent of Children with MCV
               Below 74  fl as a Function of Blood Lead Levels

-------
                               IV-18






IV.A.2.c.2.  The Relationship Between Blood Lead and FEP



     The increased interference of lead in the formation of



hemoglobin, and consequent accretion of protoporphyrins in red



blood cells, has been well documented by Piomelli et al. (1982).



Our analysis of the NHANES II data confirmed that study's results.



Annest and Mahaffey (1985) have recently analyzed the relationship




between FEP levels and blood lead in the NHANES II data and found



a strong relationship after controlling for iron status.  We also



analyzed the NHANES II data and found that, even after controlling



for iron status using transferrin saturation,  the relationship



was very strong.



     A considerable body of literature exists  suggesting that FEP



levels are exponentially related to blood lead levels (Piomelli



et al., 1973; Kammholz et al.,  1972; Sassa et  al., 1973; Lamola



et al., 1975a, b; Roels et al., 1976).  To test this relationship



in the NHANES II population, we used several alternative specifi-



cations.  We considered a linear model, a model where FEP was



proportional to both exp(blood  lead) and exp(percent transferrin



saturation), a model where FEP  was proportional to exp(blood lead)



and (transferrin saturation)3,  and a model where FEP was propor-



tional to (blood lead)Bl and (transferrin saturation)B2-  The



model that that fit best was exp(blood lead) times (transferrin



saturation)6.  We examined the  possibility of  different additive



intercepts in this model and found the highest correlation coeffi-



cient and F-statistic for a zero additive constant.  This model



suggested an exponential relationship to blood lead and a power




law dependence on transferrin saturation.  "Hockey stick" analysis

-------
                              IV-19


of the relationship between PEP and blood lead gave a threshold

of 18 ug/dl.  Figure IV-5 shows the FEP-blood lead relationship.

     We also investigated the relationship between the probability

of elevated FEP levels and blood lead, and verified previous find-

ings.  Again using NHANES II data, we performed logistic regres-

sions on the probability of FEP levels being above 50 ug/dl* as a

function of blood lead, using both blood lead and log(blood lead)

as the independent variable; we obtained a better fit with blood

lead than with log(blood lead).

     Again, we checked to see whether the relationship between

the risk of elevated FEP and blood lead held at lower blood-lead

levels, repeating the regression only for children with blood lead

levels under 30 ug/dl.  Using multiple logistic regressions, blood

lead was again extremely significant (p < 0.0001).  The coeffi-

cient of blood lead for the low group was 0.178 +_ 0.04 compared

to 0.175 + 0.018 for the regression with all blood-lead levels,

a trivial difference between the two cases.  This indicated that

the risk of seriously elevated FEP levels was strongly related to

blood lead, even at blood lead levels below the previously defined

"safety level."

     Piomelli and coworkers1 studies have suggested a threshold

for lead-induced increases in FEP levels of about 15 ug/dl blood

lead.  Taking 17.5 ug/dl of blood lead as our reference level,
* The 50 ug/dl FEP level is considered by CDC to indicate
  severe enough interference with heme processes that med-
  ical attention is usually required, even when not coupled
  with elevated blood-lead levels.

-------
                               IV-20
                            Figure IV-5

        The Relationship  Between PEP Level and Blood  Lead
       After Adjusting  for  All  Other Significant Variables
  k)Q(FEP)
   (Mg/
   3.8
   3.6
   3.4
   3.0
                          -I	L.
       15
20
25
   30

LEAD (pg/dl)
35
40
45
The regression  line  shows  the true regression line  for  all I
individual observations.   For ease of display, each point repre-
sents the mean  of  about  100  children with consecutive  blood lead
levels.

-------
                               IV-21






our regression predicted that the relative risk of FEP levels over




50 ug/dl was 1.55 (1.42-1.70 at 95 percent confidence levels) at



20 ug/dl of blood lead, and was 3.73 (2.55-4.89 at 95 percent



confidence) at 25 ug/dl of blood lead.  This was true across all



transferrin saturation levels.






IV.A.2.C.3.  The Relationship Between FEP Levels and Anemia



     To assess the implications of elevated FEP levels, we analyzed



the relationship between log(FEP) and hematocrit, hemoglobin, and



MCV.  We performed linear regressions on all three outcomes as a



function of log(FEP), controlling for race, age, and transferrin



saturation.  These analyses showed that log(FEP) was strongly



inversely related (p < 0.0001 in all cases) to hematocrit levels,



hemoglobin levels, and MCV.  We then performed logistic regres-



sions on the probability of abnormal levels of hematocrit, hemo-



globin, and MCV as a function of log(FEP), with the same controls.



They also showed that FEP was an excellent predictor (p < 0.0001)



of the probability of abnormally low levels of all three indica-



tors.  Again, we repeated our regressions using only children with



FEP values below 33 ug/dl, and FEP was still very significant



(p < 0.0001).  (FEP levels of less than 33 ug/dl are generally



associated with blood lead levels under 30 ug/dl.)  The coeffi-



cients differed by less than one standard deviation from those for



the full sample.  Thus, the relationship appeared to hold for low




FEP levels as well as high ones.



     Figure IV-6 shows the regression's prediction of the percent




of children with anemia as a function of FEP levels at normal



transferrin saturation levels for children.  The data used in the

-------
                              IV-22






regression contained FEP levels as low as 9.6 ug/dl, but we have




shown the projections only for 18 ug/dl and above.  For our



definition of anemia, we have used the hemoglobin and hematocrit



levels representing the minimum normal range levels recommended




by the Journal of Pediatrics (1977).  These definitions are



supported by the work of Yip et al. (1981), and the use of hema-



tocrit and hemoglobin levels to define anemia is standard



(Harrison, Principles of Internal Medicine, 9th Edition).



     Figure IV-6 shows that as FEP levels increase from 20 ug/dl



to 50 ug/dl, an additional 20 percent of children aged 2-6 years



would develop anemia at normal iron levels.  Our earlier regres-



sions of blood lead levels on FEP suggested that blood lead levels



of less than 15 ug/dl were necessary to keep average FEP levels



below 20 ud/dl.  Since elevated FEP is a symptom of an underlying



interference in heme synthesis, it cannot be viewed as the cause



of these abnormal hematocrits.  The causal association must be



with whatever produced the excess FEP.  As the data portrayed in



Figure IV-3 were for normal iron levels, the anemia appeared to



be the outcome of the lead exposure, for which FEP served as a



surrogate.



     In summary, blood lead levels below 30 ug/dl seem to be



associated with increased microcytosis and hypochromia in children



and increased interference with heme synthesis producing elevated



levels of free erythrocyte protoporphyrin.  Elevated FEP levels,



even below 33 ug/dl, are themselves associated with an increased



risk of anemia.  In addition, the reduced mean cell volumes and



the lower hematocrit values again indicate that lead's effect on

-------
                      IV-2 3

                   FIGURE IV-6

   Prediction of Percent of Children with Anemia as a Function
     of FEP Level at Normal Transferrin Saturation Levels
    PERCENT OF CHILDREN WITH ANEMIA
 By Age and Race at Average Transferrin Saturation Levels
12-5
                      FEP Level (/ig/dl)

-------
                              IV-24


heme synthesis continues at levels below 30 ug/dl of blood lead.

This further strengthens the case for considering elevated FEP

levels themselves, which mark lead's interference with normal

body activity, as a pathophysiological effect.


IV.A.3.  Lead's Interference with Vitamin D Metabolism and
         Associated Physiological Processes

     Another potentially serious consequence of lead exposure is

the impairment of the biosynthesis of the active vitamin D meta-

bolite, 1,25-(OH)2 vitamin D, which is detectable at blood lead

levels of 10-15 ug/dl.  Further, an inverse dose-response rela-

tionship has been reported between blood lead and 1,25-(OH)2

vitamin D throughout the range of measured blood lead values up

to 120 ug/dl (Criteria Document, p. 12-49; Rosen et al., 1980a,

1980b; Mahaffey et al., 1982).  Interference with vitamin D pro-

duction disrupts calcium, zinc, and phosphorous homeostasis,

partially resulting in the reduced absorption of these elements

from the gastro-intestinal tract.  This alters the availability

of these elements for physiological processes crucial to the

normal functioning of many tissues, cell membranes, and  organ

systems.

     The reduced uptake and utilization of calcium has two

compounding conseguences.  First, it interferes with calcium-

dependent processes that are essential to the functioning of

nerve cells, endocrine cells, muscle cells (including those in

the heart and other components of the cardiovascular system),

bone cells, and most other types of cells.  The second concern

is possible increased lead absorption resulting from decreased

-------
                               IV-2 5





calcium availability.  The latter can be expected to create a



feedback response, further exacerbating the inhibition of vitamin



D metabolism and reduced calcium availability, leading to even



greater lead absorption and greater vulnerability to increasingly



more severe lead-induced health effects (Rosen et al., 1980b;



Barton et al., 1978).  These effects are especially dangerous



for young (preschool age) children who are developing rapidly.



These children, even in the absence of lead, generally are



susceptible to calcium deficiencies because of the large amount



of calcium used for the formation of the skeletal system, as



well as several other calcium-dependent physiological processes



important in young children.



     Even moderate levels of lead exposure in children are



associated with vitamin D disturbances that parallel certain meta-



bolic disorders and other disease states,  as well as severe kidney



dysfunction (Criteria Document, p. 12-37).  At blood lead levels



of 33-55 ug/dl, 1,25-(OH)2 vitamin D is reduced to levels compar-



able to those observed in children who have severe renal insuffi-



ciency with the loss of about two-thirds of their normal kidney



function (Rosen et al., 1980a; Rosen and Chesney, 1983; Chesney



et al., 1983).  Analogous vitamin D hormone depressions are found



in vitamin D-dependent rickets (type I), oxalosis, hormone-defi-



cient hypoparathyroidism, and aluminum intoxication in children



undergoing total parenteral nutrition.



     Lead-induced interference of 1,25-(OH)2D biosynthesis



affects a wide range of physiological processes.  The vitamin D-



endocrine system is responsible in large part for the maintenance

-------
                               IV-2 6






of extra- and intra-cellular calcium homeostasis (Rasmussen and



Waisman, 1983; Wong, 1983; Shlossman et al.,  1982;  Rosen and



Chesney, 1983).  Thus, modulation in cellular calcium metabolism



induced by lead at relatively low concentrations may potentially



disturb multiple functions of different tissues that depend upon



calcium as a second messenger (Criteria Document, p. 12-40).  It



also appears that 1,25-(OH)2D participates directly in bone



turnover by orchestrating the population of  cells within the



bone (Criteria Document, p. 12-41).  An immunoregulatory role



for the vitamin D hormone is evident through the widespread exist-



ence of 1,25-(OH)2^3 receptor sites on immunoregulatory cells,



such as monocytes and activated lymphocytes  (Provvedini et al. ,



1983; Bhalla et al., 1983).



     The negative correlation between blood  lead and serum



1,25-(OH)2D, the active form of vitamin D, appears to be another



example of lead's disruption of mitochondrial activity at low



concentrations.  While serum levels of 1,25-(OH)2 vitamin D



decreased continuously as blood lead levels  increased from an



apparent threshold of 10 to 15 ug/dl, this was not true for its



precursor, 25-(OH) vitamin D.  In fact, in lead-intoxicated



children after chelation therapy, vitamin D  levels were restored,



but the precursor levels remained unchanged  (Rosen et al., 1980a,



1980b; Mahaffey et al., 1982).  This indicates that lead inhibits



renal 1-hydroxylase, the kidney enzyme that  converts the precursor



to the active form of vitamin D.  Renal 1-hydroxylase is a




mitochondrial enzyme system, which is mediated by the hemoprotein,



cytochrome P-450.  This suggests that the damage to the mitochon-

-------
                               IV-2 7






drial systems detected at 15 ug/dl has uncompensated consequences.



     If cytochrome P-450 is being inhibited at the low levels of



blood lead that the reduced renal 1-hydroxylase activity suggests,




we must consider the possibility that other physiological func-



tions related to cytochrome P-450 may also be disrupted.  For



example, reduced P-450 content has been correlated with impaired



activity of the liver detoxifying enzymes, aniline hydroxylase



and aminopyrine demethylase, which help to detoxify medications,



hormones, and other chemicals (Goldberg et al., 1978).



     While cytochrome P-450 inhibition has been found in animals,



and in humans at higher lead levels, this has not yet been examined



in children at low blood-lead levels (i.e., 10 to 15 ug/dl).  But



the disruption of vitamin D biosynthetic pathways at these levels



is suggestive of an effect.



     The reduction in heme caused by lead exposure probably



underlies the effects seen in vitamin D metabolism.  This would



explain the similarity in apparent "thresholds" for the effect of



lead on both erythrocyte protoporphyrin accumulation and decreases



in levels of serum 1,25-(OH)2D-  It would also indicate a cascade



of biological effects among many organ and physiological systems



of the body (depicted graphically in Figure IV-1).  Together, the



interrelationships of calcium and lead metabolism, lead's effects



on 1,25-(OH)2D, and the apparent disruption of the cytochrome



P-450 enzyme system provide a single molecular and mechanistic




basis for Aub et al.'s observation in 1926 that "lead follows




the calcium stream."

-------
                               IV-28






IV.B.  Neurotoxic Effects of Lead Exposure



     Lead has been known to be a neurotoxicant since the early



1800s, and neurotoxicity is among the more severe consequences of



lead exposure.  At very high-blood lead levels,  encephalopathy



and severe neurotoxic effects are well documented; the neurotoxic



effects at lower blood lead levels,  however,  are less clearly



defined.  Recent research has investigated the occurrence of



overt signs and symptoms of neurotoxicity and the manifestation



of more subtle indications of altered neurological functions in



individuals who do not show obvious  signs of  lead poisoning.






IV.B.I.  Neurotoxicity at Elevated Blood-Lead Levels



     Very high blood-lead levels (i.e., above 80 ug/dl in children)



are associated with massive neurotoxic effects that can include



severe, irreversible brain damage, ataxia (i.e., the inability to



coordinate voluntary muscular movements), persistent vomiting,



lethargy, stupor, convulsions, coma, and sometimes death.  Once



encephalopathy occurs, the risk of death for  children is signifi-



cant  (Ennis and Harrison, 1950; Agerty, 1952; Lewis et al. , 1955),



regardless of the quality of the medical treatment they receive.



      In cases of severe or prolonged nonfatal episodes of lead



encephalopathy, neurological damage  occurs that is qualitatively



similar to that often seen following traumatic or infectious



cerebral injury, with permanent and  irreversible damage being



more  common in children than adults  (Mellins  and Jenkins, 1955;



Chisolm, 1956, 1968).  The most severe effects are cortical



atrophy, hydrocephalus (an abnormal  increase  in cranial fluid),



convulsive seizures, and severe mental retardation.  Permanent

-------
                              IV-2 9





central nervous system damage almost always occurs in children



who survive acute lead encephalopathy and are re-exposed to



lead (Chisolm and Harrison, 1956) .  Even if their blood lead



levels are kept fairly low, 25-50 percent show severe permanent



sequelae including seizures, nervous disorders, blindness, and



hemiparesis (paralysis of half of the body) (Chisolm and Barltrop,



1979) .



     Even children without obvious signs of acute lead encephalo-



pathy have exhibited persisting neurological damage.  As early



as 1943, Byers and Lord's study of 20 previously lead-poisoned



children indicated that 19 later performed unsatisfactorily in



school, "presumably due to sensorimotor deficits, short attention



span, and behavioral disorders".  Effects such as mental retarda-



tion, seizures, cerebral palsy, optic atrophy, sensorimotor



deficits, visual-perceptual problems, and behavior disorders have



been documented extensively in children following overt lead



intoxication or even just known high exposures to lead (e.g.,



Chisolm and Harrison, 1956: Cohen and Ahrens, 1959; Perlstein



and Attala, 1966) .



      The extent of the later manifestations seems to relate to



the severity of the earlier observed symptoms.  In Perlstein and



Attala, 9 percent of the children studied, none of whom appeared



to have severe symptoms when diagnosed for overt lead poisoning,



were later observed to be minimally mentally retarded and 37



percent showed some lasting neurological sequelae.



     At somewhat lower blood-lead levels (i.e., 30-70 ug/dl),



substantial data confirm that a variety of neural dysfunctions

-------
                              IV-30






occur in apparently asymptomatic children.  Several studies



indicate that blood lead levels of 50-70 ug/dl are associated



with IQ decrements of 5 points.  Adverse electrophysiological




effects, including markedly abnormal EEC patterns, slow-wave



voltages, etc., are also well documented at levels of 30-70



ug/dl.



     De la Burde and Choate (1972, 1975) showed persisting



neurobehavioral deficits in children exposed to moderate-to-high



levels of lead; most of the children appear to have had blood



lead levels above 40 ug/dl.  Compared to low-lead control



children — matched for age, sex, race, parents' socioeconomic



status, housing density, mother's 10, number of children in the



family below age 6, presence of father in the home, and mother



working -- the higher lead children averaged about five points



lower in 10 and were seven times more likely to have repeated



grades in school or to have been referred to school psychologists.



Moreover, follow-up studies showed that these effects persisted



for at least three years.



     While chelation therapy may mitigate some of these persisting



effects, permanent neurological and cognitive damage seems to



result from very high lead levels, with or without encephalopathy.



In addition, these children also appear more likely to experience



neurological and behavioral impairments later in their childhood.






IV.B.2.  Neurotoxicity at Lower Blood-Lead Levels



     The adverse effects of lead on neurological functioning,



both on the microscopic (i.e., cellular and enzymatic) level and



the macroscopic  (i.e., learning behavior) level, are well docu-

-------
                              IV-31






merited.  On the micro-level, data from experimental animal studies



suggest several possible mechanisms for the induction of neural



effects, including:  (1) increased accumulation of ALA in the



brain as a consequence of lead-induced impaired heme synthesis,



(2) altered ionic balances and movement of ions across axonal



membranes and at nerve terminals during the initiation or conduc-



tion of nerve impulses due to lead-induced effects on the meta-



bolism or synaptic utilization of calcium, and (3) lead-induced



effects on the metabolism or synaptic utilization of various



neurotransmitters.



     In addition, lead-induced heme synthesis impairment,



resulting in reduced cytochrome C levels in brain cells during



crucial developmental periods, has been clearly associated with



the delayed development of certain neuronal components and



systems in the brains of experimental animals (Holtzman and



Shen Hsu, 1976).  (Cytochrome C is a link in the mitochondrial



electron transport chain that produces energy, in the form of



adenosine triphosphate (ATP), for the entire cell.)  Given the



high energy demands of neurons, selective damage to the nervous



system seems plausible.



     In addition to the effects of lead on the brain and central




nervous system, there is evidence that peripheral nerves are



affected as well.  Silbergeld and Adler (1978) have noted lead-



induced blockage of neurotransmitter (acetylcholine) release in



peripheral nerves,  a result of lead's disruption of the transport




of calcium across cellular membranes.  This disruption of cellular



calcium transport may also contribute to the effects of lead on

-------
                              IV-3 2






peripheral nerve conduction velocity.  Landrigan et al. (1976)



have noted a significant correlation between blood lead and



decreasing nerve conduction velocity in children in a smelter



community.  This effect may indicate advancing peripheral



neuropathy.



     Paralleling these cellular or biochemical effects are



electrophysiological changes indicating the perturbation of peri-



pheral and central nervous system functioning observed in children



with blood lead levels of approximately 15 ug/dl.  These included



slowed nerve conduction velocities (Landrigan et al., 1976),



reaction-time and reaction-behavior deficits (Winneke et al.



1984; Yule, 1984), as well as persistent abnormal EEC patterns



including altered brain stem and auditory evoked potentials



down to 15 ug/dl (Benignus et al. , 1981; Otto et al., 1981, 1982,



1984).  The results indicating neurological effects of lead at



such low levels are particularly important because two- and



five-year follow-up studies (Otto et al., 1982, 1984) indicated



some persistent effects.



      Aberrant learning behavior has been noted in rats with



blood lead levels below 30 ug/dl.  This behavior evidenced both



reduced performance on complex learning problems and signs of



hyperactivity and excessive response to negative reinforcement



(Winneke, 1977, 1982a).



     Finally, the cognitive effects of lead in children show



signs of a dose-response relationship.  For high level lead



poisoning, adverse cognitive effects in children are indisputable



and mental retardation is a common outcome.  For children with

-------
                              IV-33





somewhat lower blood-lead levels, de la Burde and Choate (1972,



1975) found lesser but still significant cognitive effects,



including lower mean IQs and reduced attention spans.  Several



studies discussed in more detail in Section IV.B.2.a. have found



smaller effects at lower blood-lead levels.  The precise biolo-



gical mechanisms connected with these effects are not yet clearly



defined, although hypotheses have been put forward.



     While some of these effects have only been observed at



higher blood-lead levels, in animals, or in vitro, they show a



consistent dose-dependent interference with normal neurological



functioning.  Furthermore, some of these effects have been docu-



mented to occur at low blood-lead levels in children, with no



clear threshold having been demonstrated.



     This general pattern of lead's interference in neurological



functioning on the cellular level, including effects below 30



ug/dl, form the background against which we examined the studies



that investigated changes in cognitive processes in children



at low blood-lead levels.  Because of the intrinsic difficulties



in performing such studies, and because most investigators have



not employed sample sizes that would permit unambiguous detection



of small effects, it is important to integrate these behavioral



studies with what has been discovered on the molecular and cellu-



lar levels.





IV.B.2.a.  Cognitive Effects of Moderate Blood-Lead Levels



     The literature on cognitive effects at low to moderate body-



lead levels is extensive.  However, most of the studies have



some methodological flaws, and few display indisputable results

-------
                              IV-3 4






concerning the relationship between IQ effects and changes in



low body-lead levels.  The Draft Lead Criteria Document (p. 12-65)



divided the studies into four groups:  clinical studies of high



lead children, general population studies, lead smelter area



studies, and studies of children who are mentally or behaviorally




abnormal.



     One of the larger and better designed studies of lower-level



cognitive effects was the Needleman et al. study in 1979, which



found a significant inverse correlation between tooth lead levels



and IQ after controlling for age, parent's 10, and socioeconomic



factors.  In 1983 an EPA peer-review panel asked for a reanalysis



of Needleman et al.'s data, using different model specifications.



EPA's Office of Policy Analysis also requested a reanalysis us-



ing a continuous lead exposure variable.  Needleman submitted a



reanalysis to EPA, and presented the reanalysis to the Clean Air



Science Advisory Committee (CARAC) in 1984.  The CASAC stated



that the reanalysis adequately addressed the concerns of the peer



review panel, and recommended that EPA include the study in its



criteria document process.  These latter results confirmed the



association between lead and 10 found in the original study,



and showed that the results were quite stable in response to



the inclusion or exclusion of different confounding factors.



     The summary table in Chapter 12 of the Criteria Document



(pp. 12-65 to 12-70) indicated that virtually all of the general



population studies showed high lead groups performing more



poorly on a variety of tests used to assess cognitive function.



For more than half of these tests, however, the probability of

-------
                              IV-3 5


falsely finding an effect due to chance was more than 5 percent;

i.e., less than half of them had a p-value of less than 0.05.

(Significance levels in the studies were reported as probabilities

if they were below 0.05 and as "not significant" otherwise.)

     For use in public policy making, rejecting the results of

these studies because so many fail to attain significance at the

5 percent level may be inappropriate for two reasons.  First,

policy makers need to be concerned about both type I and type II

errors.  Significance tests guard only against the first type

(falsely rejecting the null hypothesis of no effect); they help

ensure that a regulation is not imposed when there is no adverse

effect.  Type II errors (failing to reject the null hypothesis

when it is false) also can be costly, however, because they can

result in the underregulation of a real hazard.  With small sample

sizes and subtle effects, the probability of a type II error can

be large; in the case of the Smith et al. (1983) study, we

calculated it to be 62 percent if the true decrease in IQ was

2 points.*  The probability of a type II error in the other

studies would be even higher, because of their smaller sample

sizes.
*  We computed the false positive from the Smith et al. data
   as follows.  Using Pocock and Ashby (1985), we derived the
   standard deviation for the difference of the high and low
   lead groups of 1.499.  At a 5 percent chance of rejecting the
   null hypothesis when it is true, the normal one-tailed test
   statistic is 1.65.  Therefore, we would reject the null
   hypothesis only for differences greater than (1.499) (1.65)
   = 2.473.  If the difference in the groups was two 10 points,
   the probability of the difference being below 2.473 is
   given by p (z < [2.47S-2]/ 1.499) = 0.62.

-------
                              IV-3 6


     The second reason for caution in rejecting the results of

these studies is that while several fail to attain statistical

significance individually, they do show a consistent pattern: in

nearly all of them, the children in the higher lead groups showed

lower mean IQs.  Figure IV-7 plots the estimated effects, along

with the 90 percent confidence limits.*  The higher end of the

90 percent confidence limit corresponds to the critical value for

a one-tailed test at the 0.05 significance level; i.e., studies

in the figure whose upper confidence limits exceed 0 are not

statistically significant at the p = 0.05 level.

     The consistent pattern in all of the studies suggested that

combining evidence from all studies would provide a better test

for a significant effect than separate evaluations of the statis-

tical significance of the individual studies.  In applying one

of the available joint tests for the existence of a specific

effect, we began with the six general population studies found

in Table 12.1 of the draft Criteria Document.  As the result of

personal communication with Harvey we have not included this

study, due to the younger age of the children (2.5) and the

continuous nature of their study design.  We did not consider

clinic studies because of their higher lead  (typically > 70 ug/dl)

or studies of children exhibiting abnormal behavior.  To the five

remining general population studies we added the smelter study

by Winneke et al.  (1982b), as the blood lead levels in that study
* Some of the studies did not report p-values or standard errors,
  in which cases, we used the data in the study to compute these
  values.  Where we had insufficient information, we did not
  include the study.

-------
FIGURE IV-7.
Mean IQ Difference Between High Lead Groups and Controls,
Adjusted for Socioeconomic Factors (90% Confidence Intervals)
•*•
2 -
5 o
o
wX
-16 -
1 R
f
1










,



.

•






.






.



1




.
•4-

«
' \




1 O 1 1 1 I 1
^Bride Yule Smith Yule and Winneke Needlonan
et al. et al. et al. Lansdown et al. et al.
(1982) (1981) (1983) (1983) (1982) (1979)

-------
                              IV-3 8






were in the same range as the general population studies.  For



three other studies, Winneke (1983), Winneke (1984), and Yule and



Lansdown's study in Leeds, sufficient data were not available to



include their results.  These three studies generally found small




or statistically insignificant effects.  We also looked only at



Full Scale IQ measures.  While not all studies used the same 10



test, the Full Scale IQ measures employed were close enough to



allow us to compare differences between groups and across



studies.  Table IV-3 summarizes the relevant information.



     In evaluating the reported results of the combined



significance of these studies, it is important to remember that



because each study was performed using different protocols,



study populations, levels of exposure and investigations, the



resulting combined p-value should be interpreted cautiously,



and viewed more for its qualitative implications than for its



precise numerical result.  The specific numerical result (joint



p-value) also is sensitive to the studies included.  Inclusion



of the studies by Winneke and others discussed above, which were



omitted because of insufficient data, would change the p-value.



The direction of the change is difficult to predict, however,



because while these studies generally found insignicant results,



the inclusion of additional studies will raise the joint p-value



unless the p-values in the individual studies are quite large



(e.g., p > 0.25) .




     We used the Fisher aggregation procedure (Fisher, 1970,



p. 99) to estimate the combined significance of the observed

-------
TABLE IV-3.  Computation of Joint P-Value from Epidemiological Studies of Cognitive Effects
             from Low Level Lead Exposure in Children
Study
McBride et al.
(1982)
/Yule et al.
(1981)
Smith et al.
(1983)
Yule and Lansdown
(1983)
Winneke et al.
(1982a)
Needleman et al.
(1979)

Joint p-value

Internal Lead Levels
Sample Sizes Blood (ug/dl) Teeth(ppm)
Control Exposed Control Exposed Control Exposed
86 86 < 9 19-30 	 	

20 21 7-10 17-32 	 	

1AQ 1 Rl^ / 1 C N Q
IflD 1DD — — — — — x Z.D > o

80 82 7-12 13-24 	 	

26 26 	 	 2.4 7

100 58 < 10 > 20


for studies: P(X2? > 31.31) <0.005

10 Difference P-Value -2 In p
1.2a 0.25 2.77

7.6b 0.027 7.22

2.3b 0.067C 5.41

1.8*> 0.13 4.08

5b 0.10 4.82

4.5b 0.03 7.01

31.31

       a  peabody Picture Vocabulary IQ lest
       b  Welchsler Intelligence Scale for Children-Revised
       c  Smith does not report a p-value but Pocock and Ashby report a 95% confidence interval of .4 to -5.5
          for the Smith result which implies a standard error of 1.5 and a t-statistic of 1.53.
                                                                                                                       LO
                                                                                                                       10

-------
                              IV-40


effects, and to derive a joint p-value for all of the studies.

To do this, we needed the p-values for all of the individual

studies.

     For each study where p-values were not reported, we used

the standard deviation of the IQ measure to compute the p-value

for the difference in the mean IQs across groups.  We could not

use this method for the 1983 study by Smith et al.  In that

study, the full scale IQ effects were reported as "not significant1

and no standard deviation was given.  However, when we computed

the p-value using the standard deviation derived from Pocock and

Ashby, we found that the p-value was 0.067 when comparing high

and low lead groups for the Full Scale IQ.*

     The results of our application of the Fisher procedure for

computing a joint probability for the observed results are pre-

sented in Table IV-3.  The resulting probability of less than

0.005 indicates that it was extremely unlikely that we could

get the observed pattern of results if there were really no

effect.  The overwhelming preponderance of the data (virtually

all studies show high lead groups with lower cognitive ability)

was highly unlikely to have been due to chance.

     Only  if the studies were consistently biased towards finding

an effect would the robustness of our result be questionable.  In
 * Pocock and Ashby reported that in the Smith et al. study
   the 95% confidence interval for the full scale IQ effect
   was -5.5 to +0.4.  The t-statistic for 298 degrees of freedom
   is 1.9679, yielding a standard deviation of 1.499.  Our
   calculation used the difference of 2.3 reported by Smith
   et al., rather than the difference of 2.55 implied by the
   recent work by Pocock and Ashby.

-------
                              IV-41





at least one case (Smith), a procedure was used that biased against



finding an effect, and biased upward the p-values.  These authors



used a two-stage analysis of variance or covariance where the



effects of all covariates (except lead) on IQ were controlled for



in the first stage, and the only residual IQ effects were regressed



on lead in the next step.  Many of these covariates (e.g., parental



care, income, and IQ) negatively correlate with lead exposure, and



this procedure attributed all of the joint variation to the non-



lead variable.



     These facts, together with the very small p-value calculated



in the joint test, suggest that the combined evidence of cognitive



effects at moderate levels of lead exposure should be treated as



statistically significant.  We conclude that the combined results



of available studies of cognitive effect at moderate lead levels



should be taken as evidence of cognitive decrements due to lead.





IV.B.3.  The Magnitude of Lead's Impact on IQ



     The evidence described above indicates that exposure to lead



can lower children's IQs and reduce their ability to perform well



in school.  In Section IV.D, we monetize the cognitive benefits



of reducing these effects using the costs of compensatory educa-



tion.  Here, we briefly describe a more direct, but also more



speculative, approach based on the improvements in IQ that might



be expected with reduced exposure to lead in gasoline.



     The latest draft of the Criteria Document characterizes the



evidence as suggesting that, on average, blood lead levels of 30



to 50 ug/dl result in a four-point decrement in IQ, and that lead

-------
                              IV-42






levels of 50 to 70 ug/dl reduce 10 by roughly five points



(de la Burde and Choate, 1972,  1975;  Rummo et al., 1979).  If we



assume that preventing a blood-lead level over 30 ug/dl avoids,



on average, the loss of four 10 points per child, the gain in



person-IO points from limiting  lead in gasoline is substantial.



In 1986, for example, as shown  in Table III-7, we estimate that



52,000 fewer children will experience blood lead levels over 30



ug/dl as a result of the 0.10 gplg limit (assuming no misfueling).



If we assume that each child over 30  ug/dl suffers roughly a four-



point 10 loss, that implies that the  final rule will yield a gain



of about 200,000 person-IQ points in  1986.  Table IV-4 presents



year-by-year estimates for the  alternative rules.



     This approach suffers from two faults, which cut in opposite



directions.  It does not account for  the fact that some children



who are prevented by the regulation from going over 30 ug/dl will



do so by a narrow margin (e.g., their blood lead level will be



29 ug/dl when it would have been 31 ug/dl in the absence of the



rule); such children are unlikely to  receive the full four point



gain in 10.  On the other hand, this  approach attributes no bene-



fit to children whose blood lead levels are reduced from very high



levels, but not brought below 30 ug/dl, or to those whose levels



would have been under 30 ug/dl  without the rule, but whose levels



decrease further by the reduction in  lead in gasoline.






IV.C.  Fetal Effects




     Because lead passes the placental barrier, a growing



concern in the public health community is that the most sensitive

-------
                                         IV-43


TABLE IV-4.  Year-by-Year Estimates of Gain in Person-IQ Points Under Alternative
	Rules, Assuming No Misfueling (thousands of person-IQ points)	


  Rule	1985    1986    1987    1988   1989   1990   1991   1992


    Proposed                0     208     188     172    156    144    128    124
    Alternative            88     184     180     172    156    144    128    124
    Final                  88     208     188     172    156    144    128    124

-------
                              IV-44






population for lead exposure is not children, but fetuses and



newborn infants.  This concern is supported by both animal



studies and, recently, human data in the published peer-reviewed




literature.



     Crofton et al. (1980) found that the development of



exploratory behavior by rat pups exposed to lead in utero lagged



behind that  of control rats.  Average blood lead levels on the



21st postnatal day were 14.5 ug/dl for the exposed pups and 4.8



ug/dl for the controls.



     Gross-Selbeck and Gross-Selbeck (1981) found alterations in



the operant  behavior of adult rats after prenatal exposure to



lead via mothers whose blood lead levels averaged 20.5 ug/dl.  At



the time of  testing (3 to 4 months, postnatal), the lead-exposed



subjects' blood lead levels averaged 4.55 ug/dl compared to 3.68



ug/dl in the controls.  This suggested that changes in central



nervous system function may persist for months after the cessation



of exposure  to relatively low blood-lead levels.



     Several other papers (McCauley and Bull, 1978; Bull et al.,



1979) have shown that the prenatal exposure of rats to 0.2 percent



lead chloride in the mother's drinking water markedly reduced the



cytochrome C content in the cerebral cortex, thereby possibly



producing an uncoupling of the electron transport chain in the



cortex.  This reduction in cytochrome C content occurred at blood



lead levels  as low as 36 ug/dl, with delays in the development of



central nervous system energy metabolism being seen as late as 50



days after birth (Bull et al., 1983).

-------
                              IV-45






     Needleman et al. (1984) analyzed data from over 4,000 live



births at Boston Women's Hospital and reported an association




between mild congenital anomalies and umbilical-cord blood-lead



malformation and lead, but only between all minor malformations



and lead.  There also were no significant associations between



lead and any major malformations, although given the rate of such



malformations in the general population, a sample this size has



little power to detect such an effect.  Holding other covariates



constant, the relative risk of a child demonstrating a minor



malformation at birth increased by 50 percent as lead levels



increased from 0.7 ug/dl to 6.3 ug/dl (the mean cord-lead level).



This risk increased an additional 50 percent at 24 ug/dl.



(Umbilical-cord blood-lead levels are somewhat lower than, but



correspond to, maternal blood-lead levels; Lauwerys et al., 1978.)



     A recent analysis by Bellinger and coworkers (1984) also



found an association between increasing cord-lead levels and



deficits in the child's subsequent performance on the Bayley



development scales, after controlling for covariates.  Again,



the mean cord-lead levels in this study were very low (under 10



ug/dl) .



     Finally, Erickson et al. (1983) found lung- and bone-lead




levels in children who died from Sudden Infant Death Syndrome



were significantly higher (p < 0.05) than in children who died




of other causes, after controlling for age.  While this study



suggests a potential relationship between lead and Sudden Infant



Death Syndrome, this issue remains to be more fully evaluated.

-------
                              IV-46






     In addition, lead has been implicated in complications of



pregnancy, including early and still births.   Fahim et al. (1976)



found that women who had normal full-term pregnancies had average



blood-lead levels of 14.3 ug/dl, whereas women with early membrane




rupture had average blood-lead levels of 25.6 ug/dl, and women



with premature delivery had average blood-lead levels of 29.1



ug/dl.  Wibberly et al. (1977) found that higher lead levels in



placental tissues were associated with various negative pregnancy



outcomes, including prematurity, birth malformation, and neonatal



death.



      Bryce-Smith et al. (1977) found bone lead concentrations



in still births of 0.4-24.2 ppm in the rib (average: 5.7) versus



typical infant bone lead levels of 0.2-0.6 ppm.



      To assess the effect of EPA's current rulemaking on fetal



exposure, we performed logistic regressions of the probability of



adults (over 15 years) from the NHANES II survey having blood lead



levels above 30 ug/dl.  The regression variables were selected



by a stepwise logistic regression procedure that chose all varia-



bles that were significant at the p = 0.05 level.  We then used



these regressions to predict the percent of women who would be



above 25 ug/dl in 1986-1992, under the final  rule.  We multiplied



the change in the percent of women of child-bearing age by the



expected total number of live births each year to estimate the



change in the number of fetuses born to mothers exposed to more



than 25 ug/dl of blood lead; this is shown on Table IV-5.  (The



regression coefficients are included in Appendix C.)

-------
                              IV-4 7
TABLE IV-5.  Estimated Decrease in the Number of Fetuses Exposed
	in Utero to > 25 ug/dl of Blood Lead	

  1985    1986    1987    1988    1989    1990    1991    1992

 1,275   3,800   3,200   2,800   2,550   2,200   2,000   1,900
IV.D.  Monetized Estimates of Children's Health Benefits

     The health benefits of reducing children's exposure to lead

are diverse and difficult to estimate quantitatively or to value

in monetary terms.  To monetize the benefits, we focused on two

admittedly incomplete measures:  savings in expenditures for

medical testing and treatment, and savings in compensatory educa-

tion.  These measures of benefit exclude many important factors,

such as reduced pain and suffering, or higher earnings in

later life.

     In fact, many children with elevated blood-lead levels are

neither detected nor treated.  However, our estimation procedure

assumes that children who go undetected and untreated bear a

burden at least as great as the cost of testing, treating, and

providing compensatory education for those who are detected.  So,

all children with high blood-lead levels are assumed to incur

"costs", whether medical expenditure costs or personal costs in

the form of poor health, inadequate learning, etc.


IV.D.I.  Reduced Medical Costs

     To estimate the benefits of reduced medical care expenses,

we assumed that children with elevated blood lead levels would

receive the treatment recommended by Drs. Piomelli, Rosen, Chisolm,

-------
                              IV-4 8





and Graef in a recent article in the Journal of Pediatrics (1984).




Those four leading experts in the clinical treatment of lead



toxicity combined their data and clinical experience to develop




optimal diagnosis, treatment, and follow-up protocols.  They also



estimated the percentages of children at different blood-lead



levels who would require various types of treatment.  Figure IV-7



summarizes the treatment options that we considered, based on




the recommendations of Piomelli et al.



     We assumed that administrative expenditures and follow-up



tests would cost $100 for every child found to be over 25 ug/dl



at screening.  Of those children over 25 ug/dl blood lead, we



estimated, based on Piomelli et al. (1982) and Mahaffey et al.



(1982), that 70 percent would be over 35 ug/dl FEP.  Piomelli



et al. (1984) recommend provocative ethyleneamentetraacetic acid



(EDTA) testing for such children.  EDTA testing typically requires



a day in the hospital and a physician's visit; we assumed a cost



of $500 per test.  We also assumed that all children receiving



EDTA testing would receive a series of follow-up tests and physi-



cians' visits, at a combined cost of $300.



     The purpose of EDTA testing is to see if children have a



dangerously high body-lead burden (a lead excretion ratio over



0.60, per Piomelli et al.).  Table IV-6 presents Piomelli et al.'s



estimates of the percentages of children at various blood lead




levels who will require chelation therapy; it ranges from a low of



zero for those under 30 ug/dl to a high of 100 percent for those



over 59 ug/dl.

-------
                                   IV-49
FIGURE IV-8
Flow Diagram for Children with Blood Lead Levels above 25 ug/dl
                Blood Lead Levels
                    >25 ug/dl
                   elevated
                     FEP?
               no
                            yes


simple
follow-up
no

hi<
leac

                             high body
                            lead burden?*
                                 I
                         yes
                                               1
                    long
                  follow-up
                           chelation
                            therapy
                                               I
                                            repeat
                                          chelation?
                                    no
                                        yes
                                                            1
                               long
                             follow-up
                                        chelation
                                         therapy
NOTES:

*Provocative EDTA or
 other test

tChelation therapy, because
 of its severe side-effects
 and inherent dangers, cannot
 be repeated again after this
 point
                                                            I
                                          repeat
                                        chelation?
                                no
                               1
 yes
                             long
                           follow-up
chelation
 therapy
                                                                         long
                                                                       follow-upt

-------
                              IV-50
TABLE IV-6.  Percent of Children Requiring Chelation Therapy

	Blood Lead Levels	Percent	
          25-30 ug/dl                              0

          30-39 ug/dl
             age three and over                   9.6
             age under three                     11.5

          40-49 ug/dl
             age three and over                  26.0
             age under three                     37.9

          50-59 ug/dl
             age three and over                  36.0
             age under three                     49.0

          above 59 ug/dl                        100.0

-------
                              IV-51





      Based on our analysis of NHANES II, we estimated that, of



those children over 25 ug/dl blood lead, about 20 percent are



between 30 and 40 ug/dl and 10 percent are over 40 ug/dl.  Based



on those estimates and the percentages in Table IV-6, we assumed



that 5 percent of the children above 25 ug/dl would require chela-



tion therapy.  In addition, we assumed that half of those children



chelated would require a second chelation due to a rebound in



their blood lead level, and that half of those children would



require a third chelation treatment.  Thus, we assumed a total of



0.0875 chelations would be required for every child over 25 ug/dl



blood lead at screening.  We assumed that chelation would require



five days in the hospital, several physicians' visits, laboratory



work, and a neuropsychological evaluation, for a total cost of



about $2,500 per chelation.



     Multiplying each of these costs by its associated probability



and then summing them yields the estimated cost per child found



over 25 ug/dl at screening:  1.0(100) + 0.7(500) + 0.7(300) +



0.0875(2500) = $878.75, which we round to $900.  This is lower



than the amount cited in a memo to the docket describing this



new methodology (August 16, 1984).  The reason for this difference



is that, in the text of their article, Piomelli et al. recommend



chelation for all children over 50 ug/dl; our earlier memo assumed



that treatment in estimating costs.  However, their data actually



indicate that some children in that range may not require chela-



tion, and, in this document, we have made the more conservative



assumption that not all children over 50 ug/dl will receive  it.

-------
                              IV-5 2





     Because we have not included welfare losses (such as work



time lost by parents), the adverse health effects of chelation



therapy itself (such as the removal of necessary minerals and



potential severe kidney damage), or such non-quantifiables as the



pain from the treatment, we believe our estimate of the benefits



is conservative.  As mentioned previously, these medical costs



are a measure of avoidable damage for all the incremental cases



of lead toxicity, whether detected or not.





IV.D.2.  Reduced Costs of Compensatory Education



     As discussed earlier, several studies show that moderate-to-



high exposures to lead reduce cognitive ability, as measured by



IQ tests.  The studies by de la Burde and Choate (1972, 1975) also



indicate that these cognitive effects, together with lead-induced



behavioral problems, translate into poorer performance in school;



they found that children in their high lead group were seven



times more likely than similar children with lower lead levels



to repeat a grade or be referred for psychological counseling.



Supplementary educational programs may compensate for some of



these effects, though certainly not all of them.



     To estimate roughly the cost of such compensatory education,



we relied on data in a study prepared for the Department of



Education's Office of Special Education Program.  Kakalik et al.



(1981) estimate that part-time special education for children who



remained in regular classrooms cost $3,064 extra per child per year



in 1978; adjusting for changes in the GNP price deflator yields



an estimate of $4,290 in 1983 dollars.  This figure is quite

-------
                              IV-5 3





close to Provenzano's (1980) estimate of the special education



costs for non-retarded, lead-exposed children.



     Based on de la Burde and Choate's finding that cognitive



effects persist for at least three years, we assumed that each



child needing compensatory education would require it for three



years.  De la Burde and Choate's high-lead group consisted mostly



of children over 40 ug/dl blood lead, who make up about 10 percent



of all children over 25 ug/dl.  A few, however, were in the range



of 30-40 ug/dl, and other studies (as discussed above in Section



IV.B) have found cognitive effects at levels well below 40 ug/dl.



Thus, we assumed that 20 percent of all children over 25 ug/dl



are affected severely enough that compensatory education would



be appropriate.  Thus, our estimated average cost per child over



25 ug/dl is (0.20)(3)(4,290) = $2,574, which we round to $2,600.





IV.D.3.  Summary of Estimated Benefits



     Adding our estimates of compensatory education and medical



costs yields a combined benefit estimate of $3,500 per case avoid-



ed of a child's blood-lead level exceeding 25 ug/dl.  Although for



convenience we have computed the average benefit per child over



25 ug/dl, it is important to note that most of the monetized bene-



fits are attributable to reducing lead in children who would be at



much higher levels (multiplied by the fraction of children over



25 ug/dl who are at those higher levels).  It  is also critical to



reiterate that our estimates are incomplete, omitting many



important categories, and thus are likely to be significant under-



estimates of the benefits of reducing lead in gasoline.

-------
                              IV-5 4






     Table IV-7 presents year-by-year estimates of the monetized



children's health benefits of the alternative rules.  They are



simply the estimated reductions in the number of children above



25 ug/dl (from Table III-6)  multiplied by $3,500 per case.  As



before, they assume that all misfueling is eliminated in each year.



The sensitivities of the results to alternative assumptions about



misfueling are explored in Chapter VIII.

-------
                                  IV-55


TABLE IV-7.  Year-by-Year Monetized Benefits of Reducing Children's
             Exposure to Lead Under Alternative Rules,
	Assuming No Misfueling (millions of 1983 dollars)	

Category
  Rule	1985  1986  1987  1988  1989  1990  1991  1992

Medical Care
    Proposed            0    155   141    130   117   107    95    93
    Alternative         65    139   134    130   117   107    95    93
    Final               65    155   141    130   117   107    95    93

Compensatory Education
    Proposed            0    447   408    374   338   309   276   268
    Alternative       187    400   387    374   338   309   276   268
    Final             187    447   408    374   338   309   276   268

Total
    Proposed            0    602   550    504   455   417   371   361
    Alternative       252    539   522    504   455   417   371   361
      1987  1988   1989  1990 1991  1992

-------
                            CHAPTER V


                HEALTH BENEFITS OF REDUCING LEAD:
            ADULT ILLNESSES RELATED TO BLOOD PRESSURE


     Concerns about the health effects of ambient exposure to

lead traditionally have focused on children.  Although lead has a

variety of adverse effects on the health of adults, most of them

appear not to be of substantial concern except at very high blood-

lead levels.  Recently, however, two new and extensive analyses

of the NHANES II data set have shown a strong and robust relation-

ship between blood lead levels and blood pressure.  That finding

has important implications for the benefits of reducing lead in

gasoline, because high blood pressure, in turn, is linked to a

variety of cardiovascular diseases.

     This chapter analyzes the health benefits for adults of

reducing lead in gasoline, but is limited in several ways.

First, we evaluated only illnesses related to blood pressure,

although lead has other adverse effects on adults.  Second, the

analysis is restricted to males aged 40 to 59, because lead appears

to affect blood pressure only in men, not women, and because the

best data are available for that age range.  Finally, most of

the estimates cover only whites, because the existing studies of

disease associated with blood pressure have had insufficiently

large samples of nonwhites.  For these reasons, the estimates

contained in this chapter are likely to understate significantly

the adult health benefits of reducing lead in gasoline.  The most

important omissions are older males and black males of all ages.

-------
                               V-2






     The estimates presented in this chapter should be treated as



preliminary.  They rely heavily on a recent paper by Pirkle et



al. (1985) that has been published in a peer-reviewed journal




(The American Journal of Epidemiology), but has not yet been



widely reviewed.  A summary of that paper and the calculations



underlying the estimates in this chapter was placed in the docket



for this rulemaking (Schwartz, "Blood Lead and Blood Pressure",



September 7, 1984).  Another recently published paper (Harlan et



al., 1985) also reports a statistically signifcant relationship



between blood pressure in the NHANES II data set.  Until the



broader scientific community has had a chance to review these



papers and comment on their findings, EPA will not rely on blood-



pressure-related benefits for this lead-in-gasoline rulemaking.



These health effects will be considered in the Agency's ongoing



deliberations on a ban, however, and are addressed here for



information purposes.






V.A.  The Relationship Between Blood Lead and Blood Pressure



     This section analyzes the statistical relationship between



blood lead and blood pressure.  The first part provides a brief



overview of earlier studies on the subject, while the second part



provides a detailed discussion of a recently completed statistical



analysis of the NHANES II data.






V.A.I.  Earlier Studies




     Lead has long been associated with effects on blood pressure



and the cardiovascular system, including a paper in the British

-------
                               V-3





Medical Journal by Lorimer in 1886 that found that higher blood-



lead levels increased the risk of hypertension.  Most of the



studies have focused only on hypertension and relatively high



lead-exposure levels, and have not looked for a continuous effect



of lead on blood pressure.  Investigators reporting such an



effect include Beevers et al. (1980), Morgan (1976), Richet et



al. (1966), and Dingwall-Fordyce and Lane (1963).  Others have



failed to find effects of lead on hypertension that were signifi-



cant at the 95 percent confidence level, although most of them



did find a positive association.  These include Ramirez-Cervantes



et al. (1978) and Fouts and Page (1942).



     More recently, Batuman et al. (1983) found an association



between chelatable body-lead levels and hypertension in veterans,



and several recent general population studies and lower lead-



exposure studies (Beevers et al., 1976; Kromhout and Coulande,



1984) have found a significant association with blood lead.



Moreau et al. (1982) also found a significant relationship



(p < 0.001) between blood lead levels and a continuous measure of



blood pressure in 431 French policemen, after controlling for



age, body mass index, smoking, and drinking.



     An even more recent British study  (Pocock et al., in press)



found blood lead significantly related  to blood pressure at the



99 percent confidence level, but the authors felt that the small



size of their correlation coefficient suggested no noticeable



effect.  However, that conclusion appears to reflect a misunder-



standing of statistics.  It is the regression coefficient that



indicates the size of an effect.  A correlation coefficient

-------
                               V-4






confounds that measurement with the variances of the dependent




and independent variables.  While their full data set was not



available to us, Pocock et al. presented their grouped data, and



we were able to perform a regression of blood pressure versus



the log of blood lead on their group averages, both before



and after adjustment for confounders.  The regressions were



weighted by the inverse of the variance of each group, and con-



firmed their finding that blood lead was a significant predictor



of blood pressure in their data, both before and after adjusting



for covariates.  Moreover, the regression coefficient indicated



that the size of the effect was significant, suggesting a change



of 3 mm Hg (millimeters of mercury, the standard measure of blood



pressure) as blood lead goes from 5 to 15 ug/dl.



     Weeden (1975) found lead associated with the vascular



renal changes linked to essential hypertension, indicating a



possible causal pathway.  Cooper and Gaffey (1974)  analyzed



mortality data from 1,267 death certificates for 7,032 lead



workers employed between 1900 and 1969, and found a significant



excess of deaths from hypertension disease and renal disease.  A



later analysis of similar data from 1971 to 1975 also found an



increase in cardiovascular and renal disease, but it was no



longer significant at the 95 percent confidence level (Cooper,



1981) .




     Animal data also link lead to hypertension.  Victery (1982)



found lead associated with a significant elevation of blood pres-



sure in rats with blood lead levels of 41 ug/dl.  Importantly,

-------
                               V-5





this study confirms Beevers et al.'s finding of a sex differen-



tial, with male but not female rats becoming hypertensive.  Webb



(1981) examined the vascular responsiveness of tail arteries in



rats exposed to blood lead levels in the 40 ug/dl range that had



suffered increases in systolic blood pressure, and found that the



arteries in exposed rats had increased contraction in response to



stimulation by neurotransmitters.



     lannaccone et al. (1981) also reported increased blood



pressure (p < 0.001) in rats at blood lead levels of 38.4 ug/dl,



as well as significant increases in the blood pressure response



to noradrenalin.  Perry and Erlanger (1979) found that low level



exposure of rats to lead produced increases of 15-20 mm Hg in



systolic blood pressure.  Kopp (1980) repeated those findings



and found electrocardiogram changes, indicating an effect on



the heart itself.  Subsequent tissue analysis of the heart showed



reduced levels of ATP in the heart muscle, indicating that heme



synthesis inhibition by lead was affecting energy availability



in the heart itself.



     The direct cardiological effects of lead are also indicated



by electrocardiogram changes in lead-poisoned children, which



are reversed by chelation therapy (Freeman, 1965; Silver and



Rodrigues-Torres, 1968).  Williams (1977, 1978, 1979) has



shown persistant increased susceptibility to norepinephrine-



induced arrhythmias in rats exposed to lead in the first



three weeks of life.





V.A.2.  Analysis of NHANES II Data



     In light of these indications of potential effects of lead

-------
                               V-6





on the cardiovascular system, the relationship between blood



lead levels and blood pressure has recently been explored (Harlan




et al., 1985; Pirkle et al., 1985) using the NHANES II data.



The NHANES II is an excellent data base for this analysis because



of the care given to accurate measurements, the great range of



information on possible confounding factors, and because it is a



representative sample of the U.S. population.  As such it avoids



the problems of selection bias, healthy-worker effect, other



occupational exposures, and the choice of controls that confound



many occupational studies.  Harlan et al. found blood lead related



to blood pressure for males aged 12 to 74 after controlling for



the traditional variables associated with blood pressure (age,



age-squared, body mass index, race) as well as alcohol consumption,



socio-economic factors, and all nutritional variables suspected



of affecting blood pressure.  Moreover, this relationship held



in each year of the NHANES II sample, when analyzed separately,



and this relationship held for both blacks and whites.



     Pirkle et al. found that blood lead levels were a statis-



tically significant predictor of blood pressure in adult males.



This relationship held not only when blood lead was evaluated



in a regression with all known factors that have previously been



established as correlated with blood pressure, but also when



that relationship subsequently was tested against 87 additional




variables representing linear and nonlinear functions of every



dietary and serologic variable in the NHANES II survey.



     Although finaal judgments about the casual relationship



between blood lead and blood pressure must await further review

-------
                               V-7





and study, EPA believes that the Pirkle et al. study provides



a reasonable basis for estimating the potential blood-pressure-



related benefits of reducing lead in gasoline.  Our analysis



builds on the Pirkle et al. study, and on additional work by



one of its authors (J. Schwartz).





V.A.2.a.  Blood Pressure Measurements



     Three blood pressure measurements were taken during NHANES



II.  A seated measurement was taken as soon as the examinee



entered.  Later, a recumbent measurement was taken.  A second



seated measurement was taken just before the end of the examina-



tion.  It is standard medical practice to prefer the second



seated measurement, because nervousness on just entering a



medical examination center makes the first seated measurement



less stable.  All of the results presented are for the second



seated measurement.  However, almost all of the regressions and



robustness tests described were performed on all three measure-



ments, and on the average of the first and third seated measure-



ments; all of the conclusions concerning lead's significance



held for all eight regressions (four diastolic, four systolic).





V.A.2.b. Initial Analysis



     After replicating the Harlan et al. results for all adult



males, the first goal was to determine if blood lead levels were



related significantly to blood pressure in white males, 40 to 59



years old.  This subgroup was chosen because at lower ages both



blood pressure and blood lead vary with age.  This collinearity



could artificially mask or enhance the correlation between blood

-------
                               V-8





lead and blood pressure.  Between 40 and 59 years of age, however,



blood pressure is essentially independent of age.  Choosing this



subgroup avoids any collinearity problems.  We focused on whites



because data relating cardiovascular disease to blood pressure



are less extensive for nonwhites.  The established correlates of




blood pressure are age, sex, race, and one of the indices of



relative height-to-weight.  Body mass index (BMI = weight/height2)



was used in this analysis.  By limiting attention to 40 to 59 year



old white males, there was no need to control for race, sex, or,



to a large degree, age.  Although age was only occasionally signi-



ficant in the stepwise analysis, both age and age-squared were



forced into each multiple regression model to be certain any



effect of lead was independent of age.



     The natural log of blood lead was more normally distributed,



was more statistically significant, and gave a higher R2 than



untransformed blood lead, blood-lead-squared, blood lead plus



blood-lead-squared, the square root of blood lead, or blood lead



to other fractional powers (0.15, 0.2, 0.3, 0.4).  All of the



results reported here are for the natural log of blood lead, but



regressions using lead on the untransformed scale gave very



similar results.




     The initial regressions analyzed systolic and diastolic



blood pressures for white males, 40 to 59 years old, with a model



consisting of age, age-squared, BMI, and blood lead.  These




regressions were done to determine whether blood lead levels were



significantly associated with systolic and diastolic blood pres-



sures after controlling for age, sex, race, and BMI, which are

-------
                               V-9





well-documented correlates of blood pressure.  Lead was statis-



tially significant (p < 0.01) for both systolic and diastolic



blood pressures in all the regressions (unweighted, weighted, and



weighted with design effects).  The regressions also tested



whether this relationship held up when other potentially confound-



ing variables were considered.





V.A.3.  Tests of Robustness



     The regression models were expanded to incorporate additional



variables, with particular attention directed to the stability



and significance of the lead coefficient in the presence of nutri-



tional factors and blood biochemistries.



     A large set of nutritional and biochemical variables from



NHANES II was included in the stepwise regressions.  Additional



regression analyses considered potential problems of interaction



terms.  To ensure the robustness of the relationships, further



analyses were done to address marginally insignificant variables



and nonnutrition variables.  Although our analysis focused on



males aged 40 to 59, additional regressions were also performed



considering all males over age 20.





V.A.S.a.  Nutritional and Biochemical Variables



     To provide an unusually rigorous test of the independent



significance of blood lead, almost all of the nutritional and bio-



chemical variables in the NHANES II were included in stepwise



regressions.  In addition, to account for possible curvilinear



relationships, squared and natural logarithmic transformations

-------
                               V-10






of almost all of these variables were also included.  The vari-



ables are listed in Table V-l.  The objective was not to evaluate



the possible association of nutritional or biochemical measure-



ments with blood pressure, but rather to conservatively estimate



the strength and independence of the relationship between blood




pressure and blood lead.



     Including these additional 87 variables increases the proba-



bility of variables being found statistically significant due to



chance alone.  This complicates the interpretation of nutritional



and biochemical factors, but not the interpretation of the lead




variable; it only makes it more difficult for lead to maintain



its significance.



     The general procedure for variable selection was as follows.



First, weighted stepwise multiple linear regression was used to



determine which variables were significantly related (p < 0.05)



to blood pressure (using the Stepwise and MAXR options of the SAS



procedure, STEPWISE).  The MAXR procedure was the principle one



used; it determines for any given model size (i.e., number of



variables) the variables that explain the greatest amount of the



variance (i.e., maximize R^).  We chose the largest model with



all variables significantly related to blood pressure (p < 0.05).



The Stepwise option, which uses forward selection with backwards



elimination, chose very similar models, and also always chose



blood lead.  From the 87 nutritional and biochemical variables,



the weighted stepwise regression selected five additional variables



for diastolic pressure and six additional variables for systolic



pressure using a 5 percent significance test.  These were used

-------
                               V-ll
TABLE V-l. Variables Included in the Stepwise Regression Analyses
age *
age-squared *

body mass index

dietary sodium t

salt shaker sodium

dietary sodium X
  salt shaker sodium


dietary potassium t

dietary sodium -
  potassium ratio


dietary calcium t

dietary phosphorus t

dietary protein t

dietary fat t

dietary carbohydrate t

dietary cholesterol t

dietary saturated

    fatty acids t

dietary oleic acid t

dietary linoleic acid t
dietary iron t
dietary vitamin A t

dietary vitamin C t

dietary thiamine t

dietary riboflavin t

dietary niacin t

serum cholesterol+

serum vitamin C t

serum iron t

serum transferrin saturation

serum zinc t

serum copper t

serum albumin t

hemoglobin t

red blood cell count

ethanol consumption / week t

cigarettes smoked / day

total dietary grams t

total dietary calories t

cigar or pipe smoking
* forced into each regression to remove any possible age
  effects on blood pressure.

t the natural log and squared transformation of these
  variables were also included in the stepwise regression,

-------
                               V-12






as the starting model for the SAS procedure SURREGR, which addi-




tionally incorporated the survey design effects.  For both



systolic and diastolic blood pressures, one variable from the



weighted stepwise regression failed to maintain significance at



the 5 percent level after the design effects were incorporated.



The final regression results for systolic and diastolic pressures,



after accounting for the weighting and design effects, are given




in Table V-2.



     The multiple logistic regressions (of the probability of



hypertension) were also performed using programs from SAS.  The



procedure LOGIST was used for the stepwise unweighted multiple



logistic regression, and the procedure NLIN (nonlinear regression)



was used for the weighted logistic regression calculations.  The



selection process again chose the largest significant model that



explained the greatest amount of the variance.  Calculations of



threshold levels for effects were made using the procedure NLIN



on segmented regression models, which finds the threshold point



that minimizes the sum of the squares of the error terms.  The



results of the logistic regression on hypertension are shown in



Table V-3.  Note that the logistic regressions included blacks



as well as whites, because these regressions were used only to



predict the effect of lead on the probability of having hyperten-



sion and were not used to estimate the number of cardiovascular



diseases and deaths.  As noted earlier, blacks were not included



in the linear regressions because the best available coefficient



for predicting cardiovascular risks included insufficient numbers



of blacks.

-------
                                     V-13
TABLE V-2.  Regression of Diastolic and Systolic Blood Pressures
            in White Males Aged 40 to 59
Variable
Diastolic
Age
Age2
Body Mass Index
Log (blood lead)
Dietary Potassium
Hemoglobin
Albumin
Log (dietary
vitamin C)
Systolic
Age
Age2
Body Mass Index
Log (blood lead)
Albumin
Log (dietary
Vitamin C)
Log (dietary
riboflavin)
Log (dietary
oleic acid)
Log (serum
vitamin C)
Coefficient

0.2768
-0.0014
1.131
3.954
-0.0018
1.548
3.587

1.838

1.311
-0.0068
1.736
8.436
7.088

2.411

-5.509

3.992

-3.472
t-Statistic

0.17
0.10
8.55
2.85
4.92
3.90
2.50

4.65

0.57
0.30
9.42
3.24
2.50

3.84

3.07

2.49

2.47
Probability

0.8636
0.9321
0.0001
0.0080
0.0001
0.0005
0.0179

0.0001

0.5720
0.7706
0.0001
0.0028
0.0178

0.0005

0.0044

0.0183

0.0184

-------
                               V-14
TABLE V-3.  Weighted Logistic Regression on Probability of Diastolic
            Blood Pressure Greater Than or Equal to 90 nun Hg in Men
            Aged 40 to 59
Variable
Constant
Log (Blood Lead)
Albumin
Body Mass Index
Hemoglobin
Log(Vitamin C)
Dietary Potassium
Total Carbohydrates
Coefficient
-16.41
0.693
0.0873
1.700
0.0329
0.3585
-0.00058
0.00246
t-statistic
10.13
3.96
3.70
9.34
5.25
5.98
7.47
3.09
p-Value
0.0000
0.0000
0.0001
0.0000
0.0000
0.0000
0.0000
0.0010

-------
                               V-15





     After including the nutritional variables, the blood analytes,



and their curvilinear transformations, lead remained significantly



associated (p < 0.01) with both systolic and diastolic blood



pressures.  The magnitude of this relationship, adjusted for the



other significant variables, is shown graphically in Figures



V-l and V-2.  Furthermore, segmented regression analyses indicated



there was no threshold blood lead level in the data.



     These segmented "hockey stick" regressions fit two regression



lines to the data.  One, below the putative blood lead threshold



T, depends on all the variables except lead.  The other, for blood



lead levels above T, includes lead.  An iterative technique is



used to find the value of T that minimizes the sum of the squares



of the error terms over the full range of both regression lines.



In this case, the error in the regression was minimized at a



threshold of zero; that is, lead was significantly related to



blood pressure at all levels down to zero.





V.A.S.b.  Interaction Terms



     In multiple regression analysis, another consideration is



the possibility of significant interaction terms.  To evaluate



this possibility, an additional weighted stepwise regression



analysis was done for systolic and diastolic blood pressures.



The variables consisted of the linear interaction terms between



the final variables in the model (shown in Table V-2) and the



linear form of all the other variables originally selected for



the initial stepwise regression, including their log and square



transforms (Table V-l).  This meant running a stepwise regression



with 162 interaction terms added to the final regression models

-------
                                                       V-16
      140
x

E
      135
CO
CO
111
(C
Q.

o
o
o
o
CO


(0

o
IU


CO
o
<
130
      125
     120
                                               FIGURE V-l


                               Ad j lasted Systolic Blood Pressure versus Blood Lead
                1    6                 10           14       18     22    26   30   34  38


                    ADJUSTED BLOOD LEAD LEVELS  (MICROGRAMS/DECILITER)

-------
o>
X
iu
oc
D
CO
CO
UJ
oc
o.

Q
O
o

CD

O
Zj
O

&
o
ui
h-
CO
    92
 I  90
88
86
84
82
80
<  78
                                      V-17



                                  FIGURE V-2


                         Adjusted Diastolic Blood Pressure versus Blood Lead
                    6
                                 10
14
18
22    26   30  34  38
                     ADJUSTED BLOOD LEAD LEVELS  (MICROGRAMS/DECILITER)

-------
                               V-18





for systolic and diastolic pressures.  Using such a large set



of variables gave a high probability that some variables would



enter at the 5 percent level by chance.  However, the purpose



was not to determine if those variables were independently



significant, but, rather, to further test the significance and



independence of the relationship between blood pressure and



blood lead.  As expected, several interaction variables entered



the systolic and diastolic regressions, but in each regression



the lead coefficient varied less than 10 percent and remained




significant (p < 0.015).






V.A.3.C.  Marginally Insignificant Variables



     Three other analyses were done to ensure that this



relationship was robust.  First, the original weighted stepwise



regression was extended to include variables significant through



the 15 percent level to see if marginally insignificant variables




influenced the significance of lead.  For both systolic and dia-



stolic pressures, lead remained significant and there was little



change in the magnitude of the coefficient.




     Second, for diastolic blood pressure, all the variables were



included that were significant between the p = 0.05 and the 0.15



levels, and every possible combination of those variables was




considered.  All 255 combinations were added to the variables that



were statistically significant, and a regression was performed



on each one.  The coefficient of the log of blood lead varied by



only plus or minus 10 percent from the value we obtained when we



included only significant variables, and the highest p-value for



lead was still less than 0.01.

-------
                               V-19





     The last analysis was the most demanding test of the indepen-



dence of the relationship between blood pressure and blood lead.



Models for diastolic and systolic blood pressures were fit by



weighted stepwise regression to the original model variables



(Table V-l), excluding lead.  This gave all of the other variables



and their curvilinear transformations the maximum opportunity to



explain variation that could also be explained by lead.  After



obtaining this new final model without lead, a single regression



was run adding the lead variable to the variables of this new



final model.  For both systolic and diastolic pressures, lead was



still statistically significant (p < 0.016) and the magnitude of



the lead coefficient changed less than 10 percent from those ob-



tained in the original analysis.  The results of all these analyses



indicated that the strength and independence of the relationship



between blood pressure and blood lead were remarkably stable.



     Because some people have found small amounts of ethanol



associated with reduced blood pressure, ethanol was also modeled



as a quadratic function of consumption, and with two dummy



variables for light and heavy drinking.  The stepwise regression



was repeated, with no change.





V.A.3.d.  Nonnutrition Variables



     Pirkle et al. then considered nonnutrition variables that



might be associated with blood pressure.  In additional runs



completed since then, we have added several other variables.



The complete set is shown in Table V-4.  Socio-economic and demo-



graphic factors as well as additional medical history variables



were included.

-------
                               V-20





TABLE V-4.  Nonnutrition Variables Tested in the Stepwise Regression






Demographic Variables              Other Personal-History Variables




Family Income                      Tricep Skinfold



Poverty Index                      Subscapular Skinfold



Region of the Country              Recreational Exercise



Season of the Year                 Work-Related Exercise



Degree of Urbanization             Recent Weight Loss



Residence Inside Central City      Family History of Hypertension



Educational Level                  Kidney Disease



                                   Serum Creatinine






Hypertension Variables



Hypertensive Medication



Low Salt Diet

-------
                               V-21





     Hypertension medication and low salt diet were tested —



not for inclusion in a final model, as they are essentially



indicators of high blood pressure, but rather to see if the



response to lead differed in those groups.  The coefficient of



lead did not change appreciably, and lead interaction terms with



the two variables were insignificant.



     The other variables in Table V-4 were tested in two ways.



First, the stepwise regression procedure was repeated with them



using all nutritional and serum measurements that were significant



at the p = 0.15 level.  The nutritional factors were limited to



those significant at the 0.15 level to give the nonnutritional



factors a greater chance to enter the model.  Again, lead was



selected (p < 0.005) and its coefficient changed by less than 10



percent from the original model that included only age, age2,



and body mass index.



     The variables in Table V-4 were then added to all of those



on Table V-l (including their nonlinear transforms) and the step-



wise process was repeated -- with the same results.  Finally,



the stepwise procedure was rerun using all of the variables in



Tables V-l and V-4 except lead; lead was then inserted into the



model resulting from this procedure.  It was still significant



(p < 0.006), with less than a 10 percent change in its coefficient.



Because the presence of two terms to describe the curvilinear



dependence of blood pressure on age might reduce the chances of



variable correlated with age achieving significance, age was modeled



as a single curvilinear function (sine of age), and the stepwise

-------
                               V-22






regression repeated; the results were the same.  In addition,



smoking and drinking were forced into the regression, and lead



was still significant (p < 0.01), with only a 3 percent change




in its coefficient.



     Our previous studies have shown that about half of the lead



in people in the NHANES II sample came from gasoline.  Tetraethyl



lead has very little cadmium in it, so confounding with cadmium



(which is also suspected of affecting blood pressure) is unlikely.



However, we repeated the regression excluding occupationally



exposed workers, who may also have cadmium exposure.  Lead



remained significant (p < 0.01), and its coefficient increased



somewhat.  We also regressed gasoline lead directly on blood



pressure, and it was significant.



      Although all of these analyses make it clear that



collinearity is not a problem in these regressions, variance



inflation factors were computed; no significant variable had a



variance inflation factor above 1.4.  (Variance inflation factors



below 4 are considered acceptable in multiple regression analyses.)



     To ensure that the significance of lead in the regression



was not due to the presence of a few influential observations,



influence diagnostic procedures were run.  Studentized residuals



were plotted for all the observations, and the largest residuals



were clustered near the middle of the data, where their influence



is slight.  Cook's D statistics also were computed for each obser-



vation.  The highest Cook's D was 0.029, and the second highest



was 0.023, both of which are very small.  Moreover, of the 10



observations with the largest Cook's D statistics, six had posi-




tive residuals and four had negative residuals, indicating that

-------
                               V-23





the most influential observations split almost evenly on which



way they would influence the lead regression coefficient.





V.A.S.e.  Other Age Groups



     The 40 to 59 year old age group represents about one-third



of adult males, and is the only one where the confounding of age



and blood lead can be eliminated unambiguously.  Additional



regressions were performed, however, to confirm the Harlan et al.



finding of an effect in all adult males.  Tables V-l and V-4



contain several variables that Harlan et al. did not consider in



their analysis.  Therefore, the stepwise regression analysis was



repeated using all of the variables in both tables, and their



square and natural log transforms as indicated.  All males over



the age of 20 were considered.  Lead was selected by the regres-



sion, with a p-value less than 0.01.  To check whether the rela-



tionship might be substantially different for different age



groups, dummy variables for each 10-year age group (between 20



and 70 years), and interaction terms between lead and those dummy



variables, were inserted in the stepwise regression.  Such inter-



action terms check for differences in the lead/blood pressure



relationship without having to subdivide the sample.  None of the



interaction terms was significant at even the p = 0.15 level.





V.A.4.  Summary of Blood Lead - Blood Pressure Results



     The final models for blood pressure, including all



statistically significant variables, are shown in Table V-5.



The final logistic model for the probability of hypertension is



shown in Table V-6.

-------
                                      V-24
TABLE V-5.  Regression of Diastolic and Systolic Blood Pressures
            in White Males Aged 40 to 59
Variable
Diastolic
Age
Age- squared
Body Mass Index
Blood leadt
Potassium
Hemoglobin
Albumin
Dietary Vitamin Ct
Family history of
hypertension
Recreational
exercise
Coefficient
-0.210
0.003
1.082
4.609
-0.002
0.151
0.354
1.886
2.085
-1.851

F-Statistic
0.02
0.04
67.88
12.19
25.30
16.81
7.42
23.67
4.37
9.48

Probability
0.8960
0.8373
0.0000
0.0014
0.0000
0.0003
0.0104
0.0000
0.0446
0.0042

Systolic
Age
Age-squared
Body Mass Index
Blood leadt
Albumin
Dietary Vitamin Ct
Dietary Riboflavint
Dietary Oleic Acidt
Serum Vitamin Ct
Family history of
hypertension
1.142
-0.005
1.710
8.510
0.695
2.458
-5.101
3.650
3.365
3.683

0.25
0.05
85.90
10.54
6.09
13.78
8.14
5.34
5.81
4.59

0.6226
0.8208
0.0000
0.0027
0.0192
0.0008
0.0075
0.0275
0.0218
0.0399

t log transform

-------
                                  V-25
TABLE V-6.  Logistic Regression on Probablity of Blood
            Pressure Greater Than or Equal to 90 ntn Hg in
            Men Aged 40 to 59
Variable
Constant
Log (Blood Lead)
Albumin
Body Mass Index
Hemoglobin
Log(Vitamin C)
Dietary Potassium
Total Carbohydrates
Recreational Exercise
Coefficient
-15.40
0.793
0.650
0.1571
0.0265
0.3593
-0.00053
0.00286
0.3864
t-Statistic
7.0
3.20
2.06
6.57
3.19
4.22
5.33
2.86
0.128
p-Value
0.0000
0.0014
0.0399
0.0000
0.0015
0.0000
0.0000
0.0080
0.0026

-------
                                V-26






     It is noteworthy that the logarithmic form of the dose-



response relationship suggests a large initial effect, leveling



off at higher blood-lead levels.  This may explain why only about



60 percent of the occupational studies (i.e., high lead-exposure




studies) have found an effect that was significant at the 95



percent confidence level, while almost all of the studies of



lower lead levels have found the relationship to be significant.




     The other low-exposure studies, the animal data, and the



robustness of these results suggest that the relationship is



causal.  Moreover, specific analyses to determine whether there



is a lower threshold below which lead has no effect on blood pres-



sure showed that the data were fit best with a threshold of zero.






V.B.  Benefits of Reduced Cardiovascular Disease



     Reducing lead in gasoline will reduce blood lead levels,



which in turn will reduce blood pressure and the number of indi-



viduals with hypertension.  The reduction in hypertension will



have some direct benefits from reduced medical treatment expendi-



tures.  More important, however, will be the indirect benefits



in the form of reduced cardiovascular disease associated with



elevated levels of blood pressure.



     This section describes the methods used to estimate the



benefits associated with lowering blood pressure.  The first part



deals with estimating the reductions in morbidity and mortality,



while the second discusses the methods used to value those



benefits.

-------
                               V-27


V.B.I.  Reductions in Hypertension and Related Morbidity
        and Mortality


     Estimating the reduction  in hypertension and cardiovascular

disease requires several steps.  The  first  is to estimate the

impact of the reduction in gas lead on levels of lead  in adults'

blood.  For that step, we used the regression analyses of the

NHANES II data reported in Chapter III.  Those regression

coefficients were applied to the NHANES  II  data base to simulate

the effects of gasoline lead reductions  on  blood lead  levels.

In each case, the blood lead levels in the  NHANES II data were

first adjusted to reflect reductions  that have occurred since

the time of the survey.  The subsequent  steps vary with the

condition involved, and are described below.


V.B.1.a.  Hypertension

      Estimating the change  in  the number of cases of hypertension

was straightforward;  the logistic regression coefficients from

Table V-6 were applied  to the  individual NHANES II data to predict

the numbers of hypertensives at  alternative levels of  gasoline

lead.  The change due to this  regulation was calculated by sub-

tracting the number at  the  new lead level from the number at  the

original lead level  (1.10 gplg).  Table  V-7 reports  the year-by-

year  estimates for three cases:   the  Final  Rule  (0.50  gplg on

7/1/85 and 0.10 gplg  on 1/1/86),  the  Proposed Rule  (0.10 gplg on

1/1/86), and the Alternative discussed  in the Notice  of Proposed

Rulemaking (0.50 gplg on 7/1/85,  0.30 on 1/1/86,  0.20  on  1/1/87,

and 0.10 on 1/1/88).  In all three cases, Table V-7  assumes  that

the rules eliminate all misfueling; alternative  assumptions

-------
                             V-28


TABLE V-7.  Reductions in Cases of Hypertension in Males Aged  40
	to 59, Assuming No Misfueling (thousands of cases)	

Rule	1985   1986   1987   1988   1989   1990   1991    1992

Proposed        0  1,804  1,727  1,649  1,562  1,489  1,396  1,399

Alternative   639  1,527  1,600  1,649  1,562  1,489  1,396  1,399

Final         639  1,804  1,727  1,649  1,562  1,489  1,396  1,399

-------
                               V-29





about misfueling are examined in Chapter VIII.  These estimates



cover only males aged 40 to 59, but include nonwhites as well as



whites.






V.B.l.b.  Myocardial Infarctions, Strokes, and Deaths



     Estimating the impact of reduced blood pressure on morbidity



and mortality required several additional steps.  Using the NHANES



II data and the regression coefficients in Table V-5, we simulated



the changes in individual blood pressure levels due to reductions



in gasoline lead.  Coefficients from two large studies of cardio-



vascular disease were then used to estimate changes in the numbers



of first-time myocardial infarctions, first-time strokes, and



deaths from all causes.



     The relationships between blood pressure and cardiovascular



diseases are well established by several large, long-term epidemic-



logical studies.  The classic study, which was important in



establishing cholesterol as a major factor in the risk of heart



disease, was the Framingham study (McGee et al., 1976).  Extensive



analyses of these data have yielded estimates of cardiovascular



risks associated with several variables, including blood pressure.



Figure V-3 shows the age-adjusted rates of death and heart attacks



as functions of blood pressure from that study.



     In the 1970s, the National Institutes of Health funded the



Pooling Project (The Pooling Project Research Group, 1978), which



combined the Framingham data with data from five other long-term



studies to improve the accuracy of the risk coefficients for heart



attacks.  The Pooling Project tested the Framingham coefficients

-------
                         V-30
                      FIGURE V-3


Adjlasted Rates of Death and Heart Attacks versus Blood Pressure:

                    Framingham Data
              Annual incidence of death by diastolic blood pressure


                     Males 45-74 (age adjusted rate)


              250
I
LI
s,
o
§
           u
           &
           s
           §
           •O
           0
           |
225



200



175



150



125



100
                    <70   70-  75-  80-  85-  90-  95-  100- 105- 110+
                          74   79   84   89   94   99   104  109
            n
            &
           o
           o
           o
           14
           s,
           o
           .5
         Annual incidence of nyocardial infarction by
         diastolic blood pressure


         Males 45-74 (age adjusted rate)
    100  -



    90



    80



    70  -



    60



    50



    40
                    <70  70-  75-  80-  85-  90-  95-  100- 105- 110+
                         74   79   84   89   94   99   104  109

-------
                               V-31





against the other study results and found that their predictive



power was good.  It then analyzed the first occurrence of myocard-



ial infarctions (serious heart attacks) in white men who entered



the studies at ages 40 to 59 and who were followed for at least 10



years.  Our estimates of the numbers of first-time myocardial



infarctions under alternative standards employ the Pooling



Project's coefficients.



     In addition to estimating the risk of heart attacks, the



Framingham study estimated regression equations for the risks of



stroke and death as functions of blood pressure and other vari-



ables.  Because the Pooling Project did not include those end-



points, we used the Framingham study coefficients.  As with heart



attacks, the estimates for strokes cover only first-time events;



thus, our estimates for strokes and myocardial infarctions are



biased downwards because they exclude second and subsequent heart



attacks and strokes associated with elevated blood pressure.  The



regression equation for deaths covers all causes of death; it



includes deaths not just from myocardial infarctions and strokes,



but also from other causes associated with blood pressure (e.g.,



heart diseases other than myocardial infarctions).



     Levy et al. (1984) recently tested the Framingham study



regression coefficients to see how well they explained the observed



decrease in cardiovascular mortality in the United States from



1970 to 1980.  They found that the coefficients, when coupled



with changes in blood pressure and other cardiovascular risk



factors over that same period, were able to explain about 80



percent of the drop in cardiovascular mortality.

-------
                               V-32





     The Hypertension Detection and Follow-up Program (NEJM, 1983)




found that intervention leading to about a 5mm Hg change in



diastolic blood pressure produced a 20 percent reduction in



overall mortality.  The Australian National Trial on mild hyper-



tension also found reductions in morbidity and mortality resulted



from lowered blood pressure (Lancet, 1980).  The Multiple Risk



Factor Intervention Trial found that drug therapy to lower blood



pressure reduced cardiovascular disease in persons with normal



resting electrocardiograms (ECGs), but increased it in persons



with abnormal resting ECGs (JAMA, 1982).  This suggests an



adverse affect of the drugs used.



     To produce estimates for all 40 to 59 year old white males,



the individual risk of each person sampled in the NHANES II was



summed and then averaged.  Since the sampled individuals repre-



sent the U.S. population for their specific age-race-sex category,



their average risk represents the average risk for all 40 to 59



year old white men.  Because blood lead levels have dropped since



the NHANES II period, we corrected for that change and then



evaluated the effects of the new lead-in-gasoline limits.  Again,



only white men were examined because there were too few blacks



in the Framingham study, and their risk might be different from



whites.




     The fact that gasoline lead levels would slowly decline even



without new EPA actions created a slight complication.  Because



gasoline lead levels fall over time in both our base case and the




low-lead case, the difference in blood lead levels resulting from



the rule will change over time.  Therefore, we recalculated the

-------
                               V-33





risk estimates for each year, assessing the annual change in blood



lead levels due to reductions in gas lead.



     The three cardiovascular-risk regression equations all predict



risk over the next 10 years, given current blood pressure, age, and



other characteristics.  Presumably, the risk in years 2-10 was



affected by blood pressure in those years, as well as by initial



blood pressure.  Because blood pressure levels over time in the



same individual are positively correlated, it is likely that



the regression coefficient in part picked up the effect of future



blood pressure levels.  Lacking any data with which to estimate



the pure effect of a one-year change in blood pressure, we divided



the coefficient for 10-year risk by 10.  The adjusted coefficient



was then used with the year-by-year predicted changes in blood



pressure to estimate risk reductions.  This procedure almost



certainly overcompensates, lending a downward bias to the results,



because current blood pressure is not perfectly correlated with



future blood pressure.



     We adjusted the population at risk for the increases in the



U.S. population of white males aged 40 to 59.  The regression



from the Framingham study predicting deaths for men aged 40 to



54 was extended to 40 to 59 for data comparability and uniformity.



Because the death rate actually increases with age, this also will



bias the results downward.



     Table V-8 reports the resulting year-by-year estimates of



reduced myocardial infarctions, strokes, and deaths for the three



phasedown schedules, as was done in Table V-7 for hypertension.



As before, these estimates assume that misfueling is completely

-------
                             V-34
TABLE V-8.  Reductions in Numbers of Cases of Cardiovascular
            Disease and Deaths in White Males Aged 40  to  59,
	Assuming No Misfueling	
Condition
  Rule
 1985  1986  1987  1988  1989  1990  1991  1992
Myocardial
  infarctions
    Proposed
    Alternative
    Final

Strokes
    Proposed
    Alternative
    Final

Deaths
    Proposed
    Alternative
    Final
    0 5,350 5,156 4,956 4,726 4,531 4,274 4,289
1,829 4,467 4,750 4,956 4,726 4,531 4,274 4,289
1,829 5,350 5,156 4,956 4,726 4,531 4,274 4,289
    0 1,115 1,074 1,032   984   943   889   892
  382   932   990 1,032   984   943   889   892
  382 1,115 1,074 1,032   984   943   889   892
    0 5,160 4,971 4,778 4,556 4,367 4,119 4,132
1,766 4,310 4,581 4,778 4,556 4,367 4,119 4,132
1,766 5,160 4,971 4,778 4,556 4,367 4,119 4,132

-------
                               V-35





eliminated.  They indicate that the rule being promulgated will



have a large impact on the incidence of cardiovascular disease.



We estimate that the reduction of lead in gasoline in 1986 alone



will result in 5,350 fewer myocardial infarctions; 1,115 fewer



strokes; and 5,160 fewer deaths from all causes among white



males aged 40 to 59.  Extending the analysis to men of other



ages and to nonwhites would substantially increase these estimates.





V.B.2.  Monetized Benefit Estimates



     Valuing reductions in morbidity and mortality is a difficult



and, to say the least, controversial task.  For morbidity, we have



restricted our estimates to avoided medical costs and foregone



earnings associated with diseases.  These estimates clearly



are too low, for they fail to account for other important losses



associated with disease, including pain and suffering (e.g.,



the paralysis that often follows a stroke).  For valuing the



reduction in mortality risk, we have chosen a fairly conservative



estimate ($1 million) from the large range obtained from studies



of occupational risk premiums.





V.B.2.a.  Hypertension



     Whether or not it results in coronary or cerebrovascular



disease, high blood pressure is a significant chronic illness.



It also generates economic costs, in the form of drugs, physicians'



visits, hospitalization, and work loss.  We used data from the



NHANES II and from the National Institutes of Health to estimate



the value of avoiding a case of high blood pressure.

-------
                               V-36






     The NHANES II ascertained how many times per year a person



saw a physician because of high blood pressure.  The weighted



average, for males 40 to 59 years old with diastolic blood pres-



sure over 90mm, was 3.27 visits per year.   We assumed a cost of




$35 per visit, for an annual total of $114.



     The same population was forced to remain in bed an average



of 0.41 days per year because of high blood pressure.  At the



average daily wage ($80), that translates  to $33 per year.



NHANES II also found that 29 percent were  on medication for



hypertension; assuming a drug cost of $200 per year for those



on medication yields an annual cost of $58.



     The National Hospital Discharge Survey (1979)  found that,



excluding those with heart disease or cerebrovascular disease,



people with high blood pressure used 3.5 million of the occupied



hospital bed-days that year; dividing by the 60 million people



the NHANES II identified as having high blood pressure gives a



rate of 0.058 hospital bed-days per person per year.  We have



assumed that these results apply to the 40 to 59 year old age



group, as well.  Using a daily hospital cost of $400 yields an



annual cost per hypertensive of $23.



     Summing these estimates yields a total of $228 per



hypertensive per year.  It should be noted that only 29 percent



of the people with blood pressure above 90mm in the NHANES II



were on medication, in part because some of them had not pre-



viously been detected as having high blood pressure.  Therefore,




the average cost for a detected case will  be higher.  For example,



Weinstein and Stason (1977) used an average cost of $200 in 1975

-------
                               V-37





dollars, or about $450 in 1983 dollars, for treatment of patients



undergoing medical care for hypertension.  Nevertheless, we have



conservatively used $220 as the value of avoiding one case of



high blood pressure for one year.





V.B.2.b.  Myocardial Infarctions



    Our estimate of the benefits of reducing the incidence of



myocardial infarctions relies heavily on Hartunian et al. (1981),



who estimated the medical expenses and lost wages associated with



a variety of diseases.  Under the category of myocardial infarc-



tions (MI), Hartunian et al. examined three types of cases:



sudden death, fatal MI, and nonfatal MI.  ("Sudden death" was



classified as a myocardial infarction in the Pooling Project



regression coefficients we used.)



     For each category and each age group, Hartunian et al.



obtained data on the type of medical services needed (e.g., ambu-



lance and coronary intensive care unit), the fraction of cases



using each service, and the costs in 1975 dollars.  They also



determined the annualized recurrence and follow-up costs, by age,



for each condition.  These were then discounted (using a 6 percent



real discount rate) to the time of initial occurrence to estimate



the cost, in current dollars, of each new case.  The resulting



estimates were $96 for sudden death and $7,075 for both fatal and



nonfatal Mis.



     We have adjusted these 1975 estimates in three ways to



reflect 1983 conditions.  First, we inflated them to 1983 dollars.



Because most of the costs were hospital-related, with the rest

-------
                               V-38






principally being physicians' fees, we inflated the Hartunian et




al. cost estimates by a weighted average of 80 percent of the



change in the Consumer Price Index (CPI) for hospital rooms and




20 percent of the change in the CPI for physicians' charges.



Approximately 90 percent of the Hartunian et al. MI costs were



hospital-related, not physicians'  fees, and hospital costs rose



faster than physicians' fees, so this approach is conservative.



     The second adjustment in the 1975 estimates involves changing



cost indices.  Because cost indices only account for increased



costs of the same procedure, in this case principally the initial



hospitalization for a heart attack, they do not reflect the cost



of new or different procedures.  Since 1975, the fraction of people



suffering coronary heart disease who subsequently undergo coronary



bypass operations has increased substantially.  The number of by-



pass operations tripled in seven years, from 57,000 in 1975 to



170,000 in 1982, while the number of cases of coronary heart



disease has remained relatively constant (National Centers for



Health Statistics, Hospital Discharge Survey, and unpublished



data).  Based on the Hartunian et al. data, 7.1 percent of MI



cases in 1975 had subsequent bypass operations.  Assuming that



they shared proportionately in the tripling of the bypass-opera-



tion rate, we estimated that an additional 14 percent of Mis now



result in a bypass operation.  Hartunian et al. estimated the



cost of bypass operations at $6,700 in 1975 dollars, or $16,800



in 1983 dollars.  Adding 14 percent of this cost to the other



direct costs yields an estimate of the total direct costs in



1983 dollars of $20,100 for an MI and $240 for sudden death.

-------
                               V-39





     Our third adjustment involved discount rates.  Hartunian



et al. used a 6 percent real discount rate to present value the



future year costs, whereas this analysis employs a 10 percent



discount rate.  Fortunately, Hartunian et al. performed sensi-



tivity calculations for other discount rates, including 10 per-



cent.  Making all of these adjustments, the costs per case are



$18,100 for an MI and $216 for sudden death.



     Hartunian et al. also obtained data indicating the proba-



bility distribution of cases among the different categories.  Of



the total number of cases in these three categories, about 22.5



percent were sudden deaths and the remaining 77.5 percent were



fatal or nonfatal Mis.  Applying those percentages to the



medical-cost estimates derived above yields a weighted average



of $14,076 per myocardial infarction.



     Hartunian et al. calculated the present value of fore-



gone earnings based on reduced labor force participation using



data on each type of heart disease, broken down by sex and 10-year



age categories.  We have used those results, with several modi-



fications.  First, we excluded foregone earnings for fatal heart



attacks, because we valued the reduction in mortality risks



separately (see Section V.B.2.d., below).  Second, we adjusted



for the increase in average non-farm compensation from 1975 to



1983, using information from Data Resources, Incorporated.



Finally, we again used a discount rate of 10 percent, rather than



the 6 percent used by Hartunian et al. in their base case



analysis.

-------
                               V-40





     The resulting estimates of foregone earnings are $90,000 for




heart attack victims under 45; $47,000 for those between 45 and



54; and $22,000 for those over 55.  Based on data from the



Pooling Project and NHANES II, 16.1 percent of nonfatal heart



attacks in men between 40 and 59 occur in those under 45, 50.9



percent occur in those between 45 and 54, and 33 percent in those



55 and older.  Using those percentages yields a weighted average



for lost earnings of $45,670 per attack.  Combining that earnings



estimate with the earlier one for medical costs yields a total



benefit per myocardial infarction avoided of about $60,000.






V.B.2.C.  Strokes



     Our estimates of the benefits of avoiding strokes also rely



on Hartunian et al., with similar adjustments.  (Unlike myocardial



infarctions, we have not adjusted their medical cost estimates for



strokes to reflect any changes in medical treatment since 1975.)



Table V-9 presents the estimates for three types of stroke —



hemorrhagic, infarctive, and transient ischemic attacks (TIA) —



by age.  The averages are based on the distribution of types of



strokes and  incidence of strokes by age.  The overall average is



$44,000 per  stroke avoided.



     We have been unable to estimate a value for avoiding the



loss in quality of life that occurs in stroke victims.  This



is a significant omission.  For example, of the people in the



NHANES II who reported having had a stroke in the past, 45 percent



suffered paralysis in the face and 13 percent still had at least



partial facial paralysis, 54 percent suffered paralysis in at

-------
                             V-41
TABLE V-9.  Benefits of Reducing Strokes (dollars per case)
Type of Stroke
Age
Hemorrhagic
35-44
45-54
55-64
Inf arctive
35-44
45-54
55-64
Transient ischemic attacks
35-44
45-54
55-64
Medical
Expenses
12,600
13,300
17,200
17,600
18,100
23,600
3,184
3,184
3,184
Foregone
Earnings
41,000
26,000
11,000
71,000
43,000
14,000
1,114
3,076
8,280
Total
53,600
39,300
28,200
88,600
61,100
37,600
4,298
6,260
11,464
Weighted average
44,000

-------
                               V-42






least one arm and 21 percent remained paralyzed, 59 percent had



numbness in arms or legs and 28 percent had remaining numbness,



30 percent had vision impairment and 13 percent remained visually



impaired, and 50 percent had speech impairment with 22 percent



continuing to suffer from speech impairment.  While we have no




estimates of people's willingness to pay to avoid the risk of



these profound injuries, common sense suggests that it is high.






V.B.2.d.  Mortality



     Valuing reductions in mortality is highly controversial.



Over the past decade or so, a substantial literature has developed



on the subject.  Economists are in general agreement that the



best conceptual approach to use is the willingness-to-pay (WTP)



of the individuals involved.  The appropriate value is not



the amount that an individual would pay to avoid certain death,



but rather the total sum that a large group of individuals would



pay to reduce small risks that sum to one; for example, the



amount that 10,000 people would pay to reduce a risk to each of



them of one in ten thousand.



     Several studies have estimated WTP based on implicit tradeoffs



between risk and dollars revealed in market transactions.  Most



of these studies (e.g., Thaler and Rosen, 1976; Smith, 1974 and



1976; Viscusi, 1978) have studied labor markets, based on the



premise that, all else being equal, workers must receive higher



wages to accept a higher risk of being injured or killed on the



job.  Such studies typically regress wages on risk and a variety



of other explanatory variables (e.g., levels of education required,

-------
                               V-43





worker experience, whether or not the industry is unionized,



location, and non-risk working conditions).  In such regressions,



risk might be measured as the number of fatalities per 1,000



workers per year.  The coefficient for that variable is then



interpreted as the amount of extra wages needed to compensate



for a 0.001 risk of death.  Dividing the coefficient by the unit



of risk yields the estimate of WTP to avoid a statistical death.



For example, if the coefficient is $500, the estimated WTP is



$500,000 (= $500/0.001).



     A few studies have estimated WTP in nonoccupational settings,



Blomquist (1977), for example, estimated the implicit cost-risk



tradeoffs that individuals make in deciding whether or not to



take the time to put on seat belts.



     None of these studies yields definitive answers.  All suffer



from data limitations (e.g., incomplete information on possible



confounding variables and on the extent to which individuals



perceive the risks they face).  Not surprisingly, given these



problems, the studies also yield a wide range of estimates.  A



recent survey of the literature prepared for EPA found a range



of $400,000 to $7 million per statistical life saved (Violette



and Chestnut, 1983).  Based on that survey, EPA1s RIA guidelines



do not attempt to set any specific value, but rather recommend



that range.  To simplify the presentation of the results, this



RIA uses a single value from the lower end of that range, $1



million per statistical life saved.  Although we do not present



any formal sensitivity analyses on this value, the results  in



Chapter VIII show that the net benefits are so large that they

-------
                               V-44


would remain positive whatever part of that broad range were

used; even at $400,000 per statistical life saved, the estimated

benefits would be many times higher than the costs.


V.B.3.   Summary of Blood Pressure Benefits

     Table V-10 summarizes the benefits of reducing the numbers

of cases of hypertension, myocardial infarctions, strokes, and

deaths due to high blood pressure.  As in earlier tables, these

estimates assume that misfueling is eliminated.  These are con-

servative estimates for several reasons:

     (1)  The hypertension estimate covers only males aged 40
          to 59.

     (2)  The other estimates cover only white males aged 40
          to 59.

     (3)  We have not assigned any value to reduced pain and
          suffering associated with hypertension, myocardial
          infarctions, and strokes.

     (4)  We have not estimated any health benefits for adults
          other than those related to blood pressure.

In addition, of course, some readers may quarrel with the value

assigned to reduced risk of mortality; we have chosen a single

value for convenience, not because we believe any particular

value can be defended strongly.  Despite these limitations, the

estimated benefits of the final phasedown rule are large, total-

ing $5.9 billion in 1986.

     As discussed at the beginning of this chapter, these esti-

mates should be treated as tentative.  Although the two key

studies (Pirkle et al., 1985; Harlan et al., 1985) recently have

been published in peer-reviewed journals, they have not yet been

-------
                              V-45
TABLE V-10.  Year-by-Year Estimates of Blood Pressure Benefits,
	Assuming No Misfueling (millions of 1983 dollars)	

Rule	1985   1986   1987   1988   1989   1990   1991   1992

Proposed         0  5,927  5,707  5,484  5,227  5,008  4,722  4,736

Alternative  2,033  4,955  5,262  5,484  5,227  5,008  4,722  4,736

Final        2,033  5,927  5,707  5,484  5,227  5,008  4,722  4,736

-------
                               V-46






widely available for public review.  As a result, EPA has not



relied on blood-pressure-related health effects in reaching a



decision on the final phasedown rule.  These potentially serious



health effects will be considered by EPA, however, in connection



with a possible ban on lead in gasoline, and extensive



review and comments will be sought on them.

-------
                             CHAPTER VI



         BENEFITS OF REDUCING POLLUTANTS OTHER THAN LEAD






     Decreasing the amount of lead in gasoline will reduce



emissions of several pollutants in addition to lead.  Most of



these reductions will result from decreased "misfueling," the



misuse of leaded fuel in vehicles equipped with pollution-control



catalysts.  In such vehicles, leaded gasoline poisons the cata-



lysts, greatly reducing their effectiveness in controlling emis-



sions of hydrocarbons (HC), nitrogen oxides (NOX)> and carbon



monoxide (CO).  Reducing lead in gasoline should affect emissions



from misfueling in two ways.  First, it will be more expensive to



produce 89 octane leaded gasoline at 0.10 gplg than to produce 87



octane unleaded gasoline.  This change in relative manufacturing



costs should alter retail price differentials (although, as



discussed in Chapter 2, it may not make unleaded cheaper than



leaded at the pump), thus reducing the incentive to misfuel.  In



addition, even for those vehicles that continue to be misfueled,



it will take considerably longer to destroy the effectiveness of



catalysts with 0.10 gplg leaded gasoline than it does now with



1.10 gplg.



     All three of these pollutants have been associated with



damages to health and welfare, and contribute to ambient air



pollution problems covered by National Ambient Air Quality



Standards (NAAQS).  CO is itself a "criteria pollutant," covered



by a NAAQS.  NOX is the composite formula tor nitrogen oxide  (NO)



and nitrogen dioxide (N02); N02 is covered by a NAAQS.  Although



most NOX is emitted as NO, some of it is chemically transformed

-------
                               VI-2






in the atmosphere to NO2.  NOX and HC both contribute to the



formation of ozone (03), another criteria pollutant.  In addition,



certain hydrocarbons (in particular, benzene)  have been linked to




cancer.



     Independent of its effect on misfueling,  reducing lead in



gasoline also will reduce emissions of ethylene dibromide (EDB),



which is added to leaded gasoline as a "scavenger" to reduce the



build-up of lead deposits in engines.  Because EDB is added in



proportion to the amount of lead, tightening the lead standard



will reduce the amounts added to gasoline.  EDB has been linked




to increased risk of cancer.



     This chapter examines the impacts of reducing these



pollutants, focusing on the three associated with misfueling:



HC, NOX, and CO.  Section A estimates the emissions caused by



misfueling.  Section B addresses the ozone-related health and



welfare effects, which account for the vast majority of the



benefits that we were able to quantify.  Section C discusses the



health and welfare gains associated with pollutants other than



ozone.  We have tried to estimate as many of the effects of



these pollutants at ambient concentrations as  possible, but



our quantitative estimates are subject to considerable uncer-



tainty and provide incomplete coverage of the  potential effects



of emission reductions.




    In both Sections B and C, the estimates of health effects



are presented in physical rather than monetary units, but non-



health effects (such as crop losses) are estimated in dollars.




Finally, in Section D, we estimate the monetized economic benefits

-------
                               VI-3





of eliminating misfueling using two methods.  The first values



directly the health and welfare effects estimated in Sections B



and C.  That method is conceptually the more appropriate one,



but it omits some important categories because of incomplete



quantification.  These emissions are likely to lend a downward



bias to the direct estimates.  This downward bias is most obvious



in the case of CO, for which we have not monetized any benefits,



but also is of major concern for NOX, for which we have identified



but have been unable to quantify, several potentially significant



benefit categories, in particular health effects and damages from



acid deposition.  Even in the case of HC, for which we have



quantified significant ozone-related benefits, the estimates may



be biased downward significantly because of our inability to



quantify ozone's impacts on chronic health conditions and forests,



nor have we estimated direct health effects of any HCs other than



benzene.  Further, there is considerable uncertainty in those



categories we have included.



     The second method monetizes the emission reductions using



the values implied by the cost of the pollution control equipment



needed to meet the emission limits set by Congress.  Our final



monetized benefit estimates, used in later chapters to compute



the total and net benefits of alternative rules, are the averages



of these two methods.



     The basic methodology used in this chapter is the same as



that employed in Schwartz et al. (1984) and in the preliminary



RIA.  The analysis has been refined in several areas, however.

-------
                               VI-4






First, to estimate emissions associated with misfueling, we have



used the results of EPA's 1983 tampering and misfueling survey,



the results of which were not available for use in the earlier




documents.  Second, the estimates of the health effects associated



with ozone exposure rely on an updated statistical model.  Third,



in this document we do not rely on a peoperty-value-based estimate




of NOX estimates, because closer examination of the underlying



study suggested that it had not adequately accounted for con-



founding factors, in particular other pollutants whose concentra-



tions may covary with NOX.  Finally, and most importantly,



additional and more detailed analysis of the impact of HC and NOX



emissions on ozone has caused us to revise downwards significantly



our estimates of the rule's impact on ozone concentrations.



     Throughout this chapter, we estimate the effects and monetized



benefits of eliminating misfueling altogether.  For lead levels



other than a complete ban on all leaded gasoline, this is probably



an overly optimistic assumption.  Chapter VIII presents estimates



based on a broader range of alternative assumptions about the



impacts of different rules on misfueling.






VI.A.  Emissions Associated with Misfueling



     "Misfueling" or "fuel switching" refers to the use of leaded



gasoline in a vehicle originally designed and certified to use



unleaded gasoline.  Because leaded regular gasoline is cheaper




and higher in octane than regular unleaded, some drivers deliber-



ately misfuel their vehicles in an attempt to reduce expenses or



to improve vehicle performance.

-------
                               VI-5





     Misfueling can occur by removing or damaging the nozzle



restrictors installed in the fuel inlets of vehicles with cata-



lytic converters, by using an improper size fuel nozzle, or by



funneling leaded fuel into the tank.  Sometimes gasoline retailers



sell gasoline that is mislabeled or contaminated (U.S. EPA,



1983a),  but this accounts for less than 1 percent of misfueling.



     It is illegal for service stations or commercial fleet



owners to misfuel or to allow the misfueling of vehicles origin-



ally equipped with catalytic converters.  Federal law does not



apply to individuals who misfuel their own vehicles, however.



     Using leaded gasoline in vehicles with catalytic converters



damages this pollution control equipment, and can increase



emissions of HC, CO, and NOX by as much as a factor of eight.



Table VI-1 shows the emissions increases caused by misfueling  on



a per-mile basis.  The estimates distinguish between pre- and



post-1981 vehicles because emission standards changed in that  year,



leading to changes in the design of catalysts and other emission



control devices.  In vehicles manufactured before 1981, misfueling



has no effect on NOX emissions, but does cause relatively



large increases in HC and CO emissions.



     Misfueling is a significant problem.  Several recent surveys



by EPA have shown that a substantial number of vehicles are mis-



fueled with leaded gasoline.  According to the 1983 survey



(U.S. EPA, 1984e), about 15.5 percent of light-duty vehicles



designed to use unleaded gasoline are misfueled with leaded.  The



1982 survey (U.S. EPA, 1983a), showed a lower rate, about 13.5

-------
                               VI-6


TABLE VI-1.  Increase in Emissions Due to Misfueling (grams/mile)

Light-Duty Vehicle Model Years	HC	CO	NOX	

1975 to 1980                            2.67    17.85     0.0

1981 and later                          1.57    11.07     0.71


Source:  U.S.  EPA, Office of Mobile Sources, "Anti-Tampering and
         Anti-Misfueling Programs to Reduce In-Use Emissions from
         Motor Vehicles," May 23, 1983.

-------
                             VI-7





percent.  Misfueling rates apparently vary by the age of the



vehicle, by whether it is in an area with an Inspection and



Maintenance (I/M) mobile source enforcement program, by whether



it is part of a commercial fleet, and other factors.  Table VI-2



provides 1983 misfueling rates by model year of vehicle and by



T/M status.  We assumed for our analysis that the rates of



misfueling by age of vehicle would stay constant in the absence



of new regulations.



     The EPA surveys probably underestimate real misfueling rates



by a significant margin, primarily because vehicle inspections for



misfueling are voluntary, which would bias the results downward



(assuming that misfuelers are less likely to agree to have their



vehicles tested).  In some areas, the rates of drivers refusing



inspections were very high.  In the 1982 survey, the refusal rates



ranged from 1 to 8 percent in I/M areas, and from 3 to 44 percent



in non-I/M areas.



     To estimate the reduction in emissions that would be achieved



by eliminating misfueling, we combined the data in Tables VI-1 and



VI-2 with our fleet model (described in the Appendix), which pro-



jected the number of vehicles of each model year and their annual



mileage.  For each year of our projection, we first estimated the



number of vehicles that would have misfueled for the first time



in that year.   (We assumed that no emission reductions would



result from stopping the misfueling of vehicles that already had



their catalysts destroyed by misfueling in earlier years.)  We



then projected over the remaining lifetime of the vehicle the



expected excess pollutants it would have emitted due to misfuel-

-------
                               VI-8
TABLE VI-2.  Misfueling Rates in 1983 (percent)
Model
Year
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
Weighted
Overall
Misfueling Rates
1.6
5.1
8.0
10.0
9.0
19.6
19.0
22.6
25.5
25.9
Average:* 15.5
I/M Areas
8.7
4.8
3.2
7.9
6.2
17.9
7.3
23.2
13.9
16.0

Non-I/M Areas
0.0
5.2
9.0
10.5
9.7
20.0
21.6
22.4
28.0
28.0

  This weighted average does not account for the number of miles
  driven by each model year.

-------
                               VI-9





ing.  The projections account for the facts that older vehicles



drive fewer miles per year, that survival rates decline with age,



and that the efficiency of emission control devices deteriorates



with age.  These projected emission streams were then discounted



(at a 10 percent real rate) back to the year in question.  Table



VI-3 presents the resulting year-by-year estimates of the dis-



counted emissions avoided by eliminating misfueling.  Note that



the estimate for each year is not of actual emission reductions



achieved in that year, but rather the discounted value of emis-



sion reductions due to stopping the first-time misfueling of



vehicles in that year.  The numbers of tons of HC and CO controlled



remain fairly constant from 1986 through 1992, at over 300,000



tons of HC and more than 2.5 million tons of CO.  The estimates



of reduced NOX emissions increase, from 94,000 tons in 1986 to



150,000 tons in 1992, because of the increase in the proportion



of post-1981 vehicles.  (Recall that misfueling does not increase



NOX emissions in vehicles manufactured before 1981.)





VI.B.  Health and Welfare Effects Associated with Ozone



     Hydrocarbons and nitrogen oxides react photochemically to



form ozone ("smog").  Ozone in turn affects health, materials



damage, and vegetation.  To estimate the ozone-related effects of



reducing emissions of HC and NOX, we employed a two-step process.



First, as described below in Section VI.B.I, we estimated the



health and welfare effects of a 1 percent change in ozone.  Second,



as discussed in Section VLB.2, we estimated the relationship



between reducing HC and NOX and the subsequent decrease in ozone.

-------
                             VI-10


TABLE VI-3.  Year-by-Year Estimates of Reductions in Emissions,
	Assuming No Misfueling (thousands of metric tons)	

Pollutant	1985  1986  1987  1988  1989  1990  1991  1992

Hydrocarbons          155   305   303   303   303   308   320   331

Nitrogen Oxides        40    94   107   119   130   139   145   150

Carbon Monoxide     1,067 2,116 2,114 2,122 2,131 2,174 2,255 2,333

  Total             1,262 2,515 2,524 2,544 2,564 2,621 2,720 2,814

-------
                            VI-11






The material in these sections closely parallels McGartland and



Ostro (1985).  Several of the studies relied upon are EPA con-



tractor reports in progress or in draft; as such they have not



undergone full peer review and should be considered preliminary.





VI.B.1.  Effects of a 1 Percent Reduction in Ozone



    In calculating the effects of a 1 percent change in ozone, we



relied primarily on dose-response estimates.  That is, we applied



change in ambient levels.  Occasionally, as a validity check of



the benefit estimates, we interpolated from existing aggregate



damage estimates to project the impacts of a single pollutant or



of a given change in ambient levels.  Regardless of the approach,



the benefit estimates are uncertain and should be interpreted



with caution.  Unless noted otherwise, we assumed a constant



benefit per ton of pollution control over the relevant range.



     The effects of ozone on human health, vegetation, materials,



and ecosystems were summarized in the EPA Air Quality Criteria



for Ozone and Other Photochemical Oxidants (U.S. EPA, 1978).  In



addition, we have relied on the considerable amount of research



that has become available since that document was finished.  As



part of EPA's periodic review of the ozone NAAQS, the Office of



Research and Development currently is updating the Criteria



Document.  Nothing in this report is intended to prejudge or



supercede the outcome of that process.





VI.B.I.a.  Health Effects of Reducing Ozone



     Studies of the effects of ozone on human health have



investigated the relationships between changes in ozone concen-

-------
                            VI-12



trations and changes in lung function; decrements in physical




performance; exacerbation of asthma;  incidence of headaches;



respiratory symptoms, such as coughing and chest discomfort;



eye, nose, and throat irritation; and changes in blood parameters




(U.S. EPA, 1978; Goldstein, 1982; Ferris,  1978).



     Uncertainty remains about whether a threshold level exists



for ozone and, if so, at what level.   For example, McDonnell et al.



(1983) found a nonlinear relationship between health and ozone



exposure that "flattened" at ozone levels below 0.18 parts per



million (ppm) — a level above the ambient concentrations in most



metropolitan areas.  If such a threshold exists, the health bene-



fits of reducing ozone from its current levels would be minimal.



Population studies by Zagraniski et al. (1979) and Lebowitz et al.



(1984), however, suggest effects may  be occurring at ambient



levels as low as 0.08 ppm.  Moreover, other studies (Portney and



Mullahy, 1983; Hasselblad and Svendsgaard, 1975) do not support



the existence of any threshold for health effects.



     Hammer et al. (1974) found associations between increased



oxidants and respiratory symptoms (such as cough and chest discom-



fort) and other symptoms (such as eye irritation and headache) in



young, healthy adults.  They obtained the symptom rates from daily



diaries and adjusted them by excluding days on which subjects




reported fevers.  Makino and Mizoguchi (1975) found a correlation



between oxidant levels and eye irritation and sore throats in



Japanese school children.  Lippmann et al. (1983) and Lebowitz




et al. (1982, 1983, 1984) found evidence of decreased athletic



performance, increased prevalence of  acute symptoms, and




dysfunction of pulmonary systems resulting from ozone exposure.

-------
                            VI-13






     In addition to these studies of the general population,



Whittemore and Korn (1980), Linn et al. (1981), Bates and Sizto




(1983), and others have shown that asthmatics and people with



other chronic respiratory diseases may he particularly sensitive



to ozone or other oxidants.  Even low levels of exposure to photo-



chemical oxidants have been shown to provoke respiratory symptoms



in individuals with predisposing factors, such as smoking or



respiratory illness (Zagraniski et al., 1979).



     There is also evidence linking reduced respiratory function



-- measured as Forced Expiratory Volume (FEV) and Forced Venti-



lating Capacity (FVC) -- to ozone exposure.  For example, McDonnell



et al. (1983) reported an association for normal subjects while



exercising.  Folinsbee et al. (1984), Horvath et al. (1979), and



Adams and Schelegle (1983) found an association between decrements



in FEV and ozone exposure.



     Unfortunately, it is difficult to estimate from these studies



the potential economic benefits of reducing health effects related



to ozone exposure because they did not estimate dose-response



functions.  Most were designed to investigate potential thresholds,



or simply to determine if any relationship existed between ozone



and particular effects.  In addition, studies using lung function



changes as the health endpoint fail to provide a measure that can




be valued in economic terms.



     Recent work by Portney and Mullahy (1983, 1985) at Resources




for the Future (RFF) is an exception.  They considered the effect



of alternative levels of ozone on various health measures, combining



individual health data from the Health Interview Survey  (HIS)

-------
                              VI-14






with data on pollution concentrations during the same period



covered by the survey.  In the HIS, interviewees provided infor-



mation on their health status during the two weeks preceding the




survey.  As their health measure, Portney and Mullahy focused on



the number of days of restricted activity due to a respiratory



condition (RADRESP).  The RADRESP measure included days when the



symptoms were relatively minor, as well as those when they were




serious enough to confine individuals to bed or to make them



miss work.



     Portney and Mullahy regressed RADRESP on a dozen or more



independent variables, including socioeconomic and demographic



factors, chronic health status, urban variables, and other pol-



lutants, as well as ozone.  As their ozone measure, they used



the daily maximum one-hour concentration (measured in parts per



million) averaged over the two-week period covered for the indi-



vidual.  They considered several different specifications and



functional forms.  In their ordinary least squares (OLS) regres-



sions, they tried various forms of the ozone measure, including



the square and the square root as well as the untransformed vari-



able.  The first part of Table VI-4 summarizes the OLS results.



In all three specifications, the ozone coefficient was positive,



but not significant at the 95 percent confidence level.  For the



specifications using the linear ozone term, the coefficient on



ozone  represents the average change in RADRESPs per person per



two weeks for a one ppm change in ozone.  Thus, for example,




that coefficient predicts that reducing the average daily maximum

-------
                            VI-15
TABLE VI-4.  Regression Results for Portney and Mullahy Study
             on Respiratory Symptoms Related to Ozone
Ozone Specification
OLS
Linear
Square root
Squared
Ozone
Coefficient
1.2185
0.8076
0.4667
t-statistic
1.13
1.66
0.07
F-statistic
2.743
2.867
2.636
Poisson                         6.8827       1.97       N.A.

                               (log-likelihood ratio = -1395.4)

-------
                              VI-16






ozone concentration by 0.01 ppm for one year for a population of



1 million adults would decrease the number of RADRESPs by 316,800




(= 0.01 x 1.22 x 1,000,000 x 52/2).



     In subsequent analysis, Portney and Mullahy (1985) estimated



the relationship using a Poisson model, which can be written as:






                      E(RADRESP) = exp(XB),






where E(RADRESP) is the expected number of RADRESPs and XB is the




vector of independent variables and their coefficients.  The



second part of Table VI-4 summarizes the results of the Poisson



model.  Because the Poisson model is nonlinear, the ozone coeffi-



cient is slightly harder to use for extrapolation.  The Poisson




model, however, appears to fit the data better; RADRESPs have a



Poisson-like distribution.



     In separate models estimating RADRESPs for children, Portney



and Mullahy (1983) did not find any consistently significant



effects.  As a lower bound, therefore, we assumed no affect on



RADRESP for children in the general population.  However, incom-



plete data for children and the reliance on parents to report



child-related health effects may explain this result; a restric-



tion in activity probably was less likely to be reported for a



child.  Public health scientists continue to debate whether



children are as susceptible to ozone as adults.  Older children,



for example, may not be as susceptible because they typically



have large excess lung capacity.  On the other hand, damages to



the lungs of a child may result in a chronic respiratory condition

-------
                              VI-17






in adulthood.  Therefore, to place a plausible upper bound on our



estimates, we applied the adult coefficients to children as well.



     To estimate the change in RADRESPs due to a 1 percent change



in ozone, we simulated the change using the data on individual



exposures and characteristics constructed by Portney and Mullahy.



Their study matched the 1979 HIS with air quality data, weather



stations, and other area-specific data.  Using the estimated



regressions, we rolled back the exposure of each person in the



data base by 1 percent.  Using data from the Census Bureau, we



assumed a population of 230 million in 1984, with 70 percent of



the total above age 17.



     The changes in total annual RADRESPs for adults predicted by



the Poisson and linear models were quite similar, 2.1 and 2.4



million, respectively.  Because the Poisson model provided a



better fit of the data, we used it as the basis for our estimates.



For our high and low estimates of adult effects, we used plus or



minus one standard deviation of the ozone coefficient.  For child-



ren, our low estimate was zero and our high estimate was 0.90



million cases (based on the adult coefficient).  For our medium



or point estimate, we used the midpoint of those extremes, 0.45



million cases.  These results should be interpreted cautiously,



because cross-sectional studies of this type can be extremely



sensitive to model specification, functional form, omitted



and confounding variables, and the ambient air monitors used.



     Portney and Mullahy (1985) also used a multinomial logit



model to estimate the marginal impact of ozone on the two types



of RADRESPs — the more serious ones resulting in a day of bed

-------
                              VI-18






rest or of lost work,  and the less serious ones that resulted in




a more minor restriction of normal activity.  That analysis




suggested that a marginal change in ozone was three times as



likely to cause a day  of minor restricted activity than a day of




bed rest or lost work.



     To provide an alternative estimate of respiratory conditions



and a separate estimate of nonrespiratory irritations, we used the



results of a statistical reanalysis of the Hammer et al.  (1974)



study discussed earlier.  In an unpublished paper, Hasselblad



and Svendsgaard (1975) fit simple logistic regressions to estimate



the relationship between ozone concentration (measured as a



daily maximum hourly concentration) and eye irritation, headache,



coughing, and chest discomfort.  The probability of a response



at an ozone level, X,  measured in parts per hundred million



(pphm), was given as:






             p(X) = C  + (1 - C)/[l + exp (-A - BX)]






Table VI-5 presents the estimates of the parameters for the four



outcome measures.  These coefficients must be interpreted cau-



tiously, because Hasselblad and Svendsgaard did not control for



some possible confounding variables, in particular temperature



and humidity.  In a later, published paper, Hasselblad (1981)



fit multiple logistic  regression models to these same data, but



that study paper does  not report the regression coefficients we



needed to make our estimates.

-------
                              VI-19
TABLE VI-5.  Regression Coefficients Relating Respiratory
             and Non-respiratory Symptoms to Ozone
Condition
Respiratory Effects
Cough
Chest discomfort
Non-respiratory Effects
Eye irritation
Headache
A B

-2.98 0.0092
-3.53 0.0023

-4.96 0.0907
-4.88 0.0470
C

0.0450
0.0166

0.0407
0.0976

-------
                              VI-20





     We used these estimated dose-response functions with the



individual information in the Portney and Mullahy data set to



simulate the effects of a 1 percent reduction in ozone.  For



coughs and chest discomfort, the Hasselblad and Svendsgaard coeffi-



cients yielded a total of 1.54 million adult cases per year for



a 1 percent reduction in ozone, compared to the estimate of 2.1



million based on the Portney and Mullahy Poisson model.  These



estimates are remarkably consistent, as the RADRESP measure used



by Portney and Mullahy included other symptoms besides cough and



chest discomfort.  In addition, the Hammer et al. sample used by



Hasselblad and Svendsgaard consisted of student nurses, who were



young and generally healthy, while the Portney and Mullahy sample



was more representative of the general population.  For those



reasons, our estimates for respiratory conditions rely solely on



Portney and Mullahy's results.



     The results of Hasselblad and Svendsgaard also can be used to



estimate the number of nonrespiratory conditions, such as eye



irritation and headache, possibly related to exposure to ozone



and other photochemical oxidants.  There is evidence suggesting



that these symptoms are not related to ozone per se, but rather



to other oxidants, such as peroxyacetyl nitrate (PAN), whose



production may be proportional to that of ozone.



     To account for this possibility, we used the estimates based



on Hasselblad and Svendsgaard as point estimates for nonrespiratory



irritations (headache and eye irritation).  Unfortunately, these



researchers did not report the standard errors, so a confidence



interval could not be determined.

-------
                          \  /  VI-21


     These ozone health benefits reflect the likely acute effects

generated by intense, short-term exposure to ozone.  Long-terra

exposure to ozone also may affect the health of some people, but

the epidemiological evidence on chronic ozone effects is sparse.

One of the available studies, Detels et al. (1979), compared the

effects of prolonged exposure to different levels of photochemical

oxidants on the pulmonary functions of both healthy individuals and

individuals with chronic obstructive pulmonary disease.  Persons

exposed to an annual mean of 0.11 ppm of oxidant, compared to a

control group exposed to 0.03 ppm of oxidant, showed statistically

significantly increased chest illness, impairments of respiratory

functions, and lower pulmonary function.*

     While the epidemiological evidence of the effects of long-term

exposure to ozone is sparse, several animal experiments have demon-

strated effects on lung elasticity, blood chemistry, the central

nervous system, the body's ability to defend against infection, and

the rate at which drugs are metabolized (U.S. EPA, 1983f).  Unfor-

tunately, it is not possible to extrapolate those results to humans.

Therefore, we could not quantify the chronic health effects attri-

butable to ozone, but we believe that some of these effects may be

present at current ambient levels.
* At workshops related to the development of the Criteria Document
  for ozone, some shortcomings in this analysis were noted.  For
  example, the study group was also exposed to higher levels of
  NO2 and 804, and there were some questions about the adequacy of
  the measurement of ozone exposure, about the subject selection,
  and about the test measures.  Although it is both reasonable and
  likely that long-term exposures affect health, the failure to
  correct for the effects of other pollutants raises uncertainties
  about the specific findings.

-------
                              VI-22





     Table VI-6 sununarizes our estimates of the acute health



effects of a 1 percent change in ozone.   The estimates of respir-



atory effects are based on the Portney and Mullahy results, while



the nonrespiratory effects are derived from Hasselblad and



Svendsgaard's analysis.  In the latter case, high and low estimates



are not reported because Hasselblad and  Svendsgaard did not calcu-



late standard errors for their estimates.  As discussed above,



these estimates omit any quantification  of chronic health effects.





VI.B.l.b.  Ozone Agricultural Effects



     Ozone, alone or in combination with sulfur dioxide and



nitrogen dioxide, is responsible for most of the U.S. crop damage



attributed to air pollution (Heck et al., 1983).  Ozone affects



the foliage of plants by biochemical and cellular alteration,



thus inhibiting photosynthesis and reducing plant growth, yield,



and quality.



     Early studies of ozone-related damages used generalized



relationships between ozone concentrations, yield, and economic



loss.  Insufficient information precluded the construction of  an



economic model with credible dose-yield  data.  Thus, for example,



Freeman  (1982), in a general survey of the literature, could only



conclude that the total agricultural damages from ozone ranged



from $1.0 to $4.0 billion in 1978 dollars.



     Recent work by the National Crop Loss Assessment Network



(NCLAN) suggests that prior studies have underestimated ozone-



related damages.  NCLAN'S estimated dose-yield functions for



soybeans, wheat, corn, peanuts, cotton,  barley, and sorghum have

-------
                            VI-23
TABLE VI-6.  Estimated Health Effects of a 1 Percent Reduction in
	Ozone (millions of days per year)	
Condition
  Low
Estimate
  High
Estimate
 Medium
Estimate
Respiratory Effects

  Bed Rest/Work Loss

    Adults
    Children
      Subtotal

  Minor Restrictions

    Adults
    Children
      Subtotal
Nonrespiratory Effects

  Headaches
  Eye Irritation
    Subtotal
  N.A.
  N.A.
  N.A.
  N.A.

-------
                              VI-24





provided more accurate information on ozone's effects on crops.



Kopp (1983, 1984) and Adams et al. (1984) incorporated these



functions into models of agricultural production and demand to



estimate the benefits of ozone reduction strategies.



     Kopp constructed a detailed microeconomic model of farm



behavior for over 200 producing regions in the United States.



The NCLAN dose-yield functions are directly incorporated in Kopp's



model of the supply side of each crop for each region.  Because



estimates of the demand and supply elasticities for these crops



are used in the analysis, it gives a good indication of the actual



change in economic welfare.  Kopp's simulations suggest that a 1



percent reduction in ozone would produce total benefits of roughly



$110 million (1983 dollars) per year for the seven major crops



covered by NCLAN, as shown in Table VI-7.  These seven crops



accounted for only about 80 percent of the total value of U.S.



crop production  (USDA, 1982).  If we increase Kopp's estimate by



assuming that ozone damages to all other crops occur in the same



proportion as their relative value, we conclude that the benefits



of a 1 percent change in ozone are roughly $137 million annually.



     Adams et al. (1984) used a different approach.  By



incorporating the NCLAN dose-yield functions into an existing



quadratic programming model, they calculated ozone benefits for



six of the seven crops covered by Kopp (they did not include



peanuts).  They estimated that a 10 percent reduction in rural



ozone would result in annual benefits of roughly $674 million



(1983 dollars).  Assuming linearity and increasing the estimate



to account for omitted crops, we estimated $90 million in benefits

-------
                            VI-25
TABLE VI-7
Crop
Annual Agricultural Benefits of a 1 Percent Ozone
Reduction (millions of 1983 dollars)	

                                     Estimate
Soybeans

Corn

Wheat

Cotton

Peanuts

Sorghum

Barley

  Total assessed
                                        50.8

                                         7.6

                                        21.1

                                        20.4

                                         5.0

                                         4.6

                                         0.2

                                       109.8

-------
                            VI-26


for a 1 percent change in ozone levels.  Unfortunately, Adams

et al. did not calculate the benefits on a crop-by-crop basis,

so it is not possible to make a detailed comparison with Kopp's

estimates.

     For our best estimates, we used Kopp (1983, 1984) and Adams

et al. (1984), because they used the superior NCLAN data and

based their estimates on economic measures of welfare loss.  The

two estimates are fairly close.  We concluded that a 1 percent

reduction in rural ozone would produce $90 to $140 million in

annual agricultural benefits per year, with a point estimate of

$114 million.  We qualify these numbers by noting that the esti-

mates do not reflect unreported small "truck farm" sales or any

averting activities that farmers may undertake, such as planting

pollutant resistant crops; we believe that these categories are

likely to be small.  On the other hand, the welfare measures do

not reflect either the effects of crop subsidy programs at the

state and federal levels, or the effects of drought, both of which

are likely to reduce the marginal welfare impacts of ozone.  The

effects of subsidies are likely to be particularly significant.


VI.B.I.e.  Ozone Effects on Nonagricultural Vegetation

     Forests and ornamental plants also may suffer substantial

damages from exposure to ozone.  The preliminary draft of the

Ozone Criteria Document discusses the issue:

     The influence of 03 on patterns of succession and
     competition and on individual tree health is causing
     significant forest change in portions of the temperate
     zone....  Long-term continual stress tends to decrease
     the total foliar cover of vegetation, decrease species
     richness and increase the concentrations of species

-------
                             VI-27


     dominance by favoring oxidant-tolerant species. These
     changes are occurring in forest regions with ozone
     levels (1-hour maximum) ranging from 0.05 ppm (111
     ug/m3) to 0.40 ppm (785 ug/m3) (U.S. EPA, 1983).


Additional evidence of significant damages from ozone associated

with nonagricultural vegetation is provided by McLaughlin et al.

(1984).

     Unfortunately, no careful quantitative studies of the type

done by NCLAN have been performed for nonagricultural vegetation.

Heintz et al. (1976) have estimated losses to ornamental plants of

$100 million per year in 1973 dollars.  Inflating to 1983 dollars

using the Farm Products Index, and assuming linearity, yields

estimated annual benefits of $1.4 million for the reduction in

damages to ornamentals from a 1 percent reduction in ozone.

     Damage to forests is potentially a much larger concern in

terms of reduced production and decreases in recreation and aethes-

tic values.  In a very small contingent valuation study, Crocker

and Vaux (1983) found that the shift of an acre of the current mix

of severely, moderately, and unharmed timberland in the San

Bernardino National Forest into the unharmed category would gener-

ate additional annual recreational benefits of between $21 and $68

per acre per year.  These findings are difficult to generalize for

the rest of the nation because ambient ozone levels are unusually

high in the San Bernardino area (and the authors provide no dose-

response function for extrapolating to areas with lower concentra-

tions), and because other site attributes and visitors' socioeco-

nomic characteristics have very large and significant effects on

users' willingness to pay to reduce damages to forests.

-------
                              VI-28





     We also lack data on the impact of ozone on commercial



forests.  Most commercial forests are located in areas with



relatively low ozone concentrations, so damages may be small.



Weyerhauser Corporation staff recently reported "negligible or



nonexistent" damages from ozone (personal communication from Jack



Larson of Weyerhauser).  In light of the NCLAN findings with



respect to crops, however, the possibility of significant ozone



damage to forests cannot be dismissed.





VI.B.l.d.  Ozone Materials Damage



     Ozone directly damages many types of organic materials,



including elastomers, paint, textile dyes, and fibers.  It can



increase the rigidity of rubber and synthetic polymers, causing



brittleness, cracking, and reduced elasticity.  Ozone exposure



also can generate other effects, such as avoidance costs (pur-



chase of specially resistant materials) and aesthetic losses.



Only the direct costs are incorporated in this analysis, however.



     In his survey of the literature, Freeman (1982) suggested



that annual materials damages from oxidants and NOX amount to



approximately $1.1 billion (1978 dollars).  We updated that esti-



mate using the U.S. Government price indices for rubber and tex-



tile products and the index of personal consumption expenditures



for durable goods; this yielded an estimate of $2.25 billion for



1983.  Assuming linearity, a 1 percent ozone reduction generates



a benefit of roughly $22.5 million annually.

-------
                              VI-29






     We obtained an alternative estimate of the benefits of



reduced material damage by using dose-response information



incorporated in the 1978 Criteria Document for ozone.  The text



contains per capita economic damage functions for elastomers,



textiles, industrial maintenance, and vinyl paint.  We used a



population-weighted mean value of ozone of 0.03 ppm (60 ug/m3)



from the draft of the new Criteria Document (U.S. EPA, 1984), a



population estimate of 230 million, and the indices cited above.



This method yielded annual benefits of $15 million (in 1983



dollars) for a 1 percent reduction in ozone.  Averaging these



two estimates (with slightly more weight given to the lower esti-



mate which does not include any NOx-related damages)  yields a



point estimate of $18 million annually, with a range of $15 to



$22.5 million.





VI.B.I.e.  Summary of Benefits of a 1 Percent Change in Ozone



     Table VI-8 summarizes our estimates of the effects of a 1



percent reduction in ozone.  The estimated health effects include



roughly 2.5 million days of respiratory symptoms resulting in some



restriction in activity, and 4.4 million days of non-respiratory



irritations (headaches and eye irritation).  On the nonhealth side



of the ledger, increases in agricultural crop production dominate,



with a total of $114 million.  Materials damage is next, with $18



million, while ornamental plants contribute $1.4 million.  In all



cases, the estimates are subject to considerable uncertainty; the



"high" and "low" estimates provide only a partial indication of



that uncertainty, as they reflect only the ranges in available

-------
                              VI-30
TABLE VI-8.  Summary of Estimated Effects of a 1 Percent
             Reduction in Ozone                      	
Effect
  Low       High     Medium
Estimate  Estimate  Estimate
Health (millions of cases)

  Respiratory Symptoms

    Bed rest/Work loss

    Minor restrictions
  0.36

  1.07
1.23

3.60
0.64

1.91
  Nonrespiratory Symptoms
  N.A.
N.A.
4.37
Other (millions of dollars)

  Agricultural crops

  Ornamental plants

  Materials damage


Unquantified Benefit Categories

  Chronic health effects
  Forest damage
  90.0     140.0     114.0

  N.A.      N.A.       1.4

  15.0      22.5      18.0

-------
                            VI-31






estimates or statistical uncertainty in the parameter estimates



from particular studies.  A more complete accounting for



uncertainty — which would include different functional forms



for the dose-response functions, omitted categories, etc. —



would yield substantially broader ranges.  The omitted categories,



in particular chronic health effects and damage to forests,



however, suggest that these estimates are too low, perhaps by a



substantial margin.





VLB.2.  Linking NOX and HC Reductions to Ozone Effects



     The previous section estimated the effects of reducing ozone.



Linking those effects to emissions of HC and NOX requires estimat-



ing the relationships between those pollutants and ozone concen-



trations.  Section VI.B.2.a describes the general process of ozone



formation, while Section VI.B.2.b presents rough quantitative esti-



mates of the impact of HC and NOX emissions on ozone concentrations



in both urban and rural areas.





VLB.2.a.  The Process of Ozone Formation



     Ozone changes are influenced by the amount of solar



radiation and the ratio of, and changes in, the concentrations of



NOX and HC.  The general process by which ozone (03) is formed is



illustrated by the N02-NO-03 cycle.  These reaction equations can



be written as:






(VI-1)    NC-2 + sunlight —> NO  + O



(VI-2)    0 + 02  	'—> 03



(VI-3)    03 + NO	> N02 + 02

-------
                            VI- 3 2


     Ozone is not emitted in any measurable quantity.  The first

two reactions represent its only significant source.  However,

these reactions are fully reversed by the third one.  Because NO

comprises roughly 90 percent of man-made NOX emissions, reaction

(VI-3) suggests that additional NOX emissions would reduce ozone

concentrations by scavenging the ozone for one of the oxygen

atoms, thereby producing ©2 and N02 .   Therefore, increases in NOX

cannot increase ozone, unless some other species oxidizes NO to

NO2 without scavenging ozone for the  necessary oxygen atom.  This

additional reaction is:
(VI-4)  Organic radicals (from HC)  + NO — >  N©2 + miscellaneous
                                                     products


     With reactions (VI-1), (VI-2), and (VI-4), NO2 is recycled,

allowing for the generation of excess 03.  That is, HC permits

ozone formation while bypassing the ozone destructive step of NO

+ 03 — > NO2 +02.  In short, with a plentiful supply of HC , NOX

will not have to scavenge ozone to form NO2 .   Instead, NOX com-

bines with the HC to form additional NO2 , which in turn makes

more ozone.  Therefore, in areas with "excess" HC, NOX control

will reduce ozone, but controlling HC will be relatively ineffec-

tive.  However, when there is not enough HC to react with avail-

able NOX, NOX will scavenge ozone to form NO2 •  In this situation,

NOX control will result in less ozone scavenging (and more ozone).

HC control, on the other hand, will be effective in reducing

ozone because the greater the reduction in HC, the more NOV is
                                                          X

forced to scavenge ozone.

-------
                              VI-33





     Atmospheric scientists have shown that the HC/NOX ratio is



relatively small in metropolitan areas.  Thus, in such areas, a



reduction in HC will reduce ozone, but depending on the specific



HC/NOX ratio, controlling NOX may increase ozone.  In rural areas,



HC/NOX ratios are larger.  Therefore, if the ozone in such areas



results primarily from local emissions, controlling local NOX



emissions may be a more effective way of reducing ozone than local



HC controls.






VI.B.2.b. Quantitative Estimates of Impacts of HC and NOX on Ozone



     Several models have been developed to simulate the effect of



changes in HC and NOX on ozone (U.S. EPA, 1983f, 1984b; Systems



Applications, Inc.  [SAI], 1984).  EPA's Office of Air Quality



Planning and Standards (OAQPS) also has developed three sets of



regional Empirical Kinetic Modeling Approach (EKMA) isopleths for



use in the ozone NAAQS review.  Finally, the data base obtained



from state implementation plans (SIPs) for ozone summarizes all of



the modeling results from the SIPs.  A review of these models and



their general results is provided in McGartland and Ostro (1985).



From these models, we generated estimates of the "average" impact



of changes in HC and NOX on urban and rural peak ozone concentra-



tions (relevant for health effects and materials damage) and on



rural average ozone concentrations (more relevant for agricultural



effects), measured  in terms of elasticities (i.e., the percent



change in the ozone variable with respect to a 1 percent change



in emissions of the pollutant).  The results are summarized in



Table VI-9; the discussion of their derivation follows.

-------
                            VI-34
TABLE VI-9.  Estimated Ozone Reductions from 1 Percent Reduction  in
	Rural and Metropolitan HC and NOy (percent reduction)
                            Reduction
                             in Peak
                           Urban Ozone
                                         Reduction
                                          in Peak
                                         Rural Ozone
                           Reduction
                           in Average
                          Rural Ozone
HC
 1 Percent Reduction
   in Urban HC
                            0.60
            0.05
                                    0.13
 1 Percent Reduction
   in Rural HC
                               0
            0.36d
                                    0.36d,e
 1 Percent Reduction
   in Urban
 1 Percent Reduction
   in Rural N0«
-0.3 to -0.05
0
                                        -0.1 to +0.1
                                           (0.04)b
                                       0.05 to 0.2d
                                           (0.15)c
                            -0.09
                                0.05 to 0.2d,e
                                   (0.15)c
a Estimates varied by a fairly wide range.
  estimate of -0.1.
                                          We used the point
Sparse evidence does not permit the estimation of a single
point estimate with a reasonable degree of certainty.  In many
areas (e.g., the Northeast Corridor), wind trajectories would
probably transport metropolitan ozone to other urban centers
or over the ocean.  In these cases, rural ozone would not
change.  In other areas, however, the increased ozone from
urban NOx reductions would travel to rural areas, but lower
rural N©2 levels would reduce ozone.  Therefore, the overall
effect could be positive or negative.  As a conservative estimate,
we assumed that a 1 percent uniform reduction in metropolitan
    reduces rural ozone by 0.04 percent.

                   We used a conservative high point estimate
c Estimates varied.
  of 0.15.
d Results from ongoing analysis at SAI will soon narrow the
  range of estimates.

e The lack of evidence made it necessary to assume that average
  rural ozone changes by the same percent as peak rural ozone.

-------
                              VI-35





     The ozone-SIP data base presents estimates of how urban HC



affects peak urban ozone for every city not attaining the ozone



standard.  Analysis of these data by OAQPS (U.S. EPA, 1984)



suggests that a 1 percent reduction in hydrocarbon emissions will



reduce metropolitan ozone on average by 0.6 percent.  This estimate



is based on many EKMA urban simulations throughout the U.S.; as



such, it represents the best available estimate available of the



percent reduction in urban ozone from a 1 percent reduction in



HC.  However, there is considerable variance from area to area.



     Estimates of how urban HC affects rural average and peak



ozone levels are presented by SAI (1984).  SAI simulated regional



air quality under four basic scenarios — mornings and afternoons,



winter and summer.  The two summer scenarios (relevant for esti-



mating agricultural benefits) indicate that average ozone changes



by 0.13 percent for every 1 percent change in urban HC during the



agricultural growing season.  Peak rural ozone changed by much



less:  roughly 0.05 percent for every 1 percent change in urban



HC.  This is mainly because the HC and ozone plumes are widely



dispersed by the time they reach rural areas.



     Estimates of how rural HC affects rural ozone generally were



not available.  Therefore, we used the three sets of regional EKMA



isopleths developed by OAQPS.  By estimating data points at HC/NOX



ratios typical of rural areas, we estimated that, on average, a



1 percent reduction in rural HC will reduce peak rural ozone by



0.36 percent.  EKMA models are not, however, suitable for estimat-



ing changes in average ozone.  As our best estimate, we assumed



that average ozone changes by the same percentage as the peak.

-------
                              VI-36






For rural areas this may be a reasonable approximation, as rural




ozone levels do not fluctuate as much as urban ozone levels and



thhe plumes would still be highly concentrated in the same area.



     To estimate the effects of NOX reductions on urban and rural



ozone, we used SAI (1984) and the EKMA isopleths developed by



OAQPS.  By taking an average of data points from the isopleths, we



calculated that, on average, a 1 percent reduction in metropolitan




NOX will increase ozone by 0.1 percent.



     SAI estimated how rural ozone is affected by changes in NOX



emissions.  A 1 percent reduction in urban NOX increases average




rural ozone by 0.09 percent.  Peak rural ozone would probably not



change as much because the urban plume is widely dispersed when



it reaches the rural areas.  The SAI analysis showed average rural



ozone increasing by a small amount in some situations, but, on



average, there probably is a small decrease in rural ozone given a



reduction in urban NOX emissions.  As a conservative estimate, we



assumed that a 1 percent reduction in urban NOX reduces peak rural



ozone by 0.04 percent.



     We again relied on the EKMA isopleths generated by OAQPS to



estimate how rural NOX emissions affect peak rural ozone.  On



average, the isopleths indicate that a 1 percent reduction in



rural NOX will reduce peak rural ozone by 0.1 to 0.2 percent.  We



used an estimate of a 0.15 percent decrease in peak ozone for




every 1 percent decrease in rural NOX.  We also used that estimate



to quantify how average ozone reacts to changes in rural NOX.




(The rural plumes would still be concentrated in a small area.)

-------
                            VI-37


     Four caveats should be considered in connection with the

results summarized in Table VI-9:

     1.   Many of the results are based on averages of
          different simulation results.

     2.   Our estimates represent how ozone is affected on
          average across the U.S.  They should not be used
          for estimating ozone changes for a specific area.

     3.   Most of the literature reports estimates of changes
          in peak ozone.  Peak ozone is the appropriate measure
          for most of our benefit categories, but for one
          important benefit category (agriculture), changes in
          daily maximum 7-hour average ozone levels are needed.
          For rural emission changes, we assumed that the mea-
          sures of average ozone move in direct proportion to
          changes in peak ozone.

     4.   Some of our estimates of how rural emissions affect
          ozone were derived from models better suited to
          represent urban areas (e.g., EKMA).

Because of these qualifications, the estimates in Table VI-9

should be regarded as highly uncertain.  The estimates for NOX

are particularly uncertain, because much less modeling has been

done of the NOx-ozone relationship than for HC.

     We next calculated the population-weighted effects on ozone

of 1 percent changes in national emissions of HC and NOX.  (We

used population as the weighting factor because the nonagricul-

tural benefits are proportional to it.)  We calculated weighted

average elasticities using the estimates in Table VI-9 and the

fact that about two-thirds of the non-agricultural benefits are

attributed to metropolitan area and one-third to non-metropolitan

areas (based on relative populations).  For hydrocarbons, the

estimates in Table VI-9 suggest that a 1 percent reduction in

emissions would reduce metropolitan ozone by 0.60 percent

(= 0.60 + 0); similarly, non-metropolitan ozone would fall by

-------
                            VI-38






0.41 percent (= 0.05 + 0.36).  Weighting these two changes by



the split in non-agricultural benefits between metropolitan and



non-metropolitan areas yields an overall benefit-weighted elasticity




of:






(VI-5)          0.67 (0.60) + 0.33 (0.41)  = 0.537






Multiplying this weighted elasticity times the national effects



of a 1 percent change in ozone yields the same result as



computing separately, and then summing, the metropolitan and



non-metropolitan effects of 1 percent changes in hydrocarbons.



Similarly, for NOX, the estimated benefit-weighted average



elasticity for non-agricultural benefits was:






(VI-6)    0.67 (-0.1 + 0) + 0.33 (0.04 + 0.15) = -0.004






Note that reductions in NOX are predicted, on average, to



cause a slight increase in peak ozone levels.  Using the full



range of elasticity estimates in Table VI-9, however, this



weighted elasticity ranges from -0.22 to +0.066.



     For agricultural crops, we assumed that the relevant measure



would be the change in average rural ozone levels; we gave no



weight to changes in urban ozone.  For HC, the estimated elasticity



of average rural ozone with respect to national emissions was:






(VI-7)               0.13 + 0.36 = 0.49






Thus, we estimate that 1 percent reductions in metropolitan and



non-metropolitan HC emissions reduce average rural ozone by 0.49



percent.  For the NOX point estimates, the elasticity was:

-------
                              VI-39





(VI-8)               -0.09 + 0.15 = 0.06





Note that this estimate suggests that nationwide NOX reductions



will slightly decrease rural ozone levels, despite the fact that



urban NOX control, on average, is predicted to increase ozone;



the increase in urban ozone that is transported to rural areas



does not offset the reduction in ozone produced in rural areas.



Again, however, the range of uncertainty is large.  Using the



ranges in Table V-9 generates elasticity estimates for the impact



of national NOX control on average rural ozone ranging from -0.04



to +0.11.





VI.B.2.C.  Ozone-Related Effects Per Ton of HC and NOX Controlled



     We used the estimated weighted average elasticities, together



with the estimates of reductions in HC and NOX emissions to predict



changes in ozone.  We needed to predict changes in two ozone



measures:  average rural ozone (for the agricultural benefits)



and changes in peak ozone levels, averaged over the nation (for



the other ozone benefit categories).



     To apply the elasticities, we needed to convert the emission



reduction to percentages; because our two largest ozone-related



benefit categories (agriculture and restricted activity days) were



based on 1978 and 1979 air-quality data, we computed the changes



as percentages of 1979 emissions, which were about 21.9 million



tons for HC and 21.3 million tons for NOX.  Thus, the predicted



emissions due to misfueling in 1986 (from Table VI-3) translate



to 1.4 percent (= 305,000/21.9 million tons x 100%) and 0.44



percent (= 94,000/21.3 million x 100%) for NOX.  Using these

-------
                              VI-40






percentages together with the estimated elasticities for peak



ozone (equations VI-5 and VI-6) yields the following estimated



percentage change in peak ozone due to emission reductions in




1986:






(VI-9)     (0.537)(1.4%) + (-0.004)(0.44%) = 0.75 percent.






Similarly, for average rural ozone, the estimated decline in



ozone due to emission reductions in 1986 is:






(VI-10)    (0.49)(1.4%) + (0.60X0.44%) = 0.71 percent.






Table VI-10 presents the year-by-year estimates of changes in



ozone due to eliminating misfueling.



     Finally, the predicted changes in the two ozone measures



were combined with the estimates in Table VI-8 to predict the



year-by-year ozone-related effects of the rule.  The changes



in average rural ozone were used for agricultural crops, while



the changes in peak ozone were used for the other benefit



categories.  Table VI-11 presents the results for 1986.  It is



important to remember that these estimated effects are subject



to all of the uncertainties and omissions discussed earlier



in connection with our estimates of the benefits of a 1 percent



reduction in ozone.  In addition, however, they also reflect



the uncertainties in the estimates relating HC and NOX emissions



to ozone; these uncertainties are particularly great in the



case of NOX.

-------
                                VI-41


TABLE VI-10.  Estimated Ozone Reductions due to Eliminating
	Misfueling (percent of 1979 level)	

Ozone
Measure	1985   1986   1987    1988    1989    1990   1991   1992

National
peak         0.38   0.75   0.74    0.74    0.74    0.75   0.78   0.81

Rural
average      0.36   0.71   0.71    0.71    0.71    0.73   0.76   0.78

-------
                                VI-42
TABLE VI-11.  Quantified Ozone-Related Effects Due to Elimination
              of Misfueling in 1986
Effect
Health (millions of cases)
Respiratory
Bed rest/Lost work
Minor restrictions
Non respiratory
Other (millions of dollars)
Agricultural crops
Ornamental plants
Materials damage
LOW
Estimate

0.27
0.80
N.A.
63.80
N.A.
11.19
High
Estimate

0.92
2.69
N.A.
99.25
N.A.
16.79
Medium
Estimate

0.48
1.43
3.26
80.81
1.04
13.43
Unquantified Benefit Categories

  Chronic health effects
  Forest damage

-------
                              VI-43

VI.C.  Health and Welfare Effects Not Related to Ozone
     Thus far, our analysis has focused only on HC and NOX, and
only on their ozone-related effects.  In this section, we discuss
other health and welfare effects associated with those pollutants
and with CO and EDB.  In most cases, the discussion is only quali-
tative because we have been unable to make quantitative estimates.

VI.C.I. Hydrocarbons
     In addition to the benefits associated with reduced ozone,
reductions in hydrocarbon emissions may affect sulfate concentra-
tions and will reduce exposure to benzene.  Although quantitative
estimates are not possible for the former, we can make some rough
estimates of the health benefits of reduced benzene exposure.

VI.C.I.a.  Impact on Sulfates
     Hydrocarbons are a factor in the formation of sulfates.
In particular, S02 oxidizes faster when the amount of hydroxide
radicals in the atmosphere increases (which is, in turn, a function
of the amount of HC in the atmosphere).  However, the ability  to
quantify these complex relationships has just been developed,  and
experts at SAI and EPA's Office of Research and Development believe
that the total change in sulfates is highly dependent upon many
factors (e.g., cloud cover, current HC and NOX concentrations, and
oxidant and sulfur dioxide levels) for which we have only  limited
data.
      A recent modeling analysis by SAI  (1984) indicated  that  a
10 percent reduction in HC could reduce  sulfates  in urban  areas
during certain times of the year by about 2 percent.  However,

-------
                            VI-44






because of the uncertainty surrounding this estimate, and the



uncertainty in interpolating this to much smaller changes in HC,



we did not try to quantify the reduction in sulfates in this




analysis.






Vl.C.l.b.   Impact on Benzene and Other Aromatics



     Reducing lead in gasoline will affect benzene emissions in



two ways.   First, because benzene is a hydrocarbon, reducing mis-



fueling will reduce benzene tailpipe emissions along with other



hydrocarbons.  Second, to increase octane with less lead, refin-



eries will increase the severity of their catalytic reforming,



which in turn increases the amount of benzene in reformate, one



of the components blended into gasoline.  Benzene emissions are



of particular concern because benzene is believed to be a leukem-



ogen, and has been listed by EPA as a hazardous air pollutant



under Section 112 of the Clean Air Act.



     Benzene comprises about 4 percent of tailpipe hydrocarbon



emissions (U.S. EPA, 1983c).  Assuming that catalysts are as



effective in eliminating benzene as other hydrocarbons, that



would imply that benzene would be roughly 4 percent of the reduced



HC emissions from eliminating misfueling.  Thus, for example,



in 1986, we estimate that eliminating misfueling would eliminate



0.04 (305,000) = 12,200 tons of benzene.




     Reducing lead is likely to increase the overall benzene



content of gasoline because additional reforming creates more



benzene in the reformate.  Benzene (along with all other aroma-



tics) has a a poor octane response to lead, however, so as the

-------
                              VI-45





aromatic content of reformats increases it will be more economical



for refiners to increasingly direct such stocks to unleaded gaso-



line, and use less reformate and more of other higher lead-response



components (e.g., alkylate) in leaded gasoline.  Based on the DOE



refining model, which finds the blending pattern that gives the



least cost, we estimated that the benzene content of gasoline will



rise from about 1.5 percent to 1.6 percent in unleaded, but will



fall from 0.62 percent to 0.38 percent in leaded gasoline (at



0.10 gplg).



     Raising the overall benzene content of the gasoline pool will



increase evaporative emissions of benzene from gasoline marketing



facilities (service stations, etc.), but based on EPA analyses of



the impacts of regulating gasoline marketing (U.S. EPA, 1984a), we



estimated the increase to be less than 400 tons per year.  Moreover,



because of the shift of benzene from leaded to unleaded gasoline,



evaporative emissions from vehicles should fall by about the



same amount, because unleaded gasoline is used in vehicles with



evaporative control devices.  In addition, switching benzene



from leaded to unleaded should reduce tailpipe emissions of



benzene because unleaded gasoline is used in vehicles with cata-



lytic converters, which oxidize about 90 percent of the unburned



hydrocarbons in the exhaust.  In 1986, this should reduce benzene



tailpipe emissions by almost 7,000 tons.



     Table VI-12 summarizes our estimates of reductions in



emissions of benzene for 1986, assuming that misfueling is elimi-



nated.  They total about 18,000 tons, of which about two-thirds



are due to the elimination of misfueling and one-third are due

-------
                            VI-46
TABLE VI-12.   Reductions in Benzene Emissions in 1986, Assuming
	No Misfueling (tons)	

Sources of Emissions	Amount	

Reduced Misfueling

  Tailpipe emissions                          12,200

Changes in fraction of benzene
  in leaded and unleaded gasoline

  Marketing evaporative                         -366

  Vehicle evaporative                            377

  Tailpipe                                     6,230

TOTAL                                         18,441

-------
                            VI-47





to reduced tailpipe emissions resulting from the change in the



relative benzene contents of leaded and unleaded gasoline.



Evaporative emissions account, on net, for less than 0.1 percent



of the change.



     EPA1s Carcinogen Assessment Group (CAG), using a "plausible



upper bound" unit-risk estimate, estimated that in 1976, automobile



emissions of benzene resulted in up to 50.9 deaths from leukemia



(U.S. EPA, 1979a).  Since then, however, the CAG has revised its



unit-risk estimate for benzene downwards by slightly more than 8



percent, yielding an estimate of 47.3 leukemia cases for 1976.



The CAG estimate was based on an emissions estimate of 202,000 tons



of benzene.  Scaling down the CAG risk estimate of 47.3 cases by



emissions yields 4.4 (= 47.3 x 18,441/202,000) cases of leukemia



eliminated in 198fi.



     We also have performed similar calculations for emissions of



total aromatics, with similar results.  Reducing lead in gasoline



should reduce overall emissions of aromatics, both because of



reduced misfueling and because of shifts in the relative fractions



of aromatics in leaded and unleaded gasolines.  On net, emissions



of aromatics should fall by several tens of thousands of tons per



year (Schwartz, 1984, in Docket EN-84-05) as a result of reducing



lead in gasoline to 0.10 gplg.  We have not included any benefits



for these reductions, however, except as they relate to benzene




specifically.





VI.C.2.  Nitrogen Oxides



     Besides the ozone-related effects discussed above, nitrogen




oxides also may affect health and welfare directly.  NOX emissions

-------
                            VI-48






are believed to affect health and materials, to contribute to




reductions in visibility, and are associated with acid deposition.



In addition, damage to vegetation has been demonstrated experiment-




ally.  Unfortunately, specific dose-response information related



to NOX is sparse.  As a consequence, there is great uncertainty in



the benefit estimates and many effects cannot be quantified.  How-



ever, a few of the benefit calculations are presented to provide



a partial measure of the effects of NOX emissions on economic



welfare.  For health effects, only qualitative evidence is provided.



While there may be acid rain benefits as well, we have not included



them because of uncertainties about the role of NOX in acid



deposition.



     As discussed earlier, NOX represents the composite formula for



NO (nitric oxide) and NC>2 (nitrogen dioxide).  NO is the dominant



oxide released initially, but atmospheric interactions result in



the conversion of NO to NO2-  Based on the results of smog chamber



tests and modeling experiments, Trijonis (1978, 1979) concluded



that maximum and average NO2 concentrations tend to be proportional



to initial NOX contributions.






VI.C.2.a.  Visibility Benefits from Reduced NOV



     NO2 is a reddish-brown gas that reduces visibility by



absorbing and discoloring light.  In contrast, particulate




matter scatters light to reduce visual range.  While particulate



matter accounts for almost all of the damage to visibility in




the East, for some western regions NO2 may play a significant



role in determining visual range.

-------
                            VI-49






     To bound the value of improved visibility per ton of NOX



reduced, we used the results of Brookshire et al. (1976).  Their



contingent valuation study showed that recreators were willing to



pay $1.2 million per year (annualized) to avoid visibility re-



ductions that would result from a planned Kaiparowits power plant



in southern Utah.  Similar studies, by Randall et al. (1974) and



by Blank et al. (1977) produced comparable results for the



value of visibility benefits in the Four Corners region of the



Southwest.  However, these estimates apply only to this particular



region, known for its scenic vistas.



     In the Kaiparowits study, climatic conditions, emission



controls, and other factors allowed the investigators to assume



that the major visibility-related impact would be the coloration



of the sky by NO2-  Given the projected power plant emissions of



80,000 tons of NOX per year, the estimate of $1.2 million of



potential damage translates to about $2.1 million in 1983 dollars



(based on Consumer Price index).  For 1986, we project a 94,000



ton reduction in NOX emissions if misfueling is eliminated.  Using



the Brookshire et al. estimate and assuming linearity yields a



benefit estimte of about $2.4 million.  We stress that this



figure is probably an upper bound, even for sensitive regions.



Although the Kaiparowits area is not densely populated, a large



number of recreators use the site and they typically place a



high value on protecting clean areas.  Further, it is doubtful



that NC>2 has a noticeable impact on eastern visibility, where

-------
                            VI-50


range is limited by buildings and largely influenced by particulate

matter (based on a conversation with Shep Burton of Systems

Applications, Inc.).


VI.C.2.b.  Health Benefits of Reducing NOy

     Evidence of health effects related to NOX is provided in the

Air Quality Criteria for Oxides of Nitrogen (EPA, 1982b), the OAQPS

staff paper on the NAAQS for nitrogen oxides, the Clean Air

Science Advisory Committee's (CASAC's) cover letter (EPA, 1982c),

the NOX Regulatory Impact Analysis (EPA, 1982e), and the published

literature.  These studies can be divided into four classes:

     (1)  Animal toxicology studies.  Animals are exposed to
          controlled levels of NC>2.  Researchers have the option
          of using invasive techniques to investigate the effects
          of N02.

     (2)  Controlled human-exposure studies.  These are clinical
          studies in which humans are exposed to NO2 in enclosed
          chambers.  They are limited typically to examining the
          effects of a single, short-term (acute) exposure.

     (3)  Outdoor epidemiological studies.  Health indicators
          of cross-sectional groups are statistically related
          to real-world outdoor ambient concentrations.  This
          class of studies is generally most appropriate to
          assess the benefits of controlling outdoor air pollu-
          tion, since health effects are related directly to
          the control variable of interest.

     (4)  Indoor epidemiological studies.  Health measures of
          cross-sectional groups are statistically related to
          indicators of indoor pollutant concentrations.  For
          example, the "gas stove" studies investigate the
          effect of indoor air pollution on individuals living
          in homes with gas stoves (a significant source of NC«2)f
          compared with people living in homes with electric
          stoves.

     Most animal toxicology studies involving NC>2 emphasize peak

exposure, with concentrations (2 ppm to 20 ppm) roughly 40 to 400

times the annual average ambient N02 standard of 0.05 ppm.  An

-------
                            VI-51





additional limitation of these studies in assessing benefits is



that there is no generally accepted method for extrapolating expo-



sure response results from animal studies to humans.  Thus, these



studies may lend support to the human studies, but are of little



help in quantifying health effects.



     The majority of the controlled human experiments have



examined the effects of NC>2 on healthy adults by exposing them



to single, short-term concentrations in enclosed chambers.  A



very limited number of clinical studies also have examined some



potentially sensitive populations (asthmatics and chronic bron-



chitics), although others (children) have yet to be tested.



     In general, these studies, as summarized in the OAQPS staff



paper, indicate that healthy adults are not affected by concen-



trations of 1 ppm or less.  When considering the effect on more



sensitive individuals, small reductions in pulmonary function may



appear in the range of 0.5 to 1.5 ppm.  Also, some studies have



shown increased sensitivity to agents inducing bronchoconstriction



at 0.1 to 0.2 ppm levels of N02-  Unfortunately, dose-response



functions relating NOX to economically-significant health endpoints



do not exist.



     Successful outdoor epidemiological studies are scarce because



investigators must separate many confounding effects and health



hazards.  Moreover, some of the studies used in setting the exist-



ing annual standard of 0.053 ppm (100 ug/m3) have been criticized




because of NC>2 measurement problems.  Therefore, this class of



studies does not provide any adequate basis for estimating a



dose-response function.

-------
                              VI-52






     Indoor epidemiological studies have focused on comparisons




of homes with gas stoves (which emit NOX)  to those with electric



stoves.  Some of these studies reported an association between



nitrogen dioxide and either lung function changes or respiratory



effects (EPA, 1982), but the most recent studies, including Ware



et al. (1984), no longer find a statistically significant rela-



tionship between children living in homes with gas stoves and



the incidence of respiratory illness.  Ware et al., however,



continue to find small statistically-significant decreases in



in pulmonary function associated with NOX, although the estimates



may be confounded by the influence of parental education.



     Because of the mixed results concerning respiratory symptoms



and the lack of a dose-response estimate for pulmonary function



changes, no health benefits are quantified for the anticipated



reductions in NOX emissions.






VI.C.2.C.  NOy Effects on Vegetation



     Data concerning the effects of NOX on plant growth and yield



also are limited.  Nevertheless, it it reasonable to assume that



NOx-induced reductions in the assimilative capacity of plants



through altered metabolism, leaf injury, or abscission affect



plant growth.  MacLean (1975), however, concluded that average



NC>2 concentrations are well below the threshold curve for damage




to growth.  In fact, the maximum NC>2 concentrations recorded in



Los Angeles for 1966 would just begin to damage growth.  Even in



Los Angeles, however (the only major city in the U.S. that exceeds



the NO2-NAAQS), average NO2 concentrations are below the likely




threshold when averaged over longer periods.

-------
                            VI-53





     Although NC>2 by itself is unlikely to damage plants at



existing outdoor levels, several studies have demonstrated syner-



gistic and antagonistic effects.  The Criteria Document (EPA,



1982) concluded that "concentrations of N02 between 0.1 ppm and



0.25 ppm can cause direct effects on vegetation in combination



with certain other pollutants" (pp. 12-44).  But these data



indicate that rural areas are still not at risk, because, in



general, concentrations will not rise above 0.1 ppm for suffi-



cient time for damages to occur.  Therefore, we limit our estimate



of benefits to account for vegetation damages in urban areas.



     The plants most in danger would be ornamental vegetation,



but even these damages are likely to be small.  Leighton et al.



(1983) concluded from their review that "ozone appears to account



for more than 90 percent of total vegetation damages" (p. 60;



see also Heck et al., 1982, and Page et al., 1982).  If NOX



accounted for the other 10 percent of the total ornamental



vegetation damage, as an upper-bound we could use the high estimate



of ozone ornamental vegetation damages of $140 million to impute



total NOX damages of roughly $16 million.  For the 0.44 percent



reduction in NOX predicted for 1986, that would imply benefits



of only $62,000.





VI.C.2.d.  NOy Effects on Materials



     Field studies and laboratory research have demonstrated



that nitrogen oxides can significantly fade textile dyes.



Barrett and Waddell (1973) estimated damages at $280 million



(inflated to 1983 dollars using the Textile Products and

-------
                           VI-54






Apparel Index) for NOX damage in this category.  The basis for



the estimates included not only the reduced wear life of textiles



of moderate fastness due to NOX, but also the costs of research



and quality control.  The major share of the costs is the extra



expense of using dyes of higher NOX resistance and of using



inhibitors.  Additional costs are also incurred for dye application



and increased labor expenditures.  The factors relating higher



costs in the textile industry to NOX are discussed in Chapter 8



of a report of the National Academy of Sciences (1976).



     For 1986, assuming linearity, the Barret and Waddell



estimate implies that materials benefits related to NOX control



would be about $1.2 million (0.44% x $280 million).






VI.C.2.e.  Acid Deposition Benefits



     Acid deposition occurs when NOX and S02 emissions are



chemically altered into acids in the atmosphere and transported



over long distances, or when the precursor emissions are acidified



after being deposited in dry form on plant, soil, or building sur-



faces.  Both wet and dry forms of acidic deposition are harmful to



aquatic, terrestrial, and material resources.  While these damages



are potentially important in some regions, basic scientific under-



standing of the effects on resources and transport processes is



quite limited.




     The computation of damages requires several crucial pieces



of information.  Unfortunately, many of these data are so uncer-




tain that confidence intervals of damage estimates are exception-



ally wide.  Dollar estimates of damage require concentration-




response functions and a tabulation of the resources at risk.

-------
                              VI-55





Reliable concentration-response data are lacking in nearly all



resource areas, especially forestry and materials, and inventory



data are unavailable for most material resources.  Finally, the



benefit estimate requires information on the relative contribution



of nitrates and sulfates to total damages, as well as the rela-



tionship between the tonnage of NOX emissions and nitrate depo-



sition.  All of these are the subjects of intensive ongoing



research.  Until more of this research is completed, quantitative



damage estimates will not be accurate enough to be useful in a



policy context.



     Damage from acid deposition may be significant, although



there is much uncertainty.  Nitrates account for roughly 30



percent of total acidic loadings, but the damage information



available indicates that the proportion of nitrate to total



damages may be less than that implied by the relative emittant



loadings.





VI. C.3.  Carbon Monoxide



     At current ambient concentrations, exposure to CO may cause



health effects in some individuals.  Persons with cardiovascular



disease appear to be at highest risk, but those with chronic



respiratory disease, pregnant women, and the elderly also are



believed to be sensitive to CO exposure.  Unfortunately, clinical



dose-response functions relating low-level CO exposure to parti-



cular health effects, when estimated, have not been conclusive.



Therefore, it is not possible to estimate the impact of reduced



CO emissions on health endpoints.  We have, however, been able

-------
                              VI-56






to estimate roughly the impact of reducing CO emissions on the



numbers of people with carboxyhemoglobin (COHb)  levels that may




pose some risk.






VI.C.3.a.  Health Effects of CO



     Probably the greatest concern about CO exposure is its effect



on the cardiovascular system.  At moderate levels of exposure, CO



reduces exercise time before the onset of angina pectoris.  This



clinical phenomenon is believed to result from insufficient oxygen



supply to the heart muscle, and is characterized by spasmodic chest



pain, usually precipitated by increased activity or stress, and is



relieved by rest.  Typically, atherosclerosis, which causes a



narrowing of the arteries in the heart (coronary heart disease),



predisposes a person to attacks of angina.



     Angina pectoris, by definition, is not associated with



permanent anatomical damage to the heart.  Nonetheless, the dis-



comfort and pain of angina can be severe, and each episode of



angina may carry some risk of a myocardial infarction.  However,



epidemiological studies have not provided conclusive results on



the association between CO exposure and the incidence of myocard-



ial infarction.




     The health effects from exposure to CO are associated with



the percentage of total blood hemoglobin that is bound with CO,



producing COHb, which reduces the oxygen-carrying capacity of the



blood.  The median concentrations of COHb in blood are about 0.7



percent for nonsmokers and about 4 percent for smokers.  At least



one clinical study (Anderson et al., 1973) associated reduced

-------
                              VI-57


exercise time until the onset of pain with COHb levels of 2.9

percent in patients with angina pectoris.  At 4.5 percent COHb,

this same study reported an increased duration of angina attacks.

     The potential health improvements from reduced CO may be

significant, for two reasons.  First, there are many people in

the population believed to be sensitive.  EPA has estimated that

5 percent of the U.S. adult population — roughly 11.5 million

people — has definite or suspected coronary heart disease.  Of

this group, as many as 80 percent have suspected or definite

angina pectoris (U.S. EPA, 1980).  Additional large subgroups of

the population may be particularly sensitive to exposure to CO,

including individuals with pre-existing conditions that

compromise oxygen delivery to various tissues, enhance oxygen

need, or elevate the sensitivity of tissues to any oxygen

imbalance.  Other sensitive groups may include:

     0  people with peripheral vascular diseases, such
        as atherosclerosis and intermittent claudication
        (0.7 million people);

     0  people with chronic obstructive pulmonary
        diseases (17 million people);

     "  people with anemia or abnormal hemoglobin types
        that affect the oxygen-carrying capacity of the
        blood (0.1245 million people);

     0  people drinking alcohol or taking certain
        medications (e.g., vasoconstrictors);

     0  the elderly;

     0  residents of and visitors to high altitude areas;
        and

-------
                            VI-58
     0  fetuses and infants (3.7 million total live births
        per year).*

     The second reason for potentially significant impacts of

reduced CO exposure is that the blood of many people shows

concentrations of COHb above 2.9 percent, the lowest level of

COHb where adverse effects are believed to occur.  Data from the

second National Health and Nutrition Examination Survey (NHANES

II) indicated that for the U.S. population over twelve years of

age, 2 percent of those who have never smoked, 3 percent of

former smokers, and 66 percent of current smokers exceeded 2.9

percent COHb at the time of the survey (U.S.  DHHS, 1982).

     Other health effects have been reported  at comparable or

higher COHb levels.  For example, several investigators have

found statistically significant decreases in  work time until

exhaustion in healthy young men with COHb levels at 2.3 to 4.3

percent (Horvath et al., 1975; Drinkwater et  al., 1974; Raven

et al., 1974).  At higher COHb levels (5.0 to 7.6 percent and

above), investigators have reported impairment in visual percep-

tion, manual dexterity, ability to learn, and performance of

complex sensorimotor tasks in healthy subjects.  Finally, Klein

et al. (1980) showed that, at 5.0 to 5.5 percent COHb, healthy

young men had decreased maximal oxygen consumption and decreased

time at strenuous exercise before exhaustion.


* Animal studies showed that pregnant females exposed to CO
  had lower birth weights and increased newborn mortality, and
  their newborns had lower behavioral levels, even when no
  effects on the mother were detected.  In addition, a possible
  association has been reported between elevated CO levels and
  Sudden Infant Death Syndrome (Hoppenbrouwers et al., 1981.)

-------
                            VI-59






Vl.C.S.b.  Change in Numbers of People Above 2.9 Percent COHb



     For three specific subgroups of the U.S. population —




current smokers, ex-smokers, and never-smokers — we calculated



the number of people who would move below 2.9 percent COHb (the



lowest level frequently associated with specific health effects)



because of the reduction in CO emissions.  We also calculated the



number of "sensitive" individuals who would shift below 2.9



percent COHb.  This 2.9 percent level should not be construed as



a "threshold" level, but rather as a level at which some increased



health risk has been detected.



     We considered several complex relationships to make these



rough estimates.  Because of resource and data limitations, we



assumed simple one-to-one relationships in some of the linkages



between changes in CO emissions and the ultimate changes in COHb



levels.  Specifically, we assumed that a given percent change in



CO emissions would generate a similar percent change in ambient



CO.  We also assumed a linear relationship between changes in



ambient CO and the mean percent COHb level in blood.  For small



changes, these assumptions may be fairly accurate.  For larger



changes in CO emissions, there is greater uncertainty concerning



the impact on COHb levels.  For example, it is well known that



the binding affinity of hemoglobin may be nonlinearly related to



the level and change in CO.  Mage et al. (1984), however, found



a very high correlation (R^ = 0.967) between maximum COHb levels




and five-hour averages of CO concentrations, which suggests



that our assumption of linearity is a good approximation.

-------
                              VI-60


     In estimating the number of people whose COHb levels would

exceed 2.9 percent, we assumed that the geometric standard devia-

tion of the distribution of COHb levels was roughly constant for

small changes in mean COHb levels; this implies that changes in

CO emissions will shift the whole distribution proportionately.

Subject to these simplifying assumptions,  the results are indi-

cative of the potential magnitude of the effect of reduced CO

emissions on COHb levels.

     To estimate the reduction in COHb levels, we had to (1)

calculate the average annual CO emissions  from auto and resident-

ial sources* for 1976-1980; (2) project 1988 emissions from these

two sources; (3) determine the shape (mean and standard deviation)

of the frequency distribution of COHb levels for 1976-1980 from

the NHANES II for current smokers, former  smokers, and never-

smokers;  (4) adjust the mean of this distribution for 1988, based

on the changes in CO emissions; (5) calculate the change in mean

COHb levels with and without misfueling; and (6) multiply the

change in the probability distribution by  the appropriate popula-

tion group to predict the number of people now expected to be

below 2.9 percent COHb.

     The average annual CO emissions from  automobiles and

residences during 1976 to 1980 were 73.3 and 3.9 million tons,

respectively, for a total of 77.2 million  tons (U.S. EPA, 1982f).
 * We focused on auto emissions and residential sources (gas
   ovens, heating, etc.) because they have been identified as
   the most serious sources of human exposure.  Cigarette
   smoking patterns and rates are assumed to be unaffected by
   any misfueling policy.

-------
                              VI-61


The estimate of auto emissions for 1988 is 36.5 million tons.*

To estimate residential CO emissions, we determined the relation-

ship of residential CO emissions over time (U.S. EPA, 1982f),

and projected that relationship into the future.  Based on this

analysis, residential CO emissions would be 8.3 million tons in

1988.  Thus, CO emissions from these two sources were estimated

to drop to 44.8 million tons in 1988, a 42 percent decrease.

     Data on blood carboxyhemoglobin levels were collected as part

of the NHANES II study.  Because of the seasonal and geographic

pattern of the survey during the four-year period, it may not

reflect the full range of potential exposure.  Specifically, the

Northeast was not sampled during the winter,  and no high-altitude

cities were sampled.  Therefore, the survey results may somewhat

understate total CO exposure.

     For each of the subgroups — current smokers, ex-smokers,  and

never-smokers — we fit the observed COHb levels with a lognormal

distribution.  We took the log of each observation in the NHANES

and calculated the mean and the standard deviation of log(COHb)
  The baseline projections for CO for 1988 were calculated as
  follows:

  For CO we started with EPA emission factors generated in the
  draft model of MOBILE III for on-road vehicles in 1988 of
  approximately 26.65 g/mi.  The emission factor was reduced by
  0.75 (to  20 g/mi) to adjust for I/M areas and the state of
  California, which has its own, more stringent, emission con-
  trols. We assumed 159.6 million on-road vehicles traveling an
  average 11,436 miles (see Appendix B).   Multiplying:

        20  g/mi x 11,436 mi/vehicle x 159.6 x 106 vehicles
                       1 x 106 g/metric ton

                       36.5 x 106 metric  tons

-------
                            VI-62






for each of the three groups.  The mean COHb level (M) is then:






                       M = exp (u + s2/2),






where u is the mean of the lognormal distribution and s2 is the



variance.  To estimate the 1988 COHb baseline, we reduced the



mean COHb level by the predicted change in  COHb (P) over time.



We calculated the difference, log(M) - log(M-P),  for each sub-



group.  This difference was then subtracted from the log(COHb)



for each person in the subgroup to obtain a new distribution.



This procedure shifts the distribution down while keeping the



geometric standard deviation constant.  Thus, it  assumes



constant proportional changes, rather than  equal  absolute



reductions, in COHb levels across subgroups.



     To calculate P, we needed to determine the impact of the



reduction in CO emissions from 1978 to 1988 on mean COHb levels.



We assumed a linear relationship between ambient  CO and percent



COHb, with a slope coefficient of 0.16 (adapted from Ferris,



1978; U.S. EPA, 1979a; and Joumard et al.,  1981).  We used an



average ambient CO level of 3.127 ppm for 1978 (Council on



Environmental Quality, 1980).  Thus, the projected 42 percent



change in ambient CO between 1978 and 1988  would  reduce mean



percent COHb levels, on average, by 0.210 percentage points



(0.42 x 3.117 x 0.16 = 0.210).




     Given the 1988 baseline distribution of COHb levels, we



determined the effect of eliminating all misfueling and reduced

-------
                             VI-63


the mean by 5 percent, the estimated CO reduction.*  We used the

same procedure of converting to a log change, reducing the COHb

level of each population subgroup by that amount, and re-averaging.

From this new distribution, we could then calculate the change

in the number of each subgroup under 2.9 percent COHb due to

this reduction in the mean.  Since the NHANES II data are weighted

to represent the population of the United States, the weighted

average gives us the change in the total number of people in

each group who would have COHb levels above 2.9 percent.  The

results indicated that over 400,000 people would shift below

the 2.9 percent COHb level if we could eliminate misfueling:

112,000 current smokers; 62,000 exsmokers; and 227,000 never-

smokers.

     Another indication of the impact of the reduction in CO is

generated by considering the reduction in COHb levels for some of

the sensitive subgroups identified above.  This includes those

with suspected and definite angine (9.2 million), peripheral

vascular disease (0.7 million), chronic obstructive pulmonary

disease (17 million), anemia (0.1245 million), and fetuses (3.7

million).  Totalling our projections for 1988, we estimated that

there will be 31 million sensitive individuals.  In addition, we

included a subset of the elderly, who may have an increased risk
"*  The actual change in CO was calculated using the estimates of
   avoided emissions in Table VI-3 and projections of total emis-
   sions in 1988.  Assuming motor vehicles emit 86 percent of all
   CO from transportation and residential fuel combustion, the
   reduction of 2.2 million metric tons is 5 percent (2.2 x 0.86/
   36.5) of the total.

-------
                              VI-64






but who may not have been counted within these other groups.  As




an approximation, we added 25 percent of those age 65 and above,



to obtain a total population at risk of 39 million people.



     Since we did not have data on the COHb distribution for these



sensitive subgroups (including the elderly and those with coronary



heart disease), we assumed their COHb distribution was similar



to that of never-smokers.  This assumption was made because some



in this sensitive group probably take measures to reduce the



impact of air pollution and other irritants on their health.  How-



ever, it probably generates a low estimate because many in this



group continue to smoke, and others are former smokers.  Neverthe-



less, we applied the previously determined change in the portion



of never-smokers who would shift below 2.9 percent COHb because



of the reduction in misfueling (0.000865)  to the group considered



sensitive.  It suggested that 33,700 sensitive people will shift



below the 2.9 percent COHb level in 1988.






VI.C.4.  Ethylene Dibromide Emissions



     Most of the ethylene dibromide (EDB)  manufactured in the



United States is added to leaded gasoline  as a scavenger for the



lead.  Reducing the lead concentration of  gasoline would result



in an equal reduction of EDB use, which is of concern to EPA



because it is a potential human carcinogen.  EDB from leaded gas-



oline enters the air through three routes:  tailpipe emissions,



evaporative emissions from cars, and evaporative emissions from



the retail and distribution chain of gasoline.

-------
                              VI-65





     Sigsby et al. (1982) estimated EDB emissions from tailpipes



and evaporative emissions from cars.  They concluded that EDB



emissions at the tailpipe were approximately 0.37 percent of the



amount in gasoline.  However, all of the emissions tests were



done on retuned and adjusted cars, and under somewhat artificial



test procedures; the 0.37 percent survival was the average of



the tests.  The EPA federal test procedure showed an average EDB



survival of 0.69 percent.  Since EPA test of on-the-road vehicles



have generally shown substantially higher actual emissions than



in retuned cars undergoing the federal test procedure, we used



the 0.69 percent survival factor in this analysis.  Multiplying



this by the projected leaded gasoline demand in 1986 yields an



estimated reduction in EDB tailpipe emissions of 143 metric tons.



     Shed tests of the evaporative emissions of EDB generally



indicate that the EDB evaporative emissions were 1/20,000 of the



hydrocarbon evaporative emissions (Sigsby et al., 1982).  In the



absence of better data, we assumed the same ratio of refueling



evaporative emissions.  This results in an estimated 34 additional



metric tons of EDB emissions, based on our fleet model.



     Finally, EPA estimated (EPA, 1984a) that hydrocarbon emissions



in the retail and distribution chain of gasoline were 407,000 metric



tons in 1982.  Based on our estimate of 37.5 percent leaded



gasoline demand in 1986, we estimated that leaded gasoline would



produce 153,000 metric tons of hydrocarbon emissions.  Using the



ratio of EDB to hydrocarbon emissions, this suggests 7.6 metric



tons of EDB emissions from the retail and distribution stages



that would be reduced if all misfueling was eliminated.  In

-------
                              VI-66






total, we estimate 185 metric tons of EDB emissions from all



three sources (tailpipe, evaporative, and marketing emissions)




would be avoided by reducing lead in gasoline in 1986.



     In addition, EDB from gasoline storage tanks has been leaking



into underground aquifers.  Data do not exist, however, to esti-



mate the magnitude of the benefit of reducing the contamination of




ground water by EDB.






VI.D.  Monetized Benefit Estimates



     To estimate the monetized benefits of reducing emissions of



pollutants other than lead, we used two different methods.  The



first used the direct estimates of health and welfare effects,



combining the nonhealth estimates with monetized estimates of the



health effects.  As discussed earlier, that method is conceptully



correct, but suffers from substantial uncertainties and from



our  inability to quantify some potential significant benefit



categories.  Our second method uses the cost of pollution



control equipment destroyed by misfueling as a proxy for the



benefits of eliminating misfueling.






VI.D.I.  Value of Quantified Health and Welfare Benefits



      Our direct approach yields monetary estimates for only a



limited number of benefit categories, most of them related to



ozone.  The ozone-related benefits fall into two major categories:




health-related and other.  The second category already has been



estimated in monetary terms (Table VI-11) for 1986.  The ozone-



related health effects, however, thus far have been stated only



in physical terms.

-------
                              VI-67



     For the more serious category of respiratory conditions

(days of bed rest or work loss), we used the average daily

wage ($80) as a lower bound.  This is the value of the lost

output to society for an employed individual.  Because some

of these conditions also involve additional medical expense,

we used an upper bound of $120, giving a mean value of $100

per day.

     To value a day of more minor restrictions in activity due

to respiratory conditions, we relied upon Loehman et al. (1979),

whose survey results suggested a willingness to pay of $2.31 to

prevent a day of minor coughing, $4.90 to prevent minor shortness

of breath, and $8.17 to prevent minor head congestion.  We con-

verted these estimates from 1978 dollars to 1983 dollars to yield

values of $3.50 to $12.50 for avoiding a minor restricted day,

with a point estimate of $8.

     We were unable to find estimates in the literature for the

value of avoiding headaches or eye irritation.  These conditions,

however, seemed less serious than the respiratory effects dis-
                                                     «
cussed above, so we used a value of $3 per case, just below the

lower end of the range from the Loehman et al. study.

     In addition to the ozone-related benefits, we were able to

place rough monetary estimates on three additional benefit

categories:  (1) reduction in leukemia cases associated with

benzene exposure; (2) NOx-related visibility benefits; and  (3)

N0x-related materials damages.  The monetary estimates for  the

last two categories already have been discussed.  For the

-------
                              VI-68






leukemia cases, we note that the types of leukemia associated



with benzene exposure are almost invariably fatal.  As discussed



in the previous chapter, valuing reductions in risks to life is



controversial, with a wide range of values found in the litera-



ture.  For our high and low estimates, we used the range in the



EPA guidelines, $400,000 to $7 million per statistical life




saved.  For our point estimate, we used the same value employed



in the previous chapter for blood-pressure-related fatalities,



$1 milliom per case.



     Table VI-13 summarizes the monetized benefit estimates for



1986.  The total covers a broad range, from $113 million to $305



million, with a "medium" estimate of $171 million.  For the



reasons discussed earlier, plus uncertainties about the valuation



of health effects, even this broad range does not capture the



full extent of the uncertainty with respect to the categories



estimated.  Moreover, it has not been possible to monetize many



of the possible benefits; the bottom of Table VI-13 presents a



partial list of those omitted categories.






VI.D.2.  Implicit Value Based on Cost of Control Equipment



     Our second method of valuing reduced emissions of HC,



NOX, and CO is based on the cost of pollution control equipment



destroyed by misfueling.  This method assumes, implicitly, that



the benefits of controlling mobile-source emissions are at least



equal to the cost of the equipment needed to meet emission

-------
                               VI-69
TABLE VI-13.  Monetized Benefit Estimates Due to Elimination
	of Misfueling in 1986 (millions of dollars)	

                                      Low       High     Medium
Effect	Estimate  Estimate  Estimate

Quantified Benefitsf Ozone-Related

  Health

    Respiratory

      Bed rest/Lost work             21.5      110.1      47.8

      Minor restrictions              2.8       33.6      11.4

    Nonrespiratory                    9.8        9.8       9.8


  Other

    Agricultural crops               63.8       99.2      80.8

    Ornamental plants                 1.0        1.0       1.0

    Materials damage                 11.2       16.8      13.4


Quantified Benefitsf Not Ozone-Related

  Leukemia (benzene)                  1.8       30.8       4.4

  Visibility (NOX)                    0.0        2.4       1.2

  Materials damage  (NOX)              1»2        1.2       1.2


TOTAL MONETIZED BENEFITS            113.1      305.0     171.0


Unquantified Benefit Categories

  Chronic health effects due to ozone
  Forest damage due to ozone
  Direct health effects of HCs other than benzene
  Sulfate-related damages due to HC
  Acid-precipitation damages due to NOX
  Health effects due to NOX
  Vegetation damages due to NOX
  Health effects due to CO
  Health effects due to EDB

-------
                              VI-70


standards.  Emissions control equipment costs about $283 per

vehicle (U.S. EPA, 1981).*

     One complication for this method of valuing emissions is

that many catalysts are not destroyed by misfueling until they

have been in use for several years.  Thus, it was necessary to

prorate the cost of the pollution control devices.  To do so, we

first estimated the tons of emissions that would be controlled

over the life of the pollution control device.  These estimates

accounted for declines in the efficiency of the devices over

time, declines in annual miles driven as vehicles age, and scrap-

page rates.  We then discounted the emissions controlled back to

the first year (to make them comparable to the cost estimate), and

divided them into the S283 cost of pollution control equipment.

That yielded a cost of $153 per ton controlled.**  Note that this
 *The costs are "retail price equivalents," which are 30 percent
  to 50 percent of the manufacturers'  suggested retail price of
  the components (catalysts, oxygen sensors, etc.)  as replacement
  parts.  There may be a small upward  bias in this estimate, but
  we used the lower estimate in the cited report ($250, compared
  to a $425 upper bound).  Converting  to 1983 dollars gave a
  cost of $283.  (About half of the cost of oxygen sensors and
  other equipment was allocated to fuel efficiency, not pollution
  control.)

**Using 1981 emissions standards of 0.41 grams per mile for
  HC, 3.4 g/mi for CO, and 1.0 g/mi for NOX gave us 4.81 g/mi
  for all pollutants in each future year.  We multiplied 4.81
  g/mi by E/(1-E) (where E equals the  catalytic converter
  efficiency in that year) and by 10,000 mi/yr, and divided by
  1000 g/Kg to get kilograms controlled in each year by one
  catalytic converter.  We then discounted the estimate for
  each year back to the first year of  the catalytic converter's
  life.  Summing these present values  gave us an estimate of
  1.848 tons controlled by each car's  catalytic converter over
  a 10-year life.  A cost of control equipment of $283, divided
  by 1.848 tons, gave us $153 per ton.

-------
                              VI-71


estimate covers the sum of HC, NOX, and CO emissions; because

the devices control all three pollutants simultaneously, it was

not possible to generate pollutant-by-pollutant estimates using

this method.

     We then calculated year-by-year benefits by multiplying

that estimate of $153 per ton of pollutant controlled times the

projected reduction in total emissions that would be achieved in

each year if misfueling were totally eliminated.  For example,

as shown in Table VI-3, we estimate that eliminating misfueling

in 1986 would reduce the present value of emissions of the three

pollutants by 2.515 million tons.  Thus, our estimate by this

method of the benefits for 1986 is $385 million ($153 x 2.515

million tons).


VI.D.3.  Summary of Benefits of Controlling Pollutants
         other than Lead

     In this chapter, we have developed two approaches to valuing

the reductions in emissions that would be achieved by eliminating

misfueling.  The first method, based on direct estimates of

health and welfare benefits, yielded an estimate of $171 million

for 1986.  To generate estimates for other years, we scaled the

1986 estimates in Table V-13, with the scaling factor depending

on the benefit category.  For the ozone-related agricultural

benefits, we scaled the estimates in proportion to the estimated

change in average rural ozone levels  (from Table VI-10).  For

the other ozone-related benefits, we used the predicted change

in peak ozone levels (also from VI-10).  For the benzene-related

-------
                              VI-72






leukemia cases, we scaled by hydrocarbon emissions (from Table



VI-3),  because most of the reductions in benzene result from




reduced emissions of hydrocarbons from misfueled vehicles.



Finally, the two NOx-related categories, visibility and materials



damage, were scaled by NOX emission reductions (also from Table



VI-3).   This scaling procedure assumed that benefits are propor-



tional  to emission reductions over the relevant ranges.  The



results are shown in the first line of Table VI-14.



     The caveats discussed earlier with respect to the direct



estimates for 1986 apply to these estimates as well.  They



include not only the significant uncertainties associated with



the estimated categories, but also the omission of several



benefit categories due to limited data.  For ozone, the



omissions include chronic health effects and forest damage.



For HC we have not estimated any direct health effects for



hydrocarbons other than benzene, nor have we quantified the



link between hydrocarbons and sulfate formation.  For NOX, we



have not quantified benefits related to acid precipitation or



to vegetation, nor have we quantified any healt effects.  We



also have been unable to generate monetized benefit estimates



for CO, although we have made made some rough estimates of the



numbers of people whose COHb concentrations may fall below a



potential harmful exposure to EDB.  As a result of all of these



omissions, it is likely that the direct estimates are too low,



perhaps by a substantial margin.

-------
                               VI-73






     The second method, based on the implicit cost per ton of




pollutants controlled by catalysts, yielded a value of $153 per



ton for the sum of HC, NOX, and CO emissions.  The second line of



Table VI-14 shows the estimates based on that method.  The last




line shows the averages of two methods; the summary estimates in



Chapter VIII are based on those averages.  In that chapter, we



also estimate the benefits if misfueling is only partially



eliminated by the rule.

-------
                            VI-74
TABLE VI-14.  Year-by-Year Monetized Estimates of Benefits of
              Reduced Emissions of Conventional Pollutants,
              Assuming No Misfueling (millions of 1983 dollars)
Method
Direct Estimate
Control Device
Average
1985
86
193
140
1986
170
385
278
1987
169
386
278
1988
170
389
280
1989
170
392
282
1990
173
401
288
1991
180
416
299
1992
186
431
310

-------
                           CHAPTER VII

              VEHICLE MAINTENANCE, FUEL ECONOMY, AND
                    ENGINE DURABILITY BENEFITS


     Switching from leaded to unleaded gasoline, or using fuel

with a lower lead content, provides benefits to vehicle owners.

The principal benefits are lower maintenance costs from lead-

induced corrosion of exhaust systems and engines.  Reducing lead

in gasoline is also likely to increase fuel economy.  Eliminating

lead altogether, however, may cause premature valve-seat recession

in a few engines designed to rely on lead as a valve lubricant.

     Recognizing the problem of excessive deposits of lead in

engines, refiners add scavengers to leaded gasoline to prevent

such deposits.  These scavengers — primarily ethylene dibromide

(EDB) and ethylene dichloride (EDO — form compounds (e.g., halo-

gen acids and lead salts) that accelerate corrosion of exhaust

systems and engine components.  Section A of this chapter discusses

the maintenance benefits associated with reducing lead (and its

scavengers) in gasoline, and presents monetary estimates for

three categories:  exhaust systems, spark plugs, and oil changes.

     Reducing or eliminating lead in gasoline increases fuel

economy.  The refining processes used to produce octane without

lead yield gasoline that is "denser" (i.e., has a higher energy

content per gallon).  Lead fouls oxygen sensors in newer cars

that are misfueled by their owners; this also reduces fuel effi-

ciency.  Section B presents the methods used to estimate these

fuel economy benefits.

-------
                              VII-2






     Section C addresses the issue of engine durability.  Most



modern engines have hardened valve seats or other features



designed to minimize valve-seat wear.  However, many older



(pre-1971) cars, designed to operate on leaded gasoline, do



not have hardened valve seats.   In such engines, lead can play a



positive role, forming a protective veneer that "lubricates" the



exhaust-valve seat, thus reducing abrasive and adhesive wear



that can erode the seat, requiring major engine repairs.



     Concern about potential valve-seat damage in some engines



was the primary reason EPA proposed reducing lead to 0.10 gplg,



and did not propose banning it  altogether until 1995; tests



indicate that 0.10 gplg provides a margin of safety to protect



against premature valve-seat wear.  As discussed in Section C,



however, it appears that valve-seat recession may be less of a



problem than the Agency believed at the time of the August pro-



posal .



     Large studies of vehicles  in use have not detected signifi-



cant valve-wear problems when older engines are switched to



unleaded gasoline.  Moreover, studies also indicate that lead



can cause other potentially serious problems that reduce the use-



ful lives of engines, such as accelerated ring and bearing wear



and increased rates of valve burnout.  In part because of reduced



concern about potential valve-seat wear, EPA is now considering a



ban on lead in gasoline to the  take effect as early as 1988.

-------
                              VII-3





VII.A.  Maintenance Benefits



     Reducing lead in gasoline can result in less frequent



replacement of exhaust systems and spark plugs and less frequent



oil changes.  Our estimates of maintenance savings are based



primarily on tests of vehicles in use, either commercial fleets



(e.g., taxis) or vehicles owned by individuals for personal use.



In such tests, most of which were performed in the late 1960s or



the early 1970s, the maintenance records of vehicles operated on



unleaded gasoline were compared to those of vehicles using leaded



fuel.  In addition, we have supplemented such data with the



results of laboratory tests and, in a few cases, theoretical



calculations reported in the literature.



     Estimating the maintenance benefits was complicated by the



fact that in most of the studies, the leaded gasoline averaged 2.3



gplg or more (the levels that were typical when the tests were



performed), but current leaded gasoline averages only 1.10 gplg.



In addition, the tests usually examined the benefits of switching



to unleaded gasoline, not to a very low-lead gasoline of the type



permitted by the rule being promulgated.  As a result, we were



forced to interpolate from limited data, first to estimate how



many of the benefits already have been reaped in reducing leaded



gasoline to 1.10 gplgi and second to predict what additional



benefits could be reaped by further reducing, but not eliminating,



lead in gasoline.





VII.A.I.  Exhaust Systems



     Vehicles experience fewer exhaust system failures using



unleaded gasoline than leaded because of the difference in

-------
                              VII-4






acidity in exhaust gas condensates.  In cars using leaded fuel,



these condensates have a PH ranging from 2.2 to 2.6, while for




unleaded cars the range is 3.5 to 4.2 (Weaver, 1984b).  This higher



acidity accelerates corrosion in mufflers and tailpipes.



     Table VII-1 summarizes the results of four studies,



involving nine different fleets of vehicles, that examined the



effects of leaded gasoline on exhaust-system replacement rates.



(The effects on spark plug replacements were also studied; this



is discussed in Section VII.A.2.)  All of the studies found



demonstrable increases in expected lifetimes (measured in



miles) of exhaust systems in unleaded vehicles when compared



with comparable leaded vehicles.  The minimum increase in average



life of the exhaust system for unleaded vehicles was 86.5 percent



(the Wintringham et al., 1972, Detroit fleet).



     The average exhaust-system replacement rates for leaded



cars varied greatly among the different studies, ranging from 1



per 20,500 miles (Gray and Azhari, 1972, model year 1967 vehicles)



to 1 per 58,800 miles (Pahnke and Bettoney, 1969).  (The rate



per mile for the Gray and Azhari, 1972, study of consumers'



personal use cannot be computed because of inadequate data.)



Four of the fleets showed virtually no replacements of exhaust



systems in the vehicles using unleaded gasoline.  Averaging the



results of all these studies, we found about one exhaust system



replacement every 56,000 miles for cars using leaded fuel, and



essentially none for vehicles using unleaded fuel during the



test periods.

-------
                                VII-5
TABLE VII-1   Summary of On-Road Studies of Spark Plugs and Exhaust Systems
STUDY
        REPLACEMENT RATES
PER 11,000 MILES (OR PER 1 YEAR)
Pahnke & Bettoney
(DuPont, 1969)
Gray & Azhari (1972)
( Amoco )
MY 1967:
MY 1968:
Gray & Azhari (1972)
( Amoco )
Wintringham et.al.,
(Ethyl, 1972)
Detroit:
Baton Rouge:
Hick ling Partners
(Environment Canada)
(1981)
Municipal Fleets
SPARK I
UNLEADED
.534
.373
.307
.247
weight*
.440
.347
2.9 times
w/leaded
>LUGS
LEADED
.726
.840
1.085
.295
id avg .
.677
.519
as many
rehicles
EXHAUST !
UNLEADED
.0033
.149
0
.004
weightec
.155
.004
2.4 times
many for
vehicles
exclude T<
fleet)
SYSTEMS
LEADED
.187
.535
.217
.071
1 avg.
.289
.358
as
Leaded
(they
Dronto
AVG. MPV/YR
11,400
7,500
7,500
Not reported
14,575
16,850
(Unknown)
ACCUMULATED
AVG. MVP
65,000
24,000
17,000
1 to 6 yrs.
72,883
84,260
23,810 leaded
24,990 unleaded
TYPE OF SERVICE
Personal Use
Commuting and
business use
Personal Use
(Consumer
Panel)
Employee Fleet
(Business and
Personal Use)
Municipal Service
t OF VEHICLES
LENGTH OF TEST
59 matched pairs/
4.7 years
12 matched pairs/
2 and 3 years
151 matched
pairs/1-5 years
31 matched pairs
33 matched pairs/
5 years
835/5 years
LOCATION
South New Jersey
and Wilmington,
Delaware
Chicago and
suburbs
Eastern states
concentrated
in Mid-Atlantic
Detroit
Baton Rouge
Montreal
Edmonton
Toronto

-------
                              VII-6






     It is useful to look most closely at the Ethyl Corporation



(Wintringham et al., 1972) findings, as vehicles in that study



had the greatest mileage and there is a clear geographic distinc-



tion between the fleets, which highlights the effects of climate,



The Baton Rouge fleet, after an average of over 84,000 miles of



travel per car (compared to a projected lifetime of 100,000



miles), had virtually no exhaust system repairs for unleaded



vehicles, but a rate of about 1 per 31,000 miles for leaded



vehicles.  By comparison, the Ethyl Detroit fleet,  after about



73,000 miles of travel per vehicle, had a rate for  unleaded



vehicles of one exhaust system replacement per 46,000 miles, but



a rate of 1 per 24,000 miles for leaded vehicles.  The main



reason for the different experiences in Baton Rouge and Detroit,



the authors concluded, was the greater external corrosion due to



road salts in the colder climate.



     The Detroit results are consistent with the Environment



Canada findings (Hickling Partners, 1981) for two municipal



fleecs, which had 42 percent fewer exhaust system replacements



(at equivalent mileage) for cars using unleaded fuel in a cold



climate.  On the other hand, the DuPont (Pahnke and Bettoney,



1971) and Amoco (Gray and Azhari, 1972) findings, conducted in



the mid-Atlantic region, in Chicago, and in the eastern U.S.,



were closer to Ethyl's in Baton Rouge; there were virtually no



exhaust repairs for vehicles using unleaded fuel.  However, the



evidence suggests that exhaust system corrosion rates do vary



with climate and we have incorporated that variation into our



estimates.

-------
                              VII-7


     Weighting Ethyl's findings for Detroit and Baton Rouge

according to the portion of registered cars in Sunbelt versus

Snowbelt states in 1982 (43 percent and 57 percent, respectively,

according to the Motor Vehicle Manufacturers Association [MVMA],

1983), mufflers nationally would last an average of three times

longer with unleaded fuel than with leaded.  Unfortunately,

these studies were conducted on fleets of vehicles over several

years, but for less than the lifetimes of the vehicles.

It is possible, therefore, that the studies ended shortly before

many of the unleaded vehicles required exhaust system replacements

Perhaps the replacement rates for unleaded vehicles would have

increased significantly had the fleets traveled another 10,000

to 20,000 miles.  The reported findings, thus, may have overesti-

mated the differences between unleaded and leaded vehicles.

Because of this concern, we have assumed that mufflers on

vehicles using unleaded fuel would last only twice as long (in

miles) as those on vehicles using leaded fuel.

     We assumed mufflers on vehicles using leaded gasoline would

last about 50,000 miles; in the studies we reviewed, the leaded

fleets averaged about 20,000 to 60,000 miles between exhaust

system replacements.*  Applying our factor of two yielded an
* Passing references in the literature and several commenters
  have suggested that the metallurgy of exhaust systems was
  upgraded during the 1970s, e.g., changing from cold-rolled
  milled steel to chromium stainless steel.  Since the more
  durable metal would corrode less easily, the commenters
  suggested, this design improvement might affect performance
  and our estimates of benefits might be substantially over-
  stated.  However, on the improved exhaust systems, only the
  parts from the exhaust manifold to the catalytic converter are
  stainless steel.  The remaining components of the exhaust
  system (exhaust pipe, muffler, and tailpipe) are generally
  still made of rolled steel.  These are the parts that we esti-
  mated would corrode from leaded gasoline.  Thus, this tech-
  nology change should have no effect on our estimates of savings.

-------
                            VII-8






expected lifetime of 100,000 miles for vehicles using unleaded



gasoline.  Based on a cost of $120 per exhaust system replacement,



the savings per mile are ($120)(1/50,000 - 1/100,000) = $0.0012/



mile, or $12.00 per year for a vehicle driven 10,000 miles



annually.



     This estimate must be applied with care, because it is based



on comparisons of vehicles operated on leaded gasoline with a



lead content over 2 gplg to vehicles operated on unleaded gasoline.



We were uncertain as to how much, if any, of the benefits of



reduced exhaust system corrosion might already have been reaped



as a result of reducing lead to its current level of 1.1 gplg.



Fortunately, Gray and Azhari (1971) examined exhaust system cor-



rosion rates using leaded gasoline at both 2.3 gplg and 0.5 gplg;



they found no difference between the two types of leaded gasoline.



(Both showed corrosion rates 10 to 20 times higher than those



with unleaded.)  This finding suggests that no reductions in



exhaust system corrosion are reaped until the lead content falls



below 0.5 gplg.  Thus, we estimated that vehicle owners switching



from leaded gasoline at 1.1 gplg to unleaded will experience



savings of $0.0012/mile.  This estimate applies to misfuelers



who are deterred by the rule, and in the case of a complete ban.



     Estimating the exhaust system benefits for vehicle owners who



use low-lead (0.10 to 0.50 gplg) gasoline was more problematic.



For lack of better information, we assumed linearity between zero



and 0.5 gplg; e.g., at 0.10 gplg, owners would get 80 percent of



the benefits, or 0.8(0.0012) = $0.00096/mile.  This translates



to $9.60 per year for a vehicle owner driving 10,000 miles yearly.

-------
                              VII-9


VII.A.2.  Reduced Fouling and Corrosion of Spark Plugs

     The corrosive effects of lead and its scavengers also reduce

the useful life of spark plugs.  As shown in Table VII-1, all of

the fleet studies showed longer intervals between spark plug

changes for vehicles operated on unleaded than on leaded.   The

increases ranged from 19 percent (Gray and Azhari, 1972, consumer

use study) to 350 percent (Gray and Azhari, 1972, for model year

1968 vehicles); on average, the gain was about 80 percent.

     To estimate benefits, we assumed that the average interval

between spark plug changes with leaded gasoline at 2.3 gplg would

be 10,000 miles.*  That is roughly consistent with manufacturers'

recommendations in the early 1970s, before new cars used unleaded.

Applying the 80 percent improvement estimated above for users of

unleaded would allow an interval of 18,000 miles between changes.

     As with the exhaust system data, these tests used leaded

gasoline at about 2.3 gplg and unleaded, so we had to make

adjustments to account for savings due to the change from 2.3 to

1.1 gplg.  In 1971, Toyota reported (Champion, 1971) that fouling
 * In practice, consumers appear to have changed spark plugs less
  frequently; the average for leaded vehicles  in the fleet tests
  was about  15,000 miles.  Owners who delay spark plug changes,
  however, suffer losses due to decreased  fuel economy; which
  usually exceed the cost of replacing spark plugs  at the appro-
  priate interval.  For consumers who change spark  plugs less
  frequently than optimal, the benefits of unleaded or reduced
  lead gasoline will come through added fuel economy (since the
  spark plugs will degrade less on the unleaded gasoline) rather
  than reduced spark plug changes.  For example, Graver et al.
  found that spark plugs with the wrong gap and orientation led
  to decrements of up to 7 percent in fuel economy.  In general
  these benefits will be higher, so our use of the  replacement
  costs is conservative.

-------
                              VII-10


of spark plugs occurred at the same rates with leaded gasoline

at 0.2 gplg as with unleaded gasoline.  In 1972, Union Oil reported

(Champion, 1972) that spark plug performance was similar for

unleaded and leaded gasoline at 0.5 gplg.  With either type of

gasoline (unleaded or 0.5 gplg), Union reported spark plugs

lasted four times longer than with leaded gasoline containing

3.0 gplg.  These findings suggest that there is a threshold

below which further reductions in lead yield no additional gains

in spark plug life.

     For lack of better information, we assumed that the threshold

occurred at 0.5 gplg, and that the relationship between lead and

spark plug life from that level to 2.3 gplg was linear.  Thus, we

assumed that the reduction in lead from 2.3 gplg to 1.1 gplg

increases the interval by (2.3-1.1)/(2.3-0.5)(80 percent) = 53

percent, or from 10,000 miles to 15,333 miles.   At $18 per spark

plug change, we estimated that reducing lead from 1.1 gplg to 0.5

gplg or below will provide benefits of (1/15,333 - 1/18,000)($18)

= $0.000174/mile.  The annual benefit for a car owner driving

10,000 miles per year would be $1.74.*  Note that this estimate

applies both to those who switch from leaded to unleaded and to

those who use low lead gasoline.
*  By contrast, if car owners replace their spark plugs less
   frequently than they should, and the fuel economy penalty of
   increased spark plug degeneration from leaded fuel is only
   0.5 percent, the benefits would be about double this estimate.

-------
                              VII-11





VII.A.3.  Extended Oil Change Intervals



     The combustion products that deposit on engine surfaces



cause corrosion and rusting.  Engine oil accumulates much of the



debris from this corrosion, as well as some portion of the gaso-



line lead.  According to at least one estimate, up to 10 percent



of the lead in gasoline ends up in the used oil, comprising up



to 50 percent of the weight of engine oil sludge (Gallopoulos,



1971).



     The particles and corrosive products that accumulate in the



oil cause substantial wear in the engine, and the internal engine



rust may cause hydraulic valve lifter sticking (Cordera et al.,



1964).  Besides the long-term engine wear that reduces the



durability of the engine, the vehicle driver may also experience



excessive valve noise and other performance degradation due to



this premature contamination of oil.  Although rusting can occur



even in the absence of the halogen acids derived from lead salts,



these compounds are the major cause of internal rusting under



normal driving conditions (Weaver, 1984b).



     The fleet studies summarized in Table VII-1 generally did



not consider oil changes or, if they did, found little difference



between the behavior of drivers using leaded and those using



unleaded.  This result should not be surprising, as it is unlikely



that the vehicle owners in the studies were aware of the impact of



eliminating lead on engine oil.  Presumably most drivers today are



similarly unaware, and follow the recommendations in their owners'



manuals (or habit) in changing their oil.  If unleaded gasoline



increases the useful life of engine oil, however, switching to

-------
                              VII-12






unleaded will yield benefits in the form of improved engine dur-



ability, even if oil change intervals do not change.  Thus, in



these cases, our estimates may be viewed as a proxy for improved



engine durability.



     Gallopoulos (1971), of the General Motors Corporation, was



one of the first people to investigate the potential impacts of



unleaded gasoline on oil-change intervals.  He concluded that



with unleaded gasoline it might be possible to decrease the



frequency of oil changes from 2 or 3 per year to only 1 per



year, but added that further investigation was needed.



     Pless (1974), also of General Motors, reported more



conclusive results based on experiments with taxicabs under con-



ditions that took an unusually severe toll on oil quality.   In



a group of 20 taxis (1970 model year), Pless found less piston



varnish, ring wear, and used-oil insolubles for the unleaded



vehicles after 16,000 miles of stop-and-go service.



     On a fleet of 1972 taxis, Pless (1974) compared unleaded



vehicles after 16,000 miles without an oil change with leaded



vehicles (2.7 grams of lead per gallon) after 8,000 miles.   The



results indicated less sludge, oil ring deposits, compression



ring wear, cam and lifter wear, and oil degradation for the



unleaded vehicles with extended oil change intervals, compared to



the leaded taxis with "normal" oil changes.  While the unleaded



vehicles had somewhat greater plugging of oil filters, Pless



concluded that this was not a significant finding.  Finally,



another fleet traveling predominantly short trips  (closer to



typical consumer driving patterns) led Pless to conclude:

-------
                              VII-13


          A combination of unleaded gasoline and doubled
          oil change interval allowed significantly less
          ring wear, and directionally less sludge,
          varnish, and cam and lifter wear than did the
          combination of leaded gasoline and "standard"
          oil-change interval.

Subsequent to these findings, both General Motors and Chrysler

recommended lengthened periods between oil changes.

     Gergel and Sheahan (1976), of the Lubrizol Corporation,

reported results similar to those of Pless, but did not find any

significant plugging of oil filters.  They concluded that engine

wear was the limiting factor in extending oil change intervals,

suggesting a maximum of 12,000 miles between changes for unleaded

gasoline engines.

     The evidence indicates that there is a relationship between

lead additives and oil change intervals shown through reduction

in engine and engine-parts wear (from reduced abrasive lead

particles or reduced rust), oil degradation, and general engine

and engine-part cleanliness (e.g., lack of deposits and sludge).

One indication of this relationship is the fact that manufacturers'

recommended intervals between oil changes have more than doubled

since the introduction of unleaded gasoline, and in a recent

statement to EPA the MVMA stated that using unleaded gasoline

allows the doubling of oil change intervals.  (Some of the increase,

however, has reflected improved oil quality.)

     To estimate the benefits of increased oil change intervals,

we assumed an interval of 5,000 miles with leaded gasoline (at

2.3 gplg) and, following Pless1 results, a doubling of the

interval to 10,000 miles with unleaded.  Based on a cost per oil

-------
                              VII-14


change of $10.50,* the estimated benefit is then (1/5,000 -

1/10,000) ($10.50) = $0.00105/mile.  The annual benefit to a car

owner would be about $10.50, based on 10,000 miles per year.

     As before, this estimate is based on changing from gasoline

containing about 2.3 gplg of lead to gasoline with no lead.  To

estimate what the benefits would be in going from 1.1 gplg to

lower levels, we relied on Cordera et al. (1964), who examined

the effects on engine rust of varying concentrations of lead (and

its scavengers, EDB or EDC) in gasoline.  They evaluated valve-

lifter rusting at 0, 0.53, and 3.2 gplg; rusting decreased

nonlinearly with reductions in lead (and its scavengers), with

the sharpest declines occurring at low lead levels.   Fitting a

smoothed curve to their data suggests that about 12.7 percent of

the total reduction in rusting would occur in going  from 2.3 gplg

to 1.1 gplg, leaving 87.3 percent, or 0.873($0.00105/mile) =

$0.00092/mile in benefits for switching from leaded  at 1.1 gplg

to unleaded.  Based on that same curve, we estimated that going

from 1.1 to 0.10 gplg yielded 58.3 percent of the total benefit,

or 0.583(0.00105) = $0.00061/mile.


VII.A.4.  Summary of Maintenance Benefits

     Table VII-2 summarizes our maintenance estimates on a per

mile basis.  They total $0.00038/mile for changing from 1.1 gplg


* This assumes four quarts of oil at $1.50 each, plus half an oil
  filter (assuming the filter would be replaced every other oil
  change) at $4 each, plus 15 minutes of labor at $10 per hour.

-------
                            VII-15
TABLE VII-2.  Estimated Maintenance Benefits Per Mile
              (cents/mile)
Standard (gplg)
Category
Exhaust systems
Spark plugs
Oil changes
Total
0.50
0.000
0.017
0.021
0.038
0.10
0.096
0.017
0.061
0.174
0
0.120
0.017
0.092
0.229

-------
                              VII-16






to 0.5 gplg, $0.00174/mile for tightening to 0.10 gplg, and




$0.00229/mile for eliminating lead altogether.



     To calculate the benefits in each year, we combined those



estimates with estimates from our fleet model of the numbers of



miles driven by light-duty vehicles of different types.  For



1986, for example, we estimate that legal leaded users of light-



duty vehicles will travel 307 million miles and that misfuelers



will travel 174 million miles.  For a standard of 0.1 gplg,



assuming that it eliminates all misfueling, the estimated benefit



is then ($0 .00174/mile)(307 million miles) + ($0.00229/mile)(174



million miles) = $933 million.  Table VII-3 presents year-by-



year estimates for the three alternative schedules presented in



earlier chapters.  As before, the estimates assume that all



misfueling  is eliminated; alternative assumptions are explored



in Chapter VIII.



     Note that these monetized estimates of maintenance savings



apply only  to light-duty vehicles (cars and light-duty trucks),



because we did not have data on such savings for other classes of



vehicles, such as heavy-duty trucks and busses.  It is likely,



however, that such vehicles would also reap maintenance savings.



In 1986, we estimate that they will account for about one-quarter



of the demand for leaded gasoline; their share will grow in later



years.  Consequently, these estimates understate the benefits in



these categories.

-------
                            VII-17
TABLE VII-3.  Year-by-Year Estimates of Maintenance Benefits,
              Assuming No Misfueling (millions of 1983 dollars)
Category
Rule
Spark Plugs
Proposed
Alternative
Final
Exhaust Systems
Proposed
Alternative
Final
Oil Changes
Proposed
Alternative
Final
Total
Proposed
Alternative
Final
1985

0
46
46

0
95
95

0
112
112

0
252
252
1986

84
84
84

503
356
503

347
267
347

933
706
933
1987

77
77
77

473
411
473

330
287
330

880
775
880
1988

73
73
73

450
450
450

318
318
318

840
840
840
1989

69
69
69

433
433
433

309
309
309

811
811
811
1990

67
67
67

422
422
422

303
303
303

792
792
792
1991

65
65
65

415
415
415

301
301
301

780
780
780
1992

64
64
64

412
412
412

301
301
301

776
776
776

-------
                              VII-18






VII.B .   Improved Fuel Economy



     Reducing the lead content of gasoline should improve fuel



economy in three ways:  by increasing the energy content of



gasoline through more intense processing, by reducing the fouling



of oxygen sensors in misfueled late-model vehicles, and by reducing



the fouling of spark plugs.  Energy content and oxygen sensor



benefits are discussed and monetized below.  The third source of



benefits was covered, at least in part, by our estimate of increased



intervals between spark plug changes, and hence is not monetized



in this section.






VII.B.1.  Energy Content



     Increased reforming and isomerization of gasoline to replace



the octane lost through lead reductions increases the density



(energy content) of gasoline.  Unleaded gasoline also generates



more deposits in engine combustion chambers, which increases



compression and engine efficiency slightly.  Exxon (1978) has



estimated that these effects could cause a 1 to 1.5 percent



improvement in fuel economy-



     To estimate the benefits of increased fuel economy from



denser gasoline, we computed the changes in density predicted by



the DOE refinery model at different lead levels.  Because the



predicted change represents a relatively small difference between



two estimated large numbers and depends on the precise methods



used by refiners to raise octane, these estimates are subject to



substantial uncertainty.  We used a formula developed by the



Society of Automotive Engineers (1979) to estimate the change in



fuel efficiency as a function of density.  Finally, we multiplied

-------
                              VII-19





the estimated savings by the retail price of a gallon of gasoline,



(For this calculation, we used the retail price, $1.10 per gallon



in 1983 dollars, rather than the refinery gate price because a



reduction in gasoline consumption — which would result from



greater fuel economy -- yields savings in distribution and



retailing costs, as well as refining costs.)





VII.B.2.  Reduced Fouling of Oxygen Sensors



     For vehicles with oxygen sensors and closed-loop catalyst



systems, reducing lead in gasoline offers additional gains in



fuel efficiency to the extent that it reduces misfueling.   In



such vehicles, lead fouls the oxygen sensor, thus reducing its



ability to optimize engine performance for maximum fuel economy.



In a recent paper, Armstrong (1984) presented data showing that



replacing the oxygen sensor as well as the catalyst in a misfuel-



ed vehicle reduced hydrocarbon emissions, indicating that leaded



gasoline causes the oxygen sensor to require a fuel mixture that



is too rich.  She found that replacing the oxygen sensor reduced



tailpipe emissions by an average of 0.13 grams per mile.  Because



the catalyst oxidizes most of the extra hydrocarbons that the



engine wasted with a lead-fouled sensor, it is necessary to



divide that number by (1 - catalyst efficiency) to estimate the



reduction in wasted hydrocarbons.  In Armstrong's sample, the



average efficiency of the catalysts was 83.4 percent, so the



excess consumption of hydrocarbons in the misfueled vehicles was



0.13/(1 - 0.834) = 0.783 grams per mile.  If the sensor is func-



tioning properly, these hydrocarbons are burned in the engine,



increasing fuel economy.

-------
                              VII-20

     To estimate the benefits associated with reduced hydrocarbon
consumption, we estimated the number of post-1981, sensor-equipped
vehicles that would be misfueled for the first time in each year
from 1985 through 1992.  We then computed the discounted (at a
real rate of 10 percent) number of miles that such vehicles
would travel, on average, over their remaining lifetimes, multi-
plied by 0.783 grams per mile and converted the resulting grams
of hydrocarbon to equivalent gallons of gasoline.  Finally, we
multiplied total gallons of gasoline for each year by the price
of gasoline ($1.10).

VII.B.3.  Summary of Fuel Economy Benefits
     Table VII-4 presents the year-by-year estimates of fuel
economy benefits, assuming, as in past chapters,  no misfueling.
The estimates are dominated by the savings due to higher fuel
density.  These savings fall over time because as demand for
leaded fuel declines, the baseline fuel density rises.  The savings
due to reduced oxygen sensor fouling increase over time because
the number of misfueled post-1981 vehicles would  grow in the
absence of the new rule.

VII.C.  Engine Durability
     Lead in gasoline can have both positive and  negative effects
on the durability of engines.  The primary concern with unleaded
gasoline has been premature valve-seat wear in engines designed
to use leaded gasoline.  Such effects have been demonstrated in
laboratory and track tests, although tests of vehicles in normal
use have failed to find any significant acceleration of valve-seat
recession.   Section  VII.C.I  focuses  on  an  evaluation  of  valve

-------
                            VII-21
TABLE VII-4.  Year-by-Year Estimates of Fuel Economy Benefits,
	Assuming No Misfueling (millions of  1983 dollars)

Category
  Rule	1985  1986   1987  1988  1989   1990   1991  1992

Fuel Density
    Proposed            0    168    150    97    114   113    148    140
    Alternative        57    106    106    97    114   113    148    140
    Final              60    168    150    97    114   113    148    140

Oxygen Sensors
    Proposed            0    22     25    27    30    32    34    35
    Alternative        11    22     25    27    30    32    34    35
    Final              11    22     25    27    30    32    34    35

Total
    Proposed            0    190    175   124    144   145    182    175
    Alternative        68    128    132   124    144   145    182    175
    Final              68    190    175   124    144   145    182    175

-------
                              VII-22






seat recession.



     Tests indicate that lead and its scavengers can increase



the wear of other major engine components, and consequently,



shorten the useful lives of engines using leaded gasoline.



These negative effects of lead are discussed in section VII.C.2.






VII.C.I.  Valve-Seat Recession



     We reviewed two types of research in evaluating the potential



for valve-seat recession.  The first type of study was engine



tests on dynamometers, done using either unusually high engine



loads to test valve durability, or cycles that simulated typical



driving patterns, or a combination of the two.  The second type



of study involved on-road vehicle tests, ranging from high-load



studies to surveys of consumers' experiences.  The advantage of



engine tests is their greater measurement precision and control



over test conditions.  The on-road studies, on the other hand,



are more likely to reflect "real world" effects.






VII.C.I.a.  Laboratory and Track Studies of Valve-Seat Recession



     Laboratory studies suggest that exhaust-valve recession



results from abrasion and adhesion on the valve seat when engines



operate continuously under high temperatures, loads, or speeds.



(For detailed discussions of the mechanisms of valve wear, see



Godfrey and Courtney, 1971; Giles, 1971; or Kent and Finnigan,



1971.)



     Several researchers have examined rates of valve recession



as a function of engine operating variables and the amount of



lead in the fuel.  Giles (1971) and Godfrey and Courtney  (1971)

-------
                              VII-23


were consistent in finding that recession rates were a function

mostly of engine speeds.  The shape of this function apparently

varied significantly by vehicle models and years.

      Table VII-5 summarizes the available laboratory and track

studies of valve recession as a function of lead concentrations.

Note that most engine studies of valve recession were conducted

at speeds and loads much greater than normal driving patterns.

For example, Giles and Updike (1971), of TRWs Valve Division,

conducted six dynamometer tests simulating vehicle speeds from

50 to 100 mph.  These tests, combined with the other evidence,

led them to conclude that:

     exhaust valve recession in engine I accelerates rapidly
     above 70 mph....  The data shown here also indicate that
     the average driver, who seldom exceeds 70 mph, should not
     experience significant engine deterioration while using
     lead-free gasoline.  The salesman, however, who drives
     15,000 turnpike miles per year at 80 mph, may well expect
     valve train problems.  (p. 2369)

Their data showed the rate of valve-lash loss actually decreased

slightly between 50 and 70 mph (wide open throttle at 2000 and

2800 rpm, respectively).  Felt and Kerley (1971), of Ethyl

Corporation, also found that valve recession (using unleaded

gasoline) was about two-thirds lower for vehicles traveling at 60

mph than for those traveling at 70 mph, despite going 22 percent

to 280 percent more miles.

     These studies were designed either to investigate the

mechanisms causing exhaust-valve-seat recession, or to show

the importance of leaded fuel combustion products in reducing

valve wear.  They did not usually test for the likelihood of

-------
                                      VII-24
TABLE VII-5.  Summary of Findings of Track and Dynamometer Studies of Lead
              Levels and Valve Recession
     Study
    Findings and Conditions
Doelling, 1971



Felt and Kerley, 1971



Fuchs, 1971


Giles et al., 1971
Giles and Updike, 1971
Godfrey and Courtney, 1971
Kent and Finnegan, 1971
Pahnke and Bettoney, 1971
U.S. Army, 1971
Engine tests at about 65 mph showed that between
0.04 and 0.07 gplg was sufficient protection.
Lead levels of 0.14, 0.07, 0.04, and 0.0 gplg.

Excess wear in continued high speed operation
(70-95 mph).  Much lower wear rates in intermit-
tent operation using oil with metal additives.

Engine tests showed 0.5 gplg virtually eliminated
valve recession.  Lead levels of 0.5 and 0.0.

Rapid wear on engines with unhardened valve seats
at engine speeds typical of 80 mph or greater.
Little or no excess wear on unleaded with hardened
valve seats at maximum engine speed.  Limited
testing of heavy-duty truck engines with inserts
showed no increased wear.

No excess wear at less than 3000 rpm, excess wear
on unhardened valve seats above that.  No excess
valve wear at 3500 rpm with hardened valve seats.

High load and speed are cause of valve recession
on unleaded.

High load engine tests showed 0.20 gplg was
sufficient protection.  Lead levels of 3.0, 0.5,
0.2, and 0.0 gplg.

High load engine tests showed severe valve
recession at 0.0 gplg, none at 0.5 gplg.

Dynamometer test of three vehicles and three
stationary generators showed no excess valve-seat
wear.  Generators at maximum rpm and load; had
hard valve-seat inserts.  Vehicle engines at
maximum torque, wide open throttle, and 3200 rpm;
one each with unhardened, hardened, and hardened
inserts

-------
                              VII-25





valve recession under normal driving conditions.  In particular,



intermittent operation at high speeds may well result in substan-



tially different test patterns than continuous operations.  Felt



and Kerley, for example, found no significant protective effect



of metal additives in engine oil in a continuous high engine



speed test, but that metal additives reduced wear rates by a



factor of 10 in a test on the same engine with operations alter-



nating between 50 mph and high speed operation.



     Overall, the laboratory studies implied that using unleaded



gasoline exclusively in vehicles with unhardened valves designed



for leaded fuel could risk premature valve failure under severe



engine loads.  These studies indicated that such severe recession



is most likely to occur in engines operating at high loads or



speeds, which, for light-duty vehicles, would involve vehicle



speeds well above the legal speed limit of 55 mph for extended



periods of time (tens or hundreds of hours).



     The evidence indicated that conditioning a vehicle on leaded



gasoline helped to prevent valve recession during subsequent use



of unleaded gasoline for a limited time, but did not lower the



longer-term risk.  Giles (1971) measured valve wear during and



after "break-in" periods of an engine running on leaded gasoline.



He demonstrated that recession rates were high initially, even



using leaded gasoline.  But, as the leaded gasoline combustion



products built up on the valve seat, recession rates dropped to



very low levels.  Giles showed that, after switching the engine



to unleaded gasoline, recession rates continued to be low until

-------
                            VII-26






the lead deposits wore away (after about 10 hours of high engine



speed operation).  Recession then rose again to high rates.



     Giles and Updike also showed that vehicles with hardened



valves had no more wear on unleaded gasoline than vehicles with



unhardened valves had on leaded.  This result was confirmed by



the Army dynamometer studies.



     For engines equipped with hydraulic valve lifters (the vast



majority of on-highway engines), the amount of valve-seat recession



that can be tolerated before serious problems appear is about



0.07 to 0.15 inches.  Engines  without hydraulic lifters will



require adjustment after a much shorter time.  The results of



the laboratory and test-track  studies discussed above indicate



that this amount of wear can be experienced in as little as 100



hours of continuous 70 mph freeway driving (7,000 miles) in a



light-duty vehicle with unhardened valve seats.  At 60 mph (a



more typical speed for the present day), this limit could still



be reached in as little as 18,000 miles of continuous high speed



operation.  Miles accumulated  at 55 mph and less are unlikely



to contribute to seat recession.  Heavy-duty gasoline engines,



which operate at higher rpm and higher load levels, could be



affected even sooner, but most engines of that type have hardened



valve seats, or valve-seat inserts, and may have speed governors



that restrict engine rpms.  These studies were done before




modern engine oils, which contain additives to reduce such wear.



     Most of the laboratory studies compared valve wear with



unleaded gasoline to that with leaded at the levels typical of

-------
                              VII-27





the late 1960s and early 1970s (about 2.3 gplg).  Several studies,



however, also tested the effects on valve wear of using gasoline



with a lower lead content.  At least four studies concluded that



0.5 gplg of lead would provide sufficient valve-seat protection,



even under severe conditions (Kent and Finnigan, 1971; Pahnke



and Bettoney, 1971; Felt and Kerley, 1971; Fuchs, 1971).  Kent



and Finnigan (1971) also found that "as little as 0.2 g/gal of



lead was sufficient to reduce wear to substantially zero."



     Only one study examined valve wear at very low lead concen-



trations to discover how little lead was necessary to eliminate



valve recession.  Doelling (1971) conducted tests at lead levels



of 0.04, 0.07, and 0.14 gplg for 100 hours each.  Focusing on



the maximum recession of any one of the valves, Doelling found



no recession at 0.07 or 0.14 gplg, but found excess wear at 0.04



gplg.  He thus concluded that leaded gasoline would protect



exhaust valves beginning at levels between 0.04 and 0.07 grams



of lead per gallon.



     Based primarily on concerns raised by these studies, EPA's



proposed rule allowed the continued use of low-lead gasoline



through the mid-1990s.  The standard of 0.10 gplg was chosen to



provide a margin of safety in protecting against valve wear.





Vll.C.l.b.  Fleet Studies of Valve-Seat Recession



     The laboratory tests discussed above suggest that premature



valve-seat recession in some engines with unhardened valve seats



that operate at high speeds could be a serious cost of eliminating



lead in gasoline altogether, though it should not be a significant

-------
                            VII-28






problem with the 0.10 gplg standard being promulgated.  Studies



of engines under normal operating conditions, however, suggest



that even a ban on lead might not have major impacts on valve-seat



durability.  The available studies are summarized in Table VII-6.



    Several of these fleet studies found little or no incidence



of valve-seat problems with unleaded gasoline (Pahnke and Conte,



1969; Orrin et al., 1972; Gray and Azhari, 1972).  Other fleet



studies were inconclusive concerning the relative incidence of



valve-seat problems for unleaded vehicles (Pahnke and Bettoney,



1971; Grouse et al., 1971; Pless, 1974).  Wintringham et al.



(1972) also noted that reported incidents of valve problems were



rare among users of unleaded gasoline in the late 1960s (when at



least one major oil company sold a premium unleaded grade).  Two



studies, however, cited more valve-seat problems for unleaded



than for leaded vehicles (Wintringham et al., 1972; Felt and



Kerley, 1971).  Recently, EPA has become aware of a very large



test by the U.S. Army in the mid-1970s, which found no problems



using unleaded in a wide range of vehicles.   These and other



studies are discussed below.



     In the middle and late 1960s, Ethyl Corporation carried out



an extensive five-year study of leaded versus unleaded gasoline



use (Wintringham et al., 1972).  This study included 64 matched



pairs of cars, owned and driven by Ethyl Corporation employees.



One vehicle in each pair used leaded gasoline, the other used



unleaded exclusively.  The cars averaged more than 15,000 miles



per year (an average of 78,749 miles per car for the unleaded

-------
                                      VII-29
TABLE VII-6.  Summary of Findings of Consumer and Fleet Studies of Lead Levels
              and Valve-Seat Recession
     Study
   Findings and Conditions
Grouse et al., 1971
Grouse et al., 1971
Felt and Kerley, 1971


Gray and Azhari, 1971



Orrin et al., 1972


Pahnke and Bettoney, 1971



Pahnke and Conte, 1969



Pless, 1974


U.S. Army. 1975
U.S. Post Office,  1983
A 50,000 mile test of matched pairs found an
insignificant decline in valve wear on unleaded.
Lead levels of 2.6 and 0.0 gplg.

A severe service test using a state police
patrol fleet found valve recession after 10 to
15 thousand miles.  Lead levels of 3.1 and 0.0
Wintringham et al., 1972
An employee fleet test found more valve problems
at 0.0 than at 0.5 gplg.

No additional valve problems found with employee
fleet test or in a consumer survey.  Lead levels
of 2.8 and 0.0 gplg.

No extra valve problems in a study of taxi
fleets.  Lead levels of 2.8 and 0 gplg.

A consumer survey found no clear difference but
somewhat more valve problems.  Lead levels of
2.3 and 0 gplg.

No additional valve problems for employee cars
in personal use.  Lead levels of 2.8, 0.1, and
0.0 gplg.

No severe valve problems, but some valve-stem
wear with unleaded in one of the taxi fleets.

Conversion of six Army bases to unleaded produced
a valve recession rate of 1 per 10 million VMT*
for commercial vehicles.  No valve recession in
other vehicles.

Conversion of 1,562 1975-model Ford trucks with
valve seat inserts to unleaded produced valve
recession rate of 1 per 15 million VMT*.  152
International Harvester trucks experienced
no valve-seat failures on unleaded.

An employee fleet test found more expensive valve
problems with unleaded; about 1 per million VMT*.
        *Vehicle miles traveled.

-------
                            VII-30






group over the life of the test).  At that time, speed limits on



the interstate highway system were 65 or 70 miles per hour.



Despite this, only four unleaded vehicles (6 percent) required



cylinder head replacements.  One vehicle in the leaded group also



required a new cylinder head during the same period.  In addition,



the absence of lead showed a beneficial effect in reducing the



number of burned and damaged valves -- only six vehicles in the



unleaded group required valve jobs, compared with sixteen of the



vehicles using leaded gasoline.



     Three other studies, conducted about the same time, gave



similar results.  Gray and Azhari (1972) reported the results of



a small fleet test and a consumer use survey, neither of which



indicated any particular problems with valve-seat recession.



Overall, engine repair costs for the unleaded group were lower



than for the leaded group, exactly the opposite of what would



have been expected if valve-seat recession were widespread.  How-



ever, no details of repair records were provided, so the data



must be interpreted cautiously.



     Grouse et al. (1971) provide data on four cars used in a



comparison of leaded and unleaded gasoline effects on lubricants.



The cars were operated on a more-or-less normal schedule, involv-



ing home-to-work driving on weekends and turnpike driving on



weekends.  Three cars completed 50,000 miles successfully on this




schedule; the fourth suffered from valve-seat recession and had



to be dropped from the test after 34,000 miles.  None of these



cars had operated on anything but unleaded fuel.  This is signi-

-------
                              VII-31





fleant, since the researchers found that preconditioning on leaded



fuel at least doubled the mileage obtained in another test fleet



(operating under very severe patrol-service conditions) before



valve recession became a problem.



     Schwochert (1969) operated an experimental catalyst-equipped



car for 50,000 miles on unleaded gasoline in a test cycle that



simulated typical city and highway driving (the Auto Manufacturers



Association's mileage accumulation cycle).  Valve-seat recession



in this cycle did not exceed 0.02 inches.  Subsequent operation in



a very high-speed cycle (70 to 90 mph) destroyed the valve seats



in less than 12,000 miles.



     All of the tests discussed above involved light-duty vehicles.



Heavy-duty vehicles, since they often have lower power-to-weight



ratios and higher rpm at highway speeds, may suffer more severely



from valve recession with unleaded gasoline if they have unhardened



valve seats.  These concerns are also applicable to a wide range of



farm, construction, and industrial equipment, much of which also



operates at high average power ratings and rpm.



     In this regard, it is instructive to consider tests conducted



by the U.S. Army.  These involved some 7,600 vehicles — including



light-duty cars and trucks, heavy-duty trucks, tractors, jeeps,



tactical and combat vehicles, and some motorized heavy equipment



— and lasted for three years, with about half of the vehicles



being added during the last year.  Table VII-7 lists the types



of vehicles and other engines involved.  The average commercial



vehicle in the study accumulated over 10,000 miles per year, and



many accumulated more.  One class-6 truck put on 34,000 miles  in

-------
                              VII-32


the first year alone, and several pick-ups accumulated over 30,000

miles in the first year.  Military vehicles accumulated  lower

mileage, but generally operate under high load conditions.  The

study is documented  in a series of reports by the Army Fuels and

Lubricants Research  Laboratory (Moffit, 1972; Russel and Tosh,

1973; Tosh et al., 1975; Tosh, 1976).  Given the the broad assort-

ment and diverse ages of the vehicles involved, it seems likely

that many of these vehicles did not have hardened valve  seats.

     The results of  this test were negative — no untoward main-

tenance problems that could be attributed to the use of  unleaded

gasoline were experienced.  Overall, an engine failure rate of

0.5 percent was experienced.  This rate was stated as being

comparable to experience with leaded gasoline.  Only three cases

of valve-seat recession were reported, all in light-duty vehicles.

This is especially significant because the test was conducted

before  the imposition of the 55 mph speed limit, and many of the

posts were located in remote areas, so that considerable highway

driving would be expected.  The conclusions of the Army  study

are worth quoting:

         From the evaluation results, it can be concluded
         that commercial, tactical and combat vehicles,
         and all other equipment used in this program can
         operate satisfactorily during their normal day-
         to-day activities  without any fuel economy
         penalties and with no apparent increases in
         vehicle maintenance or operating costs so long  as
         unleaded gasoline  meeting VV-G-00169A Federal
         specification  is used.  (Tosh, 1976, p. 34; emphasis
         in original)

The Federal specification cited  is essentially that for  present-

day commercial unleaded gasoline.

-------
                                 VII-33
TABLE VII-7  Vehicle and Engine Types in U.S.

Commercial Vehicles
Army Unleaded Gasoline Test
Cars                                 445

Light-Duty Trucks

  0-6,000 pounds                   1,003
  6,000-10,000 pounds                429

Medium and Heavy-Duty Trucks

  10,000-14,000 pounds                68
  14,000-16,000 pounds                57
  16,000-19,500 pounds               163
  19,500-26,000 pounds                43
  26,000-33,000 pounds                63
  33,000 pounds plus                  28

Unclassified cars and trucks         411

Buses                                 87

Tractors                              84

Construction and Other Equipment

  Cranes                              38
  Graders                              5
  Fork Lifts                         256
  Generators                         527
  Miscellaneous Construction Equip.  255

Other Vehicles and Engines

  Scooters                            40
  Outboard motors                     41
  Lawn mowers                        225
  Motorcycles                          7
 Tactical Vehicles
 1/4 ton trucks*           2785
 3/4 ton trucks               8
 1 1/4 ton trucks           919
 Other tactical trucks       83
*These vehicles did not have
 hardened valve seats as of 1971;
 status of valve seats for other
 tactical vehicles unsure.

-------
                             VII-34






     In 1975, shortly after this test, all branches of the U.S.



Armed Services converted completely to unleaded gasoline wherever



it was available.  A monitoring system was set up to detect



subsequent problems and no special vehicle maintenance or other



problems were experienced since this conversion (M. DePara, U.S.



Army, Belvoir Research and Development Center, personal communi-




cation) .



     Data provided by the U.S. Postal Service tell a very similar



story for heavy-duty trucks in their service.  The Postal Service



has operated some 1,562 1975-model Ford heavy-duty trucks (22,000



pounds) on unleaded gasoline since 1980.   These trucks were



originally purchased in 1975, and travel  about 50,000 miles per



year, on average.  By 1980, most of them  were on their second or



third engine rebuild or replacement, so that there was a wide



variety of engine mileages -- from zero to about 100,000 miles --



represented  in the fleet.  All of the new and rebuilt engines in



the fleet used hardened valve seat inserts (as do most heavy-duty



trucks).



     In the  42 months or so since switching to unleaded, the



Postal Service has recorded 69 instances  of valve problems (a



valve failure rate of 4.4 percent) and 18 cases of valve-seat



problems (a  failure rate of 1.2 percent), while operating these



trucks for an average of about 175,000 miles each on unleaded



gasoline.  This would normally include at least one full engine



rebuild (M.  Sanders, U.S. Postal Service, personal communication).



For comparison, Ford indicated that its warranty data for the same



types of engines — presumably run primarily on leaded fuel —

-------
                            VII-35





showed higher valve and cylinder head failure rates  (Ford Auto-



motive Emissions Office, personal communication).  The Postal



Service has experienced no significant mechanical or operating



problems as a result of using lead-free gasoline in  its fleet.





VII.C.I.e.  Other Types of Engines



     The studies described above generally involved on-road



vehicles (cars and trucks), although the Army study also included



some construction equipment, stationary generators, motorcycles,



and outboard engines.  To investigate possible valve-seat damage



in smaller engines, such as those used in lawnmowers, snowblowers,



garden tillers, and snowmobiles, we contacted three manufacturers



of small engines (Briggs, Tecumseh, and Kohler).  All said either



that their engines could almost always use either leaded or



unleaded, or that they specifically recommend unleaded.  Repre-



sentatives from these companies also stated that they believed



that this would be true for all of the engines that they had



manufactured for at least the last 10 years, and were not aware



of any design changes that would have made this untrue even for



earlier engines.



     Marine engines are generally of two types:  inboard and



outboard.  Inboard engines typically are adapted from automobile



or truck engines, so we would expect the data on light-duty



vehicles to apply to them as well.  Outboard engines are almost



all two-stroke engines, for which the fuel is mixed with a special



type of oil.  For most such engines, unleaded does not appear to



cause any serious problems; however, for high-output two-stroke

-------
                            VII-36






engines (125 hp and above), cylinder-wall scoring and premature




bearing failures can occur (Weaver, 1984b).  Several solutions



are possible for this problem, if lead is banned.  One would be



to allow lead to remain in gasoline for marine use.  The other



would be to allow lead or another additive to be added to the oil



that is mixed with gasoline for two-stroke engine use.



     Another engine class of potential concern is farm equip-



ment.  Although diesel engines now dominate the market for tractors



and other large pieces of farm equipment, there are many older



gasoline-powered engines still in use on farms.  The Army tests



involved some farm as well as non-farm tractors that are likely



to be used under conditions similar to those on farms.  That



study also included portable generators, which should be similar



to many small engines used on farms to power stationary equipment.



To the extent that the Army data are applicable, it seems that



the Agency's action should not have a significant impact on the



durability of engines used on farms.






Vll.C.l.d.  Alternatives to Lead to Avoid Valve Recession



     As noted earlier, most engines manufactured over the last



decade have used induction-hardened valve seats, hardened valve-



seat inserts, or other mechanisms to eliminate potential valve-



seat recession problems without lead.  It is not feasible to modify



existing engines in those ways, however, except during major



engine rebuilds.  Thus, the most promising way of coping with



potential valve-seat durability problems in the total absence of



lead would be alternative additives.

-------
                              VII-37






     Relatively little research has been done on such alternatives,



presumably because there is little incentive to develop and market



them so long as lead remains available.  (Because lead is a rela-



tively cheap octane booster, it is a "free" valve lubricant.)



Limited work on the subject, however, suggests several possible



alternatives, the most promising of which is phosphorus.



     Several experiments suggest that phosphorus in unleaded



gasoline could reduce or eliminate the risk of valve recession at



high speeds.  Specifically, at about 0.06 or 0.07 grams of phos-



phorus per gallon, valve wear proceeds at one-half to one-third



the rate occurring with no additives (Giles and Updike, 1971;



Kent and Finnigan, 1971; Felt and Kerley, 1971; Wagner, 1973).



The tests were run primarily under unusually high loads or speeds,



similar to conditions used in the previously-described studies



of valve recession.



     Amoco (Wagner, 1971) reported that its road tests of heavily



loaded 1970-vintage cars, for 20,000 to 30,000 miles at average



speeds of 60 mph  (and up to 70 mph), found that 0.07 g/gal of



phosphorus was effective in controlling valve recession for nearly



90 percent of the cars tested.  The phosphorus more than halved



the rates of recession for the cars that, without lead or phos-



phorus, had experienced sinkage rates of more than 0.01 inches



per 10,000 miles.  Kent and Finnegan found, however, that at lower



load conditions and 2300 rpm for 80 hours, phosphorus was fully



protective against any valve-seat widening or oxidation.  Cordera



et al. (1964) found the presence of phosphorus in the gasoline was



critical to exhaust valve-life durability.  All of these results

-------
                              VII-38






indicated that adding phosphorus to unleaded gasoline would



substantially reduce the risk of valve recession for those vehicles



at risk.  Because phosphorus has a negative impact on catalysts,



however, it would be necessary either to have a special grade of



unleaded gasoline with phosphorus for older engines, or to make



phosphorus available as a separate additive.



     In considering a possible ban on leaded gasoline, EPA is



soliciting comments on phosphorus and other possible additives



to deal with potential valve-seat recession.  The Agency is



asking for comments also on other possible approaches, such as



making leaded gasoline available on a very limited basis (e.g.,



at marine terminals).






VII.C.2.  Megative Effects of Lead on Engine Durability



     Wear in engines may be due either to physical processes



(abrasive wear) or to chemical effects (corrosive wear).  Abrasive



wear results from the rubbing contact between two parts.  Corrosive



wear is a phenomenon akin to engine rusting — it occurs where



chemicals can attack a surface subject to wear, and either dissolve



it directly, or combine with it to form a less wear-resistant



material.



     Lead and its salts are effective solid lubricants.  Thus, it



might be expected that engine components exposed to lead deposits



might suffer less abrasive wear.  However, discussions with TRW



(H. McCormick, TRW Piston Ring Division, personal communication)



indicate that lead deposits may actually increase abrasive wear



of piston rings.  In addition, the acid combustion products of

-------
                            VII-39






lead scavengers contribute to corrosive wear, especially if water



is present.  Hudnall et al. (1969) have commented that corrosive



wear can be much greater than abrasive wear, especially in cold



operation.  Heavy-duty engines, however, which are less subject



to rusting due to their higher operating temperatures, also suffer



less from corrosive wear (Hudnall et al., 1969).



     Several investigators have compared the levels of wear



observed with leaded and unleaded gasolines.  Cordera et al.



(1964) compared wear results with the standard scavenger mixture



containing both chlorine (from EDC) and bromine (from EDB) with



wear using only bromine.  They found that eliminating the chlorine



reduced wear rates by about 40 percent.  In another test, they



examined the effects of different lead (and lead-scavenger)



concentrations on wear rates.  They found that going from 3.0 to



1.5 grams of lead per gallon reduced wear rates by around 40 per-



cent, with a small additional improvement at 0.5 grams per gallon.



Going to unleaded gasoline from 0.5 grams actually increased wear



rates, although wear was still lower than at 3 grams.  Gagliardi



and Ghannam (1969) obtained similar results in an 18,000-mile



fleet test.  They found that piston ring wear was lowest at 0.5



grams per gallon of lead, and increased slightly for both zero



and 1.5 grams.  Wear at 3 grams per gallon was 70 to 200 percent



greater than at 0.5 grams.



     The reduction in wear with low-lead gasoline is not



surprising, but the observed increase in wear when going from low-



lead to unleaded gasoline is.  One reasonable explanation for this



increase would be the solid lubricating effects of lead deposits

-------
                            VII-40






on the cylinder walls, which would be present with low-lead



gasoline but not with unleaded.  Alternatively, some differences



in combustion or lube-oil chemistry due to the presence of lead



might account for the difference in wear.  The available labora-



tory data give conflicting impressions as to the degree of in-



creased wear in changing from low-lead to unleaded gasoline.



Cordera and coworkers found a rather large increase, while



Gagliardi and Ghannem reported only a small effect.  To better



evaluate the magnitude of this effect in actual use, it is in-



structive to consider the results of in-use fleet testing.



     Orrin et al. (1972) and Carey et al. (1978) have reported



the results of two tests of leaded vs. unleaded gasoline in taxi



fleets.  One fleet operated in Oakland, California for 48,000



miles, and the other operated in Montreal, Canada for 80,000



kilometers.  These tests would be expected to favor leaded gaso-



line.  As taxis generally operate nearly continuously for 8 to 24



hours per day, they spend a comparatively small amount of time in



warm-up and cold operation -- the conditions that tend to favor



corrosion.  Despite this, the results of these tests showed a



distinct advantage for unleaded gasoline.



     In each case, wear measurements in taxis using leaded fuel



were 70 to 300 percent greater than those for taxis using unleaded.



Piston-ring and cylinder-bore wear, perhaps the most critical




areas, ranged from 70 percent to 150 percent greater with leaded



than with unleaded fuel.  Neither fleet experienced any overt



problems with rust, possibly indicating that significant corrosive



wear can occur even in the absence of visible rust.  Alternatively,

-------
                              VII-41






some of the difference might be due to increased abrasive wear



due to lead deposits, as suggested by TRW.



     These data, which closely match those of Gagliardi and



Ghannam (1969) in laboratory tests, appear to indicate that the



actual decrease in wear with unleaded gasoline is almost as



great as that found with low-lead fuel.  Since lead deposits



appear to form similarly at 0.1 and 0.5 grams per gallon, it is



probable that wear at 0.1 gram per gallon would be similar to or



lower than that at 0.5.  The data also indicate that the recent



reduction to 1.10 grams per gallon should have produced a signifi-



cant decrease in corrosive wear, at least with regular oil changes,



However, oil changes are frequently irregular, and blowby volume



in worn engines is much greater than in the new engines on which



these test were conducted.  Both of these factors would tend to



increase corrosive wear rates, even at the current 1.10 gram per



gallon level.  Thus, either lower-lead (0.5 and 0.1 gplg) or



unleaded fuel could be expected to produce a significant reduction



in wear rates from those observed with regular leaded gasoline,



even at 1.1 gram per gallon.



     The economic significance of reduced wear rates would be



considerable.  At present, worn-out piston rings and cylinder



bores are one of the major causes of failure in gasoline engines.



They result in poor fuel economy, poor performance, and increased



emissions.  Repairing this condition requires an engine overhaul,



at a cost of $500 to more than $1,000, depending on the engine.



Many older vehicles with these problems are simply junked and



not repaired.

-------
                              VII-42





     In new engines, the use of unleaded gasoline can extend



piston-ring lives significantly — by as much as a factor of two



(H. McCormick, TRW Piston Ring Division, personal communication).



It is not clear, however, how to extrapolate from this finding to



estimate the effects on the service lives of engines now in use.



If corrosive wear is the major factor in piston-ring wear, then



one would expect it to get worse over time as blowby rates increase



and, generally, maintenance practices degrade.   The general shift



to shorter trips and less annual mileage with increasing age



would also increase corrosion.  Thus, a car that had run for



half of the expected lifetime of its piston rings would probably



have accumulated somewhat less than half its lifetime wear, and



a radical decrease in wear might have more than proportional



benefits.



     On the other hand, if abrasive wear due to lead deposits is



the dominant factor, these deposits would last  for some time



after the switch to unleaded, and would thus result in less than



proportional increases in service life.  Overall, the effect of



unleaded gasoline in increasing the remaining service life of the



piston rings is probably best estimated as being linearly propor-



tional to the remaining life.  The typical service life for



piston rings in cars using leaded gasoline is about 70,000 to



80,000 miles.  A car driven 50,000 miles on leaded gasoline



could expect perhaps another 25,000 miles before needing an



overhaul.  Switching to unleaded would probably increase this



by 70 to 150 percent, giving a new expected time-to-overhaul of



about 43,000 to 67,000 miles.  Since parts other than piston

-------
                            VII-43





rings can fail, the actual increase in engine life would probably



be closer to 70 percent than 150 percent.  Because a substantial



part of the oil change benefits that have been monetized may



reflect reduced engine wear rather than longer oil change inter-



vals, we have not included any monetized benefits for this



category.





VII.C.3.  Summary of Engine Durability Effects



     We have made no attempt to monetize the potential engine



durability benefits or costs of reducing or eliminating lead in



gasoline.  The net impact of lead on engine durability is unclear.



For most engines, it appears that lead does substantially more



harm than good.  For some, however, lead may play an important



role in reducing the risk of valve-seat recession at high loads



and speeds, although tests of vehicles in use suggest that few



engines need lead under normal operating conditions.  It appears



that the low valve-seat wear rates in in-use fleets are due both



to the  low proportion of time spent at high rpm, and the ability



of engine oil additives to build up a protective coating during



the low rpm use which then protects the engine during intermittent



high speed operation.  The 0.10 gplg standard provides a margin



of safety to protect against potential recession at high loads



and speed since it also has been shown to build up a protective



layer during low and moderate rpm use.



     For the longer run, when EPA proposes to ban lead in gasoline,



several solutions may be possible for those few engines that need



protection against valve-seat recession.  First, phosphorus or



some other additive may prove to provide effective protection.

-------
                              VII-44






Second, it may be possible to make leaded gasoline available on a



very limited basis, so that its use is restricted to those engines



that truly need it.



     If acceptable alternatives are not developed, the Agency may



be forced to accept some increased risk of premature valve-seat



recession in some engines as the price of eliminating the severe



health and environmental consequences associated with lead in




gasoline.  As part of its continued deliberations on a possible



ban, the Agency will attempt to develop quantitative estimates



of the magnitude of this problem.

-------
                           CHAPTER VIII




       COST-BENEFIT ANALYSIS OF ALTERNATIVE PHASEDOWN RULES






     EPA considered many alternative phasedown schedules before



deciding on the final rule.  This chapter compares the costs and




benefits of those alternatives, based on the methods and results



described in earlier chapters.  Section A of this chapter sum-



marizes the estimates of benefits and costs, and compares them



to the estimates contained in the Preliminary Regulatory Impact



Analysis (RIA) issued when the rule was proposed in August 1984.



Section B presents the cost and benefit estimates for different




lead standards under various assumptions about the impact of the



rule on misfueling.  Section C examines the impact of the proposed



banking rule on the costs and benefits of the final phasedown



rule.  Finally, Section D summarizes the conclusions and EPA's



rationale for selecting the final phasedown schedule.



     Throughout this chapter, we present benefit estimates with



and without adult blood-pressure-related benefits.  The estimation



of a dose-response relationship between blood lead and blood pres-



sure is very recent.  The paper presenting those results (Pirkle



et al., 1985) has just been published in a peer-reviewed journal.



A summary of the results of that study and their application to



this rule was also placed in the docket for this rulemaking




several weeks before it closed for public comment.  Until the



scientific community has had an opportunity for more intensive



review, however, EPA is not relying on these results to reach




final regulatory decisions on lead in gasoline.  As the results



in this chapter show, these blood-pressure-related benefits

-------
                              VIII-2






greatly increase the total estimated benefits, but even when



they are not included the benefits of the rule exceed the costs




by a large margin.






VIII.A.  Summary of Cost and Benefit Estimates



     Table VIII-1 summarizes the cost and benefit estimates for



the 0.10 gplg standard in 1986; these estimates assume that the



rule would eliminate all misfueling.  As shown in the table, the




benefits total $7.9 billion, while the estimated cost is only



$607 million, resulting in net benefits of $7.3 billion.  About




75 percent of the estimated benefits are attributable to reduc-



tions in cardiovascular diseases associated with elevated blood



pressure.  Even if these benefits associated with adult health



are excluded from the calculation, however, the benefits still



exceed the costs by more than a three-to-one margin.



     Table VIII-1 also presents the cost and benefit estimates



contained in the Preliminary RIA that accompanied the proposed



rule.  The most striking difference, of course, is in adult blood



pressure benefits, which were not included in the Preliminary RIA.



     All of the other categories show some changes as well,



reflecting changes made in response to comments or to newly



available information.  The higher cost estimates reflect several



changes in base-case assumptions.  The two most important are:



reduced yields from reformers operated at high severity and



reduced segregation of naphthas for optimal allocation to pro-



cessing units.  Partly offsetting those changes are the use of




newer, more efficient catalysts in FCC units, and lower oil



prices, as discussed in Chapter II.

-------
                            VIII-3
TABLE VIII-1.  Costs and Monetized Benefits of 0.10 gplg  in 1986,
               Assuming No Misfueling:  Comparison of Current and
               Draft RIA Estimates (millions of 1983 dollars)

MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED
Current

602
5,927
278
933
190
7,930
Draft RIA

271
N.A.
348
840
360
1,819
    BENEFITS
TOTAL REFINING COSTS
  607
  575
NET BENEFITS
7,323
1,244
NET BENEFITS EXCLUDING
  BLOOD PRESSURE
1,396
1,244

-------
                              VIII-4






     The category of "Children's Health Effects" is higher for




several reasons.  First, the CDC recently reduced the blood lead



and FEP levels that define lead toxicity from 30 ug/dl to 25 ug/dl;



this greatly increases the number of children requiring at least



some follow-up medical testing or treatment.  Second, in estimat-



ing medical costs, we have relied on recently published recommend-




ations that call for more extensive testing and treatment than



assumed in the Preliminary RIA.  Finally, because of the change



in the CDC definition of lead toxicity and the greater weight



that the most recent draft of the Lead Criteria Document gives



to cognitive effects at blood lead levels in the range of 30 to



50 ug/dl blood lead, we have increased our estimate of the number



of children likely to warrant compensatory education in the



absence of further reductions in gasoline lead.



     As discussed earlier, the estimates of the benefits of



reduced emissions of conventional pollutants are the average of



two estimation methods:  the value implied by the cost of the



pollution control equipment destroyed by misfueling and a direct



valuation of some of the health and welfare effects associated



with these pollutants.  (The second method, direct valuation,



is based on an incomplete quantification of these health and




welfare effects.)  The reduction in the overall estimates com-



pared to the Preliminary RIA reflects a decrease in the direct



estimate, as discussed in Chapter VI.




     The changes in the maintenance estimates reflect refinements



in the fleet model used to estimate the number of miles traveled

-------
                              VIII-5





by different classes of vehicles.  The reduction in the fuel-



economy estimate is caused by changes in the predicted fuel



density, which in part results from changes in the inputs to the



refinery model.



     Table VIII-2 presents some important non-monetary measures



of the estimated benefits of the 0.10 gplg standard in 1986



(again, assuming that the rule eliminates all misfueling).  Note



that we estimate that the rule will reduce by 172,000 the number



of children above the new CDC blood-lead limit of 25 ug/dl.



The reductions in the numbers of children at lower, but still



possibly harmful, blood-lead levels are even greater; we estimate



that 1.7 million fewer children will experience blood lead levels



over 15 ug/dl  in 1986 as a result of the rule.



     If the rule eliminates misfueling, we estimate that it will



eliminate over 2.5 million tons of excess emissions of HC, NOX,



and CO.  The current estimates are higher than those made in the



preliminary RIA because we have used the results of the 1983 EPA



tampering and misfueling survey, which show higher rates than the



1982 survey employed earlier, and because of refinements in the



fleet model used to estimate the number of miles driven by dif-



ferent classes of vehicles.



     Table VIII-2 also reports estimates of the numbers of reduced



health effects among adults in 1986.  As already discussed, these



estimates are  restricted to males aged 40 to 59, and the estimates



for myocardial infarctions, strokes, and deaths apply only to



white males in that age range.  Despite these limitations, the



estimated benefits are large, ranging from 1.8 million fewer



cases of hypertension to 5,160 fewer deaths from all causes.

-------
                            VIII-6
TABLE VIII-2.  Nonmonetary Measures of Health and Environmental
               Benefits of 0.10 gplg in 1986, Assuming No Mis-
               fueling:  Comparison of Current and Draft RIA
               Estimates

Reductions in thousands of
children above selected
blood lead levels
30 ug/dl
25 ug/dl
20 ug/dl
15 ug/dl
Reductions in thousands of
tons of pollutants
Hydrocarbons
Nitrogen oxides
Carbon monoxide
Reductions in adult male
health effects
Thousands of
hypertensives
Myocardial
infarctions
Strokes
Deaths
Current


52
172
563
1,726

305
94
2,116


1,804
5,350
1,115
5,160
Draft RIA


52
172
563
1,726

247
81
1 ,646


N.A.
N.A.
N.A.
N.A.

-------
                              VIII-7





VIII.E.  Comparisons of Alternative Lead Levels



     In the August 1984 proposal, EPA discussed a range of alterna-



tive schedules and presented two specific possibilities:  (1) a



one-step reduction to 0.10 gplg starting January 1, 1986 and (2)



a phasedown with several steps — 0.50 gplg on July 1, 1985;



0.30 gplg on January 1, 1986; 0.20 gplg on January 1, 1987;  and



0.10 gplg on January 1, 1988.  The final rule imposes the 0.10



gplg as of January 1, 1986, and also requires a reduction to



0.50 gplg as of July 1, 1985.  In addition to these three sched-



ules, however, the Agency considered many other possibilities.



     For 1986, the Agency considered levels between 0.1 gplg and



0.5 gplg.  Table VIII-3a presents the cost and benefit estimates



for those alternatives, assuming in each case that all misfueling



would be eliminated.  Net benefits are maximized at the tightest



of those limits, 0.10 gplg, whether or not adult blood pressure



benefits are included.  Table VIII-3b shows the estimates assum-



ing that the rule fails to have any impact on misfueling; again,



net benefits are maximized at 0.10 gplg.



     Neither of these two extreme assumptions about misfueling



(that it will be eliminated entirely or that it will continue



unabated, even at very low lead levels) appears realistic.  EPA



believes it is more likely that the number of misfuelers is a



declining function of the lead level, primarily because the



manufacturing cost differential between leaded and unleaded



declines and then reverses as the lead limit is tightened.  We



expect changes in manufacturing costs to be at least partly



reflected in retail prices.  Although we cannot be certain that

-------
                            VIII-8
TABLE VIII-3a.  Costs and Monetized Benefits of Alternative Lead
                Levels in 1986, Assuming No Misfueling
                (millions of 1983 dollars)

0.50
Lead
0.40
Level (
0.30
Igpig)
0.20

0.10
MONETIZED BENEFITS

   Children's health effects     466    504    539    571    602

   Adult blood pressure        4,018  4,483  4,955  5,436  5,927

   Conventional pollutants       278    278    278    278    278

   Maintenance                   517    608    706    808    933

   Fuel economy                  119    119    128    136    190

   TOTAL MONETIZED             5,398  5,992  6,606  7,229  7,930
    BENEFITS


TOTAL REFINING COSTS             243    305    386    472    607


NET BENEFITS                   5,155  5,687  6,220  6,757  7,323
NET BENEFITS EXCLUDING         1,137  1,204  1,265  1,321  1,396
 BLOOD PRESSURE

-------
                            VIII-9
TABLE VIII-3b.  Costs and Monetized Benefits of Alternative Lead
                Levels in 1986, Assuming Full Misfueling
	(millions of 1983 dollars)	

                                 	Lead Level (gplg)	
                                 0.50   0.40   0.30   0.20   0.10
MONETIZED BENEFITS

  Children's health effects       403    455    504    550    592

  Adult blood pressure          3,328  3,920  4,526  5,144  5,778

  Conventional pollutants          00000

  Maintenance                     186    329    482    642    838

  Fuel economy                     44     97     97    115    177

  TOTAL MONETIZED BENEFITS      3,961  4,801  5,609  6,451  7,385


TOTAL REFINING COSTS              178    260    350    460    627


NET BENEFITS                    3,783  4,541  5,259  5,991  6,758
NET BENEFITS EXCLUDING
  BLOOD PRESSURE                  455    621    733    847    980

-------
                              VIII-10





unleaded will be priced below leaded at the retail level even if



its manufacturing cost is lower, we do expect the price differential



between leaded and unleaded to decline and that the decline will



cause some misfuelers to switch to unleaded gasoline.



     Figure VIII-1 plots three possibilities for how the percent-



age of current misfuelers might decline as a function of the



lead limit.  Each of the three curves assumes that 50 percent of



misfueling would be eliminated at 0.25 gplg, the estimated point



at which the manufacturing costs of leaded and unleaded intersect.



They differ, however, in the assumed rates at which misfueling



changes.  Curve A assumes that misfueling declines linearly from



0.50 gplg to 0 gplg; at 0.10 gplg, misfueling is 20 percent of



its base level.  Curve B also assumes that the decline in misfuel-



ing begins at 0.50 gplg and ends at 0 gplg, but that the rate of



decline is most rapid over the intermediate range.  Curve C is



similar to B, but the change is compressed to the range 0.40



gplg to 0.10 gplg; it assumes no misfueling at 0.10 gplg.



     Figure VIII-2 plots the net benefits of the alternative lead



limits for all five assumptions about misfueling:  no misfueling,



full misfueling, and the three intermediate cases.  In all five



cases, not surprisingly given the results in Tables VIII-3a and



VIII-3b, the net benefits peak at 0.10 gplg.  Figure VIII-3



presents similar estimates, but excludes the adult blood pressure



benefits; again, net benefits peak at 0.10 gplg, though they are



much smaller than with the inclusion of the adult health effects.



     It is impossible to determine which assumption about

-------
c
0)
D
«*-
V)
c

-------
 0)

 o
.a
S.X

V)
!Z
«4-
0)
c
0)
00
         8



         7



         6
         5 -
4 -
3 -
        2 -
         1  -
          0.5
                        No Misfueling
                                           Full Misfueling
                     0.4
         0.3

Lead Content (gplg)
0.2
                                                                                        <
                                                                                        M
                                                                                        I
                                                                                        M
                                                                                        ro
0.1
            FIGURE VIII-2
                    Net Benefits (Including Blood-Pressure-Related Effects)  as
                    Functions of Lead Level and Misfueling

-------
 CO
 c
 o
-O
N^

 OT
.±i
»*-
 o>
 c
 0)
CO

•M
 0)
                                                                 Partial Misfueling
Full Misfueling
                                                 H
                                                 U)
                                                 0.3


                                        Lead  Content (gplg)

            FIGURE VIII-3.  Net Benefits (Excluding Blood-Pressure-Related Effects) as
                            Functions of Lead Level and Misfueling
                                       0.1

-------
                             VIII-14






misfueling is most accurate.  To do so would require knowing how




gasoline retail prices will respond to changes in production



costs, and how misfuelers will respond to changes in prices.



Thus, most of our estimates in this chapter and the next one



present both extreme possibilities with respect to misfueling.



     It is useful, however, to have a standard "partial misfuel-



ing" case for making comparisons.  For that case, we have settled



on the simplest of the three curves in Figure VIII-1, curve A,



which assumes that misfueling declines linearly from 100 percent



of its current level at 0.50 gplg and above, to zero at 0 gplg.



Under that assumption, 20 percent of the misfuelers continue to



use leaded gasoline at 0.10 gplg.  We believe this is a reasonable



estimate, as some gasoline stations are likely to continue to



sell leaded gasoline at a lower price than unleaded, and some



misfuelers may continue to buy leaded even if it costs more than



regular unleaded, either because they desire higher octane or



because they mistakenly believe that leaded gasoline is better



for their engines.  Table VIII-3c presents the cost and benefit



estimates under this assumption, i.e., that misfueling declines



from 100 percent of its current level at 0.50 gplg to 20 percent



at 0.10 gplg.




     For 1987, EPA considered two alternative levels:  0.20 gplg



and 0.10 gplg.  Tables VIII-4a through VIIl-4c present the cost



and benefit estimates for the no-misfueling, full-misfueling, and



partial-misfueling cases, respectively.  As in the earlier tables,



net benefits are maximized at 0.10 gplg, whether or not adult



blood pressure benefits are included.

-------
                           VIII-15
TABLE VIII-3c.  Costs and Monetized Benefits of Alternative  Lead
                Levels in 1986, Assuming Partial Misfueling
	(millions of 1983 dollars)	

                                 	Lead Level  (gplg)	
                                 0.50    0.40   0.30   0.20   0.10
MONETIZED BENEFITS

  Children's health effects

  Adult blood pressure

  Conventional pollutants

  Maintenance

  Fuel economy

  TOTAL MONETIZED BENEFITS


TOTAL REFINING COSTS


NET BENEFITS
  403    465    518    563    600

3,328  4,033  4,698  5,319  5,897

   0      56    111    167    222

  186    385    572    742    914

   44    101    109    128    187

3,961  5,039  6,008  6,918  7,821
  178
269
364
467
608
3,783  4,770  5,643  6,451  7,213
NET BENEFITS EXCLUDING
  BLOOD PRESSURE
  455
738
946  1,131  1,316

-------
                           VIII-16
TABLE VIII-4a.  Costs and Monetized Benefits of Alternative
                Lead Levels in 1987, Assuming No Misfueling
                (millions of 1983 dollars)

MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED
BENEFITS
TOTAL REFINING COSTS
NET BENEFITS
Lead Level
0.20

522
5,262
278
775
132
6,968
452
6,516
(gpig)
0.10

550
5,707
278
880
175
7,590
553
7,037
NET BENEFITS EXCLUDING
  BLOOD PRESSURE                1,255             1,330

-------
                           VIII-17
TABLE VIII-4b.  Costs and Monetized Benefits of Alternative
                Lead Levels in 1987, Assuming Full Misfueling
                (millions of 1983 dollars)

MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED
BENEFITS
TOTAL REFINING COSTS
NET BENEFITS
Lead Level
0.20

501
4,940
0
596
106
6,143
441
5,702
(gpig)
0.10

539
5,543
0
111
150
7,009
578
6,431
NET BENEFITS EXCLUDING
  BLOOD PRESSURE                  762               888

-------
                           VIII-18
TABLE VIII-4C.  Costs and Monetized Benefits of Alternative
                Lead Levels in 1987, Assuming Partial
                Misfueling (millions of 1983 dollars)

MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy
TOTAL MONETIZED
BENEFITS
TOTAL REFINING COSTS
NET BENEFITS
Lead Level
0.20

513
5,133
167
703
122
6,638
448
6,191
(gpig)
0.10

547
5,675
222
859
170
7 ,474
558
6,916
NET BENEFITS EXCLUDING
  BLOOD PRESSURE               1,058             1,241

-------
                             VIII-19





     For 1985, EPA considered five alternative levels:  1.10 gplg



(i.e., no change), 0.80 gplg, 0.70 gplg, 0.60 gplg, and 0.50 gplg.



Table VIII-5 presents the estimated costs and benefits, assuming



full misfueling.  (Estimates are not presented for the no-misfuel-



ing case, as we doubt that standards above 0.50 gplg will have



enough impact on the price of leaded gasoline to make a signifi-



cant difference in misfueling.)  Note that the estimates cover



only half a year, as none of the standards considered for 1985



would take effect until the middle of that year (July 1).  Again,



net benefits are maximized at the tightest standard discussed in



the August proposal, 0.50 gplg.



     The net benefits of the 0.50 gplg standard in 1985 are



substantial:  $264 million if blood-pressure-related benefits are



not included, and $2.0 billion if they are.  Moreover, as discussed



in Chapter II, all available measures indicate that the refining



industry can comply easily with that portion of the rule; reducing



lead to 0.50 gplg should require minimal adjustments in refinery



operations.



     Table VIII-6 compares the present values of the costs and



benefits of three phasedown schedules over the period 1985-1987:



the original primary proposal; the more gradual, illustrative



phasedown presented in the Notice of Proposed Rulemaking; and



the schedule in the final rule.  In all cases, the costs and



benefits have been discounted at 10 percent (real) to the begin-



ning of 1985.  (The 1985 estimates were discounted for half a



year, the 1986 benefits for a full year, and the 1987 benefits



for two years.)  All of the schedules yield substantial

-------
                           VIII-20
TABLE VIII-5.  Costs and Monetized Benefits of Alternative
               Lead Levels in 1985, Assuming Full Misfueling
               (millions of 1983 dollars)

MONETIZED BENEFITS
Children's health effects
Adult blood pressure
Conventional pollutants
Maintenance
Fuel economy

0.80

124
837
0
37
-5
Lead Level
0.70

159
1,126
0
54
0
(gpig)
0.60

193
1,423
0
80
31

0.50

223
1,724
0
102
35
  TOTAL MONETIZED
    BENEFITS
993
1,339
1,727
2,084
TOTAL REFINING COSTS

NET BENEFITS

NET BENEFITS EXCLUDING
  BLOOD PRESSURE
 44       56       75       96

949    1,283    1,652    1,988
112
  157
  229
  264

-------
                           VIII-21
TABLE VIII-6.  Present Values of Costs and Monetized Benefits:
               Comparison of Proposed, Alternative, and Final
               Schedules for 1985-1987 (millions of 1983 dollars)

With No Misfueling
Costs
Benefits
Net benefits
Net benefits, excluding
blood pressure
With Full Misfueling
Costs
Benefits
Net benefits
Net benefits, excluding
blood pressure
With Partial Misfueling
Costs
Benefits
Net benefits
Proposed
1,009
13,482
12,473
2,368
1 ,048
12,506
11.459
1,625
1,014
13,287
12,273
Alternative
845
14,377
13,532
2,743
774
12,160
11,386
1,547
793
12,932
12,139
Final
1,130
16,095
14,965
2,924
1,139
14,490
13,351
1,876
1,105
15,271
14,166
  Net benefits, excluding
    blood pressure              2,222       1,985       2,473

-------
                             VIII-22






net benefits, in excess of $11 billion with blood-pressure-



related benefits and over $1.5 billion without them.  Compared



to the other schedules, the final rule has higher costs but even



higher benefits (whether or not blood-pressure-related benefits



are included), with the result that it has the highest net bene-




fits of the three schedules.



     Tables VIII-7a through VIII-7c present year-by-year estimates



of the costs and benefits of the final rule over the period 1985




to 1992, under the three assumptions about misfueling.  The costs



fall from 1986 to 1992 because we project that the demand for



leaded gasoline would fall even in the absence of the rule, as a



result of retirement of older cars.  For that same reason, most



of the estimated annual benefits also decline over time.  The



major exception is conventional pollutants, because we expect the



amount of misfueling to increase in the absence of the rule, as



the number and average age of catalyst-equipped vehicles increase.



     Table VIII-8 shows the present values of the final rule



under the different assumptions about misfueling.  The estimated



net benefits, not including blood-pressure-related benefits, range



from $4.1 billion if the rule has no impact on misfueling, to



$6.7 billion if the rule eliminates all misfueling.  Under the



more realistic "partial misfueling" assumption, the present value



of the net benefits is $5.9 billion.  If the adult blood pressure



benefits are included, the present value of the net benefits is



much higher, $33.4 billion under the partial misfueling assumption,

-------
                                     VIII-23
TABLE VIII-7a.  Year-by-Year Costs and Monetized Benefits of Final Rule, Assuming
	No Misfueling (millions of  1983 dollars)	

	1985    1986     1987     1988     1989     1990     1991     1992

MONETIZED BENEFITS

  Children's
    health effects     251     602     550     504      455     417      371     361
  Adult blood
    pressure         2,033   5,927    5,707    5,484   5,227   5,008   4,722   4,736


  Conventional
    pollutants          140      278      278      280      282     288     299     310


  Maintenance           252      933      880      840      811     792     780     776


  Fuel economy           68      190      175      124      144     145     182     175
  TOTAL MONETIZED
    BENEFITS          2,744    7,930    7,590    7,232   6,919   6,649   6,354   6,358
 TOTAL REFINING
  COSTS                 127      607      553      530     502     468     442     440


 NET BENEFITS          2,617    7,323    7,037    6,702    6,417   6,181   5,912   5,918
NET BENEFITS
  EXCLUDING BLOOD
  PRESSURE              584    1,396    1,330    1,218    1,190    1,174    1,190    1,182

-------
                                     VIII-24
TABLE VIII-7b.  Year-by-Year Costs and Monetized Benefits of Final Rule, Assuming
	Full Misfueling (millions of 1983 dollars)	

	1985    1986    1987    1988    1989    1990     1991     1992

MONETIZED BENEFITS

  Children's
    health effects    223     592     539     494     445     406      361      350
  Adult blood
    pressure        1,724   5,778   5,543   5,303   5,031   4,798   4,521   4,512
  Conventional
    pollutants
  Maintenance         102     838     777     730     694     668     650     640
  Fuel economy         35     177     150      70      96     113     131     122
  TOTAL MONETIZED
    BENEFITS        2,084   7,385   7,009   6,597   6,266   5,985   5,663   5,624
TOTAL REFINING
  COSTS                96     627     578     539     514     485     451     443


NET BENEFITS        1,988   6,758   6,431   6,058   5,752   5,500   5,212   5,181


NET BENEFITS
  EXCLUDING BLOOD
   PRESSURE           264     980     888     755     721     702     691     669

-------
                                     VIII-25
TABLE VIII-7c.  Year-by-Year Costs and Monetized Benefits  of Final  Rule, Assuming
	Partial Misfueling (millions of 1983 dollars)	

	1985    1986     1987    1988    1989    1990     1991     1992

MONETIZED BENEFITS

  Children's
    health effects    223      600     547      502      453      414      369      358
  Adult blood
    pressure         1,724   5,897    5,675    5,447    5,187   4,966   4,682   4,691
  Conventional
    pollutants          0      222      222      224      226      230     239     248


  Maintenance          102      914      859      818      788      767     754     749


  Fuel economy          35      187      170      113      134      139     172     164
  TOTAL MONETIZED
    BENEFITS         2,084    7,821    7,474    7,105    6,788   6,517   6,216   6,211
 TOTAL REFINING
  COSTS                 96      608      558      532      504     471      444     441


 NET BENEFITS         1,988    7,213    6,916    6,573    6,284    6,045    5,772    5,770
NET BENEFITS
  EXCLUDING BLOOD
  PRESSURE             264    1,316    1,241    1,125    1,096    1,079    1,090    1,079

-------
                           VIII-26
TABLE VIII-8.  Present Values of Costs and Benefits of Final
	Rule, 1985-1992 (millions of 1983 dollars)

                           No         Full      Partial
	Misfueling  Misfueling  Misfueling

MONETIZED BENEFITS

  Children's
    health effects        2,582       2,506       2,546

  Adult blood
    pressure             27,936      26,743      27,462

  Conventional
    pollutants            1,525           0       1,114

  Maintenance             4,331       3,634       4,077

  Fuel
    economy                 856         643         788

  TOTAL
    MONETIZED
    BENEFITS             37,231      33,526      35,987
TOTAL REFINING
  COSTS                   2,637       2,678       2,619

NET BENEFITS             34,594      30,847      33,368

NET BENEFITS EXCLUDING
  BLOOD PRESSURE          6,658       4,105       5,906

-------
                             VIII-27






VIII.C.  Impact of Banking on Costs and Benefits of Final Rule



     The analysis thus far in this chapter has assumed that



refineries will follow the phasedown schedule being promulgated



in this final rule.  EPA has proposed the use of "banking" of



lead rights during 1985 for use in 1986 and 1987.  Under banking,



refineries would have the option of using less lead than allowed



in 1985 and "banking" it for use in 1986 or 1987.  As discussed



in Chapter II, this provision would give individual refineries



extra flexibility in reducing lead, without increasing the total



allowable amount of lead in gasoline between 1985 and 1987.



Although refiners would be under no obligation to use the right



to bank, EPA expects that most would, because the marginal value



of lead to refiners will be higher in 1986 and 1987 than in 1985,



for the reasons discussed in Chapter II.  This section briefly



examines the impact that banking may have on the costs and bene-



fits of the phasedown rule.  This analysis is for illustrative



purposes only, as EPA has determined that the final rule is



feasible and not unduly expensive without banking, and should



be promulgated independent of the banking rule.



     Table VIII-9 compares the schedule without banking to the



two possible alternatives with banking that were presented in



Chapter II.  Both alternatives are based on the partial misfuel-



ing assumption; i.e., that misfueling continues unabated through



1985 and is reduced by 80 percent in 1986 and 1987.



     Alternative 1 assumes that refiners would not start banking



until the second quarter of 1985, at which point they would use



an average of 0.60 gplg, thus banking 0.50 grams, on average, for

-------
                           VIII-28



TABLE VIII-9.  Alternative Phasedown Patterns with Banking (gplg)

                       1985(by quarter       1986        1987
Alternative	I    II III-IV	

Without Banking      1.10  1.10  0.50        0.10        0.10

With Banking

  Alternative 1      1.10  0.60  0.40        0.25        0.19

  Alternative 2      0.80  0.60  0.45        0.30        0.21

-------
                             VIII-29


each gallon of leaded produced.*  Under alternative 1, we also

assume some banking in the last half of the year, with leaded

gasoline averaging 0.40 gplg, slightly below the limit of 0.50

gplg.  A total of 7.0 million grams of lead (about 22 percent

of the total allowed in 1985) would be banked during 1985, allow-

ing refiners to average 0.25 gplg in 1986 and 0.19 gplg in 1987-

Note that shaving 0.10 gplg from the annual average in 1985

translates into a larger per gallon increase in 1986 or 1987,

because the amount of leaded gasoline produced in the later years

is smaller; the total amount of lead use allowed over the three

years, however, would be the same as without banking.

     Alternative 2 assumes that some refiners would be able to

reduce lead more quickly, so that banking would begin in the

first quarter of 1985.  Compared to Alternative 1, those extra

banked rights from the first quarter would be used to reduce

banking slightly in the last half of 1985 and to achieve slightly

higher lead levels in 1986 and 1987.  The amount banked would be

9.1 million grams.  Again, the total amount of lead used would

be the same as without banking.

     Table VIII-10 presents the year-by-year cost and benefit

estimates for the two alternatives with banking.  Compared to

the estimates without banking (see Table VIH-7c), the benefits

and costs would be higher in 1985, but lower in 1986 and 1987.
* Note that in California, where a state-imposed rule will limit
  leaded gasoline to 0.8 gplg in the first half of 1985, refiners
  producing 0.60 gplg leaded gasoline would only be able to bank
  0.20 gplg in the second quarter; refiners would not be allowed
  to take credit for reductions mandated by state laws.

-------
                                     VIII-30


TABLE VIII-10.  Costs and Monetized Benefits of Alternative (Banking) Phasedown
                Patterns, Assuming Partial Misfueling (millions of 1983 dollars)
Alternative 1*
1985
MONETIZED BENEFITS
Children's
health effects 347
Adult blood
pressure 2,745
Conventional
pollutants 0
Maintenance 110
Fuel economy 95
TOTAL MONETIZED
BENEFITS 3,297
TOTAL REFINING
COSTS 176
NET BENEFITS 3,121
NET BENEFITS
EXCLUDING BLOOD
PRESSURE 376
1986

550
5,122
222
716
123
6,733
420
6,313
1,191
1987

522
5,244
222
752
126
6,866
463
6,403
1,159
Alternative 2*
1985

395
3,008
0
97
87
3,587
170
3,417
409
1986

532
4,870
222
661
114
6,399
378
6,021
1,151
1987

515
5,162
222
732
126
6,757
452
6,305
1,143
  *See Table VIII-9 for description of  alternatives.

-------
                             VIII-31

Table VIII-11 compares the present values of the costs and bene-
fits with banking to those without.  In present value terms,
banking reduces refining costs by $173 million under alternative
1 and $226 million under alternative 2, a reduction of over 20
percent.  As discussed in Chapter II, this is likely to be an
underestimate of the real savings that can be achieved with bank-
ing, as it does not account for the extra flexibility banking
allows in meeting unexpected problems.  It is also interesting
to note that the present value of the cost of the final rule with
banking is actually lower than the cost of the August proposed
rule without banking (cf. Table VIII-6), despite the fact that
the final rule eliminates significantly more lead.
     Banking has a slight negative impact on the present value of
estimated benefits.  This reduction reflects several factors.  The
most important of these is that our estimates of the relationship
between blood pressure and blood lead employs the logarithm of
blood lead; thus it predicts slightly higher benefits from concen-
trating the reduction in lead in 1986 and 1987 rather than
spreading it over 1985-1987.  The same holds true for mainten-
ance benefits, because exhaust system corrosion appears to fall
most sharply as lead is reduced at low levels.  In contrast,
children's health benefits are higher with banking than without,
primarily because the lead reductions are achieved earlier with
banking.
     Overall, banking has virtually no impact on net benefits.
The small differences shown in Table VIII-11 are well within the
"noise" of the estimates and should be regarded as insignificant.

-------
                           VIII-32
TABLE VIII-11.  Present Values of Costs and Benefits of
                Alternative Phasedown Patterns, 1985-87,
                Assuming Partial Misfueling
                (millions of 1983 dollars)

MONETIZED BENEFITS
Children's
health effects
Adult blood
pressure
Conventional
pollutants
Maintenance
Fuel
economy
TOTAL
MONETIZED
BENEFITS
Without
Banking

1,210
11,693
385
1,638
344
15,271
With
Alt. 1

1,262
11,605
385
1,377
307
14,936
Banking
Alt. 2

1,285
11,558
385
1,298
291
14,818
TOTAL REFINING
  COSTS                   1,105         932         879

NET BENEFITS             14,166      14,007      13,939

NET BENEFITS EXCLUDING
  BLOOD PRESSURE          2,473       2,399       2,381

-------
                              VIII-33







Coupled with the important, but nonmonetized, cost saving that



banking provides in the form of flexibility to meet unexpected



refining problems, these results indicate that banking is a



desirable policy.





VIII.D.  Summary



     The results presented in this chapter indicate that the



final phasedown rule being promulgated has the highest net



benefits of the alternatives considered.  This conclusion holds



whether or not the recently developed estimates of blood-



pressure-related benefits are included, and whether or not it is



assumed that the rule will eliminate misfueling.  Although many



of the estimates are subject to uncertainty, the magnitude of the



estimated monetized benefits relative to the costs indicates that



this conclusion is very robust.  Moreover, the monetary benefit



estimates represent an incomplete tabulation of the benefits



likely to result from the rule; in short, the benefits are



under-estimated.



     Three limitations deserve particular notice.  First, we have



not estimated any benefits for children at blood lead levels



below the CDC cutoff of 25 ug/dl, and our monetized estimates for



children above that level cover only medical and compensatory



education costs.  Second, the direct estimates of benefits



associated with reduced conventional pollutants omit several



important categories, including benefits associated with ozone's



effects on nonagricultural vegetation, chronic health effects



related to ozone, and the effects of nitrogen oxides and carbon

-------
                             VIII-34








monoxide on health.  Last, the estimates of adult health benefits



cover only blood-pressure-related effects, and only males aged



40-59.  In the case of myocardial infarctions, strokes, and



deaths, only whites males in this age range are included.  (This



is because of limited data on the cardiovascular risks associated



with high blood pressure in nonwhites.)   These unquantified bene-



fits add further strength to the conclusion that rapid reductions



of lead in gasoline are amply justified.

-------
                            REFERENCES


Abell RF. Operation and application of hydraulic valve lifters.
  SAE Paper #690347, Society of Automotive Engineers, Warrendale,
  PA 1969.

Adams R, McCarl B. Assessing the benefits of alternative oxidant
  standards on agriculture:  the role of response information.
  Prepared for U.S. EPA, 1983 September.

Adams R, et al. Economic effects of ozone on agriculture.
  Final Report to U.S. EPA, 1984.

Adams WC, Schelegle ES. Ozone toxicity effects consequent to
  prolonged, high intensity exercise in trained endurance
  athletes.  J App Physiol Respir Environ Exercise Physiol
  1983; 55:  805-12.

Agerty HA. Lead poisoning in children. Med Clin North Am 1952; 36:
  1587-97.

American Petroleum Institute. Unpublished Carcinogenicity Study
  on Unleaded Gasoline in Vapor in Fischer 344 Rats and B6C3F1
  Mice. 1982.

Amoco, submission to EPA Central Docket EN-84-05.

Anderson E, et al. Effect of low-level carbon monoxide exposure
  on onset and duration of angina pectoris:  a study on 10 patients
  with ischemic heart disease. Annals of Internal Medicine 1973;
  79: 46-50.

Anderson G, et al. Development and application of methods for
  estimating:  effects of industrial emission controls on air
  quality impact of reactive pollutants.  U.S. EPA 1984 July.

Angle CR, Mclntire MS, Swanson MS, Stohs SJ. Erythrocyte
  nucleotides in children - increased blood lead and
  cytidine triphosphate. Pediatric Research 1982; 16: 331-4.

Annest JL, Pirkle JL, Makuc D, Neese JW, Bayse DD, Kovar MG.
  Chronological trends in blood lead levels between 1976 and
  1980. New England Journal of Medicine 1983; 308: 1373-7.

Armstrong JA. Investigations into the emissions effects of
  vehicle misfueling.  Presented to: The North American Motor
  Vichicle Emission Control Conference.  New York, 1984 April
  2.

Aub JC, Fairhall Lt, Minot AS, Reznikoff P, Hamilton A. Lead
  Poisoning in Baltimore, MD.  The Williams and Wilkins Company
  (Medicine monographs:  V.7.)

-------
                                -2-
The Australian Therapeutic Trial in Mild Hypertension:
  Report by the Management Committee.   Lancet 1980;  2:  1261-7.

Barrett LB, Waddell T.  The Cost of Air Pollution Damages:  A
  Status Report.  U.S. EPA; Research Triangle Park,  NC  1973

Barton JC, Conrad ME, Nuby S,  Harrison L. Effects of
  iron on the absorption and retention of lead.   Journal of
  Laboratory and Clinical Medicine 1978; 92:  536-47.

Bascunana J, Stahman R. Impact of gasoline characteristics on
  fuel economy and its measurement.  Report 76-10,  Technology
  Assessment and Evaluation Branch, ECTD, EPA,  1976  December.

Bates D, Sizto R. Relationship between air pollution levels and
  hospital admissions in southern Ontario.  Canadian J  of Pub
  Health 1983 March/April; 79-

Batuman V, Landy E, Maesaka JK, Wedeen RP. Contribution of lead to
  hypertension with renal impariment.  N Engl  J  Med  1983; 309:
  17-21.

Beevers DG, Erskine E, Robertson M, Beattie AD,  Campbell
  BC, Goldberg A, Moore MR, Hawthorne VM. Blood-lead and
  hypertension. Lancet 1976; 2(7975):  1-3.

Beevers DG, Cruickshank JK, Yeoman WB, Carter GF, Goldberg A,
  Moore MR. Blood-lead and cadmium in human hypertension.
  J Environ Pathol Toxicol 1980; 4: 251-60.

Bell AG, Keene JA, Reders K. Road antiknock performance of low-
  lead and nonleaded gasoline  in european cars.  SAE  Paper #710625,
  Society of Automotive Engineers, Warrendale,  PA 1971.

Bellinger D, Needleman H, Leviton A, Waternaux C, Rabinowitz M,
  Nichols M.  Early sensory-motor development and prenatal exposure
  to lead.  Neurobehavioral Toxicology and Teratology 1984; 6:
  387-402.

Benignus VA, Otto DA, Muller KE, Seiple KJ. Effects  of  age and
  body lead burden on CNS function in young children:  II.
  EEG Spectra. Electroencephalog Clin Neurophysiol  1981;
  52:240-8.

Bennett PA. Engine rusting in  service and dynamometer tests.
  Lubrication Engineering 1960 November: 529.

Benson JD. Some factors which  affect octane requirement increases.
  SAE Paper #750933, Society of Automotive Engineers, Warrendale,
  PA 1975.

-------
                                -3-


Benson JD. Manganese fuel additive (MMT) can cause vehicle prob-
  lems.  SAE Paper #770655, Society of Automotive Engineers,
  Warrendale, PA 1977.

Bernard JP. Short-trip engine oil rust testing. SAE Paper #710839,
  Society of Automotive Engineers, Warrendale, PA 1971.

Bert JA, Gething JA, Hansel TJ, Newhall HK, Peyla RJ, Voss PA.
  Gasoline additive concentrate removes combustion chamber depos-
  its and reduces vehicle octane requirement. SAE Paper #831709,
  Society of Automotive Engineers, Warrendale, PA 1983.

Betts PR, Astley R, Raine DN. Lead intoxication in children in
  Birmingham. Br Med J 1973; 1(5850): 402-6.

Bhalla AK, Amento EP, Clemens T, Holick MF, Krane SM. Specific
  high-affinity receptors for 1,25-dihydroxyvitamin 03 in human
  peripheral blood mononuclear cells:  presence in monocytes and
  induction in T lymphocytes following activation.  J Clin
  Endocrinol Metab 1983; 57: 1308-10.

Bigley HA Jr, Keller BD, Kloppe MG. Octane requirement increase in
  1971 model cars with leaded and unleaded gasolines. SAE Paper
  #710675, Society of Automotive Engineers, Warrendale, PA 1971.

Bigley HA Jr, Benson JD. Octane requirement increases in 1971 model
  cars - with and without lead. SAE Paper #730013, Society of
  Automotive Engineers, Warrendale, PA 1973.

Billick IH, Curran AS, Shier DR. Analysis of pediatric blood
  lead levels in New York City for 1970-1976. Environmental
  Health Perspectives 1979; 31:183-90.

Billick IH, et al. Predictions of pediatric blood lead levels
  from gasoline consumption. U.S. Department of Housing and
  Urban Development 1982.

Blank FM, et al. Valuation of aesthetic preferences:  A case
  study of the economic value of visibility.  Draft report
  to the Electric Power Research Institute, Palo Alto, Ca 1977.

Blomquist G. Valuation of Life:  Implications of Automobile
  Seat Belt Use.  Ph.D. Dissertation.  University of Chicago, 1977.

Brennan MJW, Cantrill RC. y-Aminolevulinic Acid is a potent
  agonist for GABA auto receptors.  Nature 1979; 280: 514-5.

Brookshire DS, et al. The valuation of aesthetic preferences.
  J Environ Econ Manag 1976; 3: 325-46.

Brown C. Equalizing differences in the labor market.
  Quarterly Journal of Economics 1980; 94.

-------
                                -4-


Bryant EC.  Statistical Analyses (2nd Edition)  New York:
  McGraw Hill 1966.

Bryce-Smith D, Deshpunde RR, Hughes J, Waldron  HA. Lead and
  cadmium levels in still births. Lancet 1977;  1: 1159.

Bull RJ, Stanaszek PM, O'Neill JJ,  Lutkenhoff SD. Specificity of
  the effects of lead on brain energy metabolism for substrates
  donating a cytoplasmic reducing equivalent. Environ Health
  Perspect 1975; 12:89-95.

Bull RJ. Effects of trace metals and their derivatives on the
  control of brain energy metabolism. Lee SD ed. Biochemical
  Effects of Environmental Pollutants.  Ann Arbor, Mi: Ann Arbor
  Science 1977: 425-40.

Bull RJ, Lutkenhoff SD, McCarty GE, Miller RG.  Delays in the
  postnatal increase of cerebral cytochrome concentrations in
  lead-exposed rats. Neuropharmacology 1979; 18: 83-92.

Bull RJ. Lead and energy metabolism. Singhal PL, Thomas JA eds.
  Lead Toxicity. Baltimore:  Urban and Schwarzenberg, Inc 1980:
  119-68.

Bull RJ, McCauley DT, Tayler DH, Crofton KM. The effects of
  lead on the developing central nervous system of the rat.
  Neurotoxicology 1983; 4(1): 1-17.

Bureau of Food and Drug Administration Compliance Program Report
  of Findings, Total Diet Studies,  1973-1980 reports.

Byers RK. Lead poisoning, review of the literature and report
  on forty-five cases. Pediatr 1959; 23:585-603.

Caracciolo F, Sperot JA. The effects of phosphorus-containing
  engine oil additives on exhaust oxidation catalyst degradation.
  SAE Paper #760562, Society of Automotive Engineers, Warrendale,
  PA 1976.

Carey LR, Stover WH, Murray DW. Extended drain passenger car
  engine  oils. SAE Paper #780952, Society of Automotive Engineers,
  Warrendale, PA 1978.

Champion  Spark Plug Co. Champion Ingnition and Engine Performance
  Conferences.  Volumes 1971-1976.

Chance RL, Ceselli RG. Corrosiveness of exhaust gas condensates.
  SAE Paper #830585, Society of Automotive Engineers, Warrendale,
  PA 1983.

-------
                                -5-


Cherrie JM. Factors influencing valve temperature in passenger-car
  engines. SAE Paper #650484, Society of Automotive Engineers,
  Warrendale, PA 1965.

Chesney RW, Rosen JF, DeLuca HF. Disorders of calcium metabolism
  in children. Chiumello, Sperling M eds. Recent Progress in
  Pediatric Endocrinology. New York, NY: Raven Press 1983: 5-24.

Children living in the vicinity of a primary lead smelter.
  Environmental Research 1980; 22: 81-94.

Chisolm JJ Jr, Harrison HE.  The exposure of children to lead.
  Pediatrics 1956; 18: 943-58.

Chisolm JJ Jr. The use of chelating agent in the treatment of acute
  and chronic lead intoxication in children.  J Pediatr (St. Louis)
  1968; 73: 1-38.

Chisolm JJ Jr, Barltrop D. Recognition and management of children
  with increased lead absorption. Arch Dis Child 1979; 54: 249-62.

Chock D, et al. Effect of NOX emission rates on smog formation in
  the California south coast air basin.   Environmental Science
  and Technology 1981 August; 15(8).

Cohen C, et al. Respiratory symptoms, spirometry, and oxidant air
  pollution in nonsmoking adults.  Amer Rev Resp Disease 1972;
  105: 251-61.

Cohen GJ, Ahrens WE. Chronic lead poisoning: a review of seven
  years' experience at the Children's Hospital, District of Colum-
  bia.  J Pediatr (St Louis) 1959; 54: 271-84.

Colyer C, Harding K. The 1980s — a decade of challenge for engine
  oils.  SAE Paper # 811410, Society of  Automotive Engineers,
  Warrendale, PA 1981.

Cooper WC, Gaffey WR. Mortality of lead  workers. Cole JF ed.
  Proceedings of the 1974 Conference on  Standards of Occupational
  Lead Exposure; February 1974; Washington, DC. J Occup Med 1975;
  17:  100-7.

Cooper WC. Mortality in employees of lead production facilities
  and lead battery plants, 1971-1975.  Lynam Dr, Piantanida LG,
  Cole JF Eds. Environmental Lead;  Proceedings of the second
  international symposium on environmental lead research;
  Cincinnati, OH, 1978 December.  New York:  Academic Press 1981.

Cordera FJ, et al. TEL scavengers in fuel affect engine performance
  and durability. SAE Paper #877A, Society of Automotive Engineers,
  Warrendale, PA 1964.

-------
                                -6-

Council on Environmental Quality. Environmental Quality - 1980.
  Government Printing Office, 1980 December.

Graver RJ, Podiak RS, Miller RD. Spark-plug design factors and
  their effect on engine performance.  SAE Paper #700081, Society
  of Automotive Engineers, Warrendale, PA 1970.

Crocker T, Vaux H. Some economic consequences of ambient oxidant
  impacts on a natural forest.  Prepared for U.S. EPA, Office of
  Policy Analysis, 1983 August.

Crofton KM, Taylor DH, Bull RJ,  Snulka DJ,  Lutkenhoff SD. Develop-
  mental delays in exploration and locomotor activity in male rats
  exposed to low level lead. Life Science 1980; 26:  823-31.

Grouse WW, Johnson RH, Reiland WH. Effect of unleaded fuel on
  lubricant performance. SAE Paper #710584, Society  of Automotive
  Engineers, Warrendale, PA 1971.

Danis LJ. Engine valve cooling.  SAE Paper #730055, Society of
  Automotive Engineers, Warrendale, PA 1973.

De la Burde B, Choate MS Jr. Does asymptomatic lead  exposure in
  children have latent sequelae. Journal of Pediatrics 1972;
  81: 1088-91.

De la Burde B, Choate MS Jr. Early asymptomatic lead exposure
  and development at school age. Journal of Pediatrics 1975;
  87: 638-42.

Delves HT. A micro-sampling method for the  rapid determination of
  lead in blood by atomic absorbtion spectrophotometry. Analyst
  1970: 431-3.

DePalma TV, Bailey CH. Application of  catalytic converters to
  heavy-duty gasoline powered trucks.   SAE  Paper #750902, Society
  of Automotive Engineers, Warrendale, PA 1975.

De Para M, Belvoir Research and Development Center,  U.S. Army.
  personal communication.

Detels R, et al. The UCLA population studies of chronic obstructive
  respiratory disease.  American Journal of Epidemiology 1979; 109-

Dingwall-Fordyce I, Lane RE. A follow-up study of lead workers.
  Br J Ind Med 1963; 20: 313-5.

Doelling RP. Engine's definition of unleaded gasoline. SAE Paper
  8401, Society of Automotive Engineers, Warrendale, PA 1971.

Dresner DL, Ibrahim NG, Mascarenhas BR, Levere RD. Modulation of
  bone marrow heme and protein synthesis by trace elements.  Environ
  Res 1982; 28: 55-66.

-------
                                -7-


DRI. The Data Resources Inc review of the U.S. economy.   Various
  issues, 1983 and 1984.  DRI,  Lexington, MA.

Drinkwater B, et al. Air pollution, exercise, and heat stress.
  Archives of Environmental Health 1974; 28: 177-81.

Durand D. Stable Choas. Morristown, NJ:  General Learning
  Corporation 1971.

Eccleston BH, Hurn RW. Comparative emissions from some leaded and
  prototype lead-free automobile fuels.   Report of Investigation
  #7390, U.S. Department of the Interior, Bureau of Mines
  Petroleum Research Center, Bartlesville, OK 1970.

Ellis JC. Future automotive fuels. SAE Paper #720617,  Society of
  Automotive Engineers, Warrendale, PA 1972.

Ellis JC. Gasoline for low-emission vehicles. SAE Paper  #730616,
  Society of Automotive Engineers, Warrendale, PA 1973.

Energy and Environmental Analysis, Inc.  Historical and projected
  emissions conversion factor and fuel economy for heavey-duty
  trucks 1962-2002. Motor Vehicle Manufacturers Association,
  Detroit, Michigan 1983.

Ennis JM, Harrison HE. Treatment of lead encephalopathy  with  BAL
  (2,3-dimercaptopropanol).  Pediatrics  1950; 5: 853-68.

Erickson MM, Pokus A, Canter CE, Dickson AW, Hillman  LS.  Tissue
  mineral levels in victims of sudden infant death syndrome.  I.
  toxic metals - lead and cadmium. Pediatric Research 1983;
  17: 784-99.

ETA Engineering, Inc. Assessment of benefits from new source
  performance standards for volatile organic compounds.   Prepared
  for U.S. EPA, Office of Policy Analysis, 1983 September.

Ethyl Corporation. Monthly Retail Gasoline Sales Bulletins,
  1976-1982.

Evans, et al. Ozone measurement from a network of remote sites.
  Journal of the Air Pollution Control Association 1983  April;
  32(4).

Expert Committee on Trace Metal Essentiality. Independent peer
  review of selected studies by Drs. Irchgessner and  Reichlmayr-
  Lias concerning the possible nutritional essentiality  of lead:
  official report of findings and recommendations of  an  interdis-
  ciplinary expert review committee, 1983. Available  for inspection
  at: U.S.  Environmental Protection Agency, Environmental Criteria
  and Assessment Office, Research Triangle Park, NC.

-------
                                -8-


Exxon Memo, Re:  Gulf/East Coast Gasoline.   1978 January.

Fachetti S, Geiss F. Isotopic lead experiment status report.
  Luxembourgh: Commission of the European Communities 1982;
  Publication No. EUR8352EN.

Faggan JE, et al. An evaluation of manganese as an antiknock in
  unleaded gasoline. SAE Paper 1750925, Society of Automotive
  Engineers, Warrendale, PA 1975.

Fahim MS, Fahim Z, Hall DG. Effects of subtoxic lead levels  on
  pregnant women in the state of Missouri.  Res Commun Chemical
  Pathol Pharmacol 1976; 13(2): 309.

Faucett Associates, Draft Report:  Review and Critique of Previous
  OMSAPC Cost-Effectiveness Analysis,  1983  March.

Felt AE, Kerley RV. Engines and effects of  lead-free gasoline.
  SAE Paper #710367, Society of Automotive  Engineers, Warrendale,
  PA 1971.

Ferris B Jr. Health effects of exposure to  low levels of regulated
  air pollutants. Journal of the Air Pollution Control Association
  1978 May; 28(5) .

Fisher RA. Statistical Methods for Research Workers.  New York:
  Hafner Press 1970.

Florey C du V, et al. The relation between  respiratory illness
  in primary schoolchildren and the use of  gas for cooking.  III.
  nitrogen dioxide, respiratory illness and lung infection.
  Int J Epidemiol 1979; 8: 347-53.

Folinsbee LJ, et al. Effect of 0.62 ppm N02 on cardiopulmonary
  function in young male non-smokers.   Environmental Research
  1978; 15: 199-205.

Folinsbee LJ, Bedi JF, Horvath SM. Pulmonary function changes
  in trained athletes following 1-hour continuous heavy exercise
  while breathing 0.21 ppm ozone.  J Appl Physiol Respir Environ
  Exercise Physiol  (in press) 1984.

Ford Automotive Emissions Office, personal  communication.

Forthofer RN. Investigation of non-response bias in NHANES
  II. American Journal Of Epidemiology (Bult.) 1983

Fouts PJ, Page IH. The effect of chronic lead poisoning on arterial
  blood pressure in dogs. Am Heart J 1942;  24: 329-31.

Fouts PJ, Page IH. The effect of chronic lead poisoning on
  arterial blood pressure in dogs.  American Heart Journal 1942;
  24: 329-31.

-------
                                -9-


Fowler BA. General subcellular effects of lead,  mercury,  cadmium,
  and arsenic. Environ Health Perspect 1978;  22: 37-41.

Fowler BAf Kimmel CA, Woods JS, McConnell EE, Grant LD.   Chronic
  low-level lead toxicity in the rat: III. an integrated  assessment
  of long-term toxicity with special reference to the kidney.
  Toxicol Appl Pharmacol 1980; 56: 59-77.

Freas W. A digital solution to city-specific  ekma isopleth diagrams
  in ozone regulatory analysis.  U.S. EPA, Research Triangle Park
  1983 July.

Freeman AM III.  Air and Water Pollution Control:  A Benefit-
  Cost Assessment.  New York: John Wiley and  Sons 1982.

Freeman R. Reversible myocarditis due to chronic lead poisoning  in
  childhood.  Arch Dis Child 1965; 40: 389-93.

Friedlander SK. Chairman, Clean Air Scientific Advisory Committee
  (CASAC).  Review and Closure of the OAQPS Staff Paper for Nitro-
  gen Oxides, 1982 July 6.

Fuchs EJ. Unleaded versus leaded fuel results in laboratory engine
  tests. SAE Paper #710676, Society of Automotive Engineers,
  Warrendale, PA 1971.

Gagliardi JC. Effect of fuel antiknock compounds and deposits  on
  exhaust emissions. SAE Paper #670128, Society  of Automotive
  Engineers, Warrendale, PA 1967.

Gagliardi JC, Ghannam F. Effects of tetraethyl lead concentration
  on exhaust emissions in customer-type vehicle  operation. SAE
  Paper #690015, Society of Automotive Engineers, Warrendale,  PA
  1969.

Gallopoulos NE. Projected lubricant requirement  of engines oper-
  ating with lead-free gasoline. SAE Paper #710585, Society of
  Automotive Engineers, Warrendale, PA 1971.

General Motors Corporation. General Motors Corporation comments
  on EPA report 'Costs and Benefits of Reducing  Lead in Gasoline1
  letter from T.M. Fisher (GM) to R.D. Morgenstern (EPA/OPA),  1984
  June 18.

Gergel WC, Sheahan TJ. Maximizing petroleum utilization  through
  extension of passenger car oil drain periods - what's  required?
  SAE Paper #760560, Society of Automotive Engineers, Warrendale,
  PA 1976.

Gerking S, Stanley L, Weirick W. An economic  analysis of  air
  pollution and health:  the case of St. Louis.  Prepared  for
  U.S. EPA, Office of Policy Analysis, 1983 July.

-------
                                -10-


Giles WS. Valve problems with lead free gasoline.  SAE Paper
  #710368, Society of Automotive Engineers,  Warrendale, PA 1971.

Giles WS, Updike SH. Influence of low lead fuels on exhaust
  valve performance. SAE Paper #710674, Society of Automotive
  Engineers, Warrendale, PA 1971.

Glasson WA, Tuesday CS.  Environ Sci Technol  1970; 4: 37.

Glasson W. Effect of Hydrocarbons and NOX on photochemical smog
  formation under simulated transport conditions. Journal of the
  Air Pollution Control  Association 1981 November 11;  31(11).

Godfrey D, Courtney RL.  Investigation of the mechanism of exhaust'
  valve seat wear in engines run on unleaded gasoline.  SAE Paper
  #710356, Society of Automotive Engineers,  Warrendale, PA 1971.

Goldberg AM, Meredith DA, Miller S, Moore MR, Thompson CG. Hepatic
  drug metabolism and heme biosynthesis in lead-poisoned rats.
  British Journal of Pharmacology 1978; 62:  529-36.

Goldstein BD. The relation between respiratory illness in school-
  children and the use of gas for cooking. II. Factors affecting
  nitrogen dioxide levels in the home.  Int J Epidemiol 1979;
  8: 339-45.

Goldstein E, et al. Fate and distribution of inhaled nitrogen
  dioxide in rhesus monkey.  Am Rev Respir Dis 1977; 115: 403-12.

Goldstein E, Moderator.  Photo Chemical Air Pollution.  Inter-
  departmental Conference sponsored by the Department of Medicine,
  University of California, School of Medicine; Davis, California
  1982.

Goth G. Recent developments in hardfacing alloys for internal
  combustion engines. SAE Paper #831287 in State-of-the-Art on
  Design and Performance of Diesel Power Cylinder Components,
  SP-552, Society of Automotive Engineers, Warrendale, PA 1983.

Goyer RA, Rhyne BC. Pathological effects of  lead. Internal Rev
  Exp Pathol 1973; 12: 1-77.

Gray DS, Azhari AG. Saving maintenance dollars with lead-free
  gasoline. SAE Paper #720084, Society of Automotive Engineers,
  Warrendale, PA 1972.

Green AES, et al. An interdisciplinary study of the health, social,
  and environmental economics of the sulfur oxide pollution in
  Florida.  Interdisciplinary Center for Aeronomy and (other)
  Atmospheric Sciences,  Univ of Florida; Gainesville, FL 1978.

-------
                                -11-


Gross-Selbeck E, Gross-Selbeck M. Changes in operant behavior of
  rats exposed to lead at the accepted no-effect level. Clinical
  Toxicology 1981; 18: 1247-56.

Gunter EW, et al. Laboratory Procedures Used by the Clinical
  Chemistry Division, Centers for Disease Control for the Second
  National Health and Nutrition Examination Survey.  Atlanta
  Centers for Disease Control 1981.

Hackney JDr et al. Experimental studies on human health effects
  on air pollutants. II.  Four hour exposure to ozone alone and
  in combination with other pollutants.  Arch Environ Health
  1975a; 33: 379-84.

Hackney JD, et al. Experimental studies on human health effects
  on air pollutants. III.  Two hour exposure to ozone alone and
  in combination with other gases.  Arch Environ Health 1975b;
  30: 176-81.

Hackney JD. Effects of atmospheric pollutants on human physiologic
  function.  Final report, U.S. EPA 1976.

Hackney JD, et al. Experimental studies on human health effects
  of air pollutants. IV.  Short-term physiological  and clinical
  effects of nitrogen dioxide exposure.  Arch Environ Health 1978;
  33: 176-81.

Hall CA, Felt AE, Brown WJ. Evaluating effects of fuel factors  on
  stabilized exhaust emissions levels. SAE Paper #  690014,  Society
  of Automotive Engineers, Warrendale, PA 1969.

Hammer D, et al. Los Angeles student nurse study. Daily symptom
  reporting and photochemical oxidants. Archives of Environmental
  Health 1974; 28.

Hare CT, Springer KJ, Huls TA. Exhaust emissions from farm
  construction, and industrial engines and their impact. SAE Paper
  #750788, Society of Automotive Engineers,  Warrendale, PA  1975.

Harlan WR, Landis JR, Schmouder RL, et al.  Relationship of
  blood lead and blood pressure in the adolescent and adult
  U.S. population.  JAMA 1985 January 25.

Harrison.  Principles of Internal Medicine.   Ninth  Edition.

Hartunian NS, Smart CN, Thompson MS.  The Incidence and Economic
  Costs of Major Health Impariments.  Lexington, MA:   Lexington
  Books 1981.

-------
                                -12-


Harvey P, Hamlin M, Kumar R. The Birmingham blood lead study.
  Presented at: annual conference of the British Psychological
  Society, Symposium on lead and health, 1983.  Available for
  inspection at: U.S. Environmental Protection Agency, Environ-
  mental Criteria and Assessment Office, Research Triangle Park,
  N.C.

Hasselblad V. Modeling dose response relations for health effects
  data.  Environmetrics 81: Selected Papers, SIAM, Philadelphia,
  1981.

Hasselblad V, Svendsgaard D. Reanalysis of the Los Angeles student
  nurse study. Unpublished paper, U.S.  EPA, Health Effects Research
  Lab, Research Triangle Park, NC 1975  August.

Heck W, et al. Nitrogen dioxide: time-concentration model to
  predict acute foliar injury.  U.S. EPA,  Corvallis, Oregon.
  Pub. No. 600/3-79-057, 1979-

Heck W, et al. A reassessment of crop loss from ozone.
  Environmental Science and Technology  1983; 17(12).

Heintz HT, Hershaft A, Horak GC. National  Damages of Air and
  Water Pollution.  Report submitted to U.S. EPA 1976.

Hickling Partners, Inc.  Final Report on the Assessment of the
  Economic Impact on the Automotive Parts/Service Industry of
  Proposed Gasoline Lead Content Regulations, submitted to Policy
  Planning and Assessment Directorate,  Environment/Canada 1981
  March.

Hernberg S, Nikkanen J. Enzyme inhibition  by lead under normal
  urban conditions. Lancet 1970; 1: 63-4.

Holtzman D, Shen Hsu J. Early effects of inorganic lead on immature
  rat brain mitochondrial respiration.   Pediatric Res 1976;  10:
  70-5.

Holtzman D, Shen Hsu J, Desautel M. Absence of effects of lead
  feedings and growth-retardation on mitochondriae and microsomal
  cytochromes in the developing brain.  Toxicol Appl Pharmacol
  1981; 58: 48-56.

Hoppenbrouwers T, et al. Seasonal relationships of sudden infant
  death syndrome and environmental pollutants.  American Journal
  of Epidemiology 1981; 113(6).

Hornbeck DD, et al. Advantages of lead  in  gasoline for european
  cars - the lead road bonus.  SAE Paper #750936, Society of
  Automotive Engineers, Warrendale, PA  1975.

-------
                                 -13-


 Horvath  S,  et  al.  Maximal  aerobic  capacity  at different levels
   of  carboxyhemoglobin.  Journal of Applied Physiology 1975;
   38:  300-3.

 Horvath  SM, Folinsbee  LJ.  Effects  of pollutants on cardiopulmonary
   function.  Report  to U.S.  EPA  1979.

 Hospadaruk V.  Fundamental  machanisms of  automobile corrosion. SAE
   Paper  #780909, Society of  Automotive Engineers, Warrendale, PA
   1978.

 Hudnall  JR, et al. New gasoline  formulations provide protection
   against corrosive  engine wear. SAE Paper  #690514, Society of
   Automotive Engineers, Warrendale, PA 1969.

 Hudson PJ.  Fitting segmented curves whose join points have to be
   estimated.   Journal  of the Amercian Statistical Association
   1966;  61: 873-80.

 Hypertension Detection and Follow-up Program Cooperative Group.
   The Effect of Treatment  on Mortality in Mild Hypertension.
   N Eng  J Med  1983;  307: 976-80.

 lannacone A, Carmigmani M, Boscolo P.  Cardiovascular reactivity
   in  the rat following chronic exposure  to  cadmium and lead.  Ann
   1st Supr  Sanrta  1981; 17:   655-60.

 Innes WB.   Environmental Science and Technology 1981; 15: 933.

 Isringhaus  EA  Jr,  Henderson  BM.  Engine rusting in short-run
   passenger-car service. SAE Paper #650868, Society of Automotive
   Engineers, Warrendale, PA  1965.

 Jacobs KM.  Energy  usage and  other  comparisons between gasoline
   and diesel medium  duty trucks.  SAE Paper #770757, Society of
   Automotive Engineers, Warrendale, PA 1977.

/Janney A.   The relationship  between gasoline lead emissions and
   blood  poisoning  in Americans.  Prepared for U.S. EPA, Office of
   Policy Analysis, 1982.

 Joumard  R, et  al.  Mathematical models of the uptake of carbon
   monoxide  on  hemoglobin at  low  carbon monoxide levels.
   Environmental Health Perspectives 1981; 41.

 Journal  of  Pediatrics  (Editorial).  New  approaches to screening
   for iron deficiency.  1977; 90:  678.

 Kagawa J, et al. Photochemical air pollution:  its effects on
   respiratory  function of  elementary school children.  Arch
   Environ Health 1975; 30: 117-22.

-------
                                -14-


Kakalik JS, et al. The Cost of Special Education.   Rand Corporation
  Report N-1791-ED, 1981.

Kammholz LP, Thatcher LG, Blodgett FM, Good TA.  Rapid protoporphy-
  rin quantitation for detection of lead poisoning.  Petiatrics
  1972; 50: 625-31.

Keller BD, et al. ORI of today's vehicles.  SAE Paper #760195,
  Society of Automotive Engineers, Warrendale, PA  1976.

Keller MD, et al. Respiratory illness in household using gas and
  electricity for cooking. II. Symptoms and objective findings.
  Environ Res 1979; 19: 504-15.

Kent WL, Finnigan FT. Effect of some fuel and operating parameters
  on exhaust valve seat wear. SAE Paper #710673, Society of Auto-
  motive Engineers, Warrendale, PA 1971.

Kerr HD, et al.  Effects of nitrogen dioxide on  pulmonary function
  in human subjects.  Environmental Chamber Study.  U.S. EPA
  Research Triangle Park, NC 1978.

Kerr JCS, et al. Effects of nitrogen dioxide on  pulmonary function
  in human subjects.  Environmental Chambers Study.  Environ Res
  1979; 19: 392-404.

Killus JP, et al. Application of a regional oxidant model to the
  northeast United States.  Presented at the International
  Conference on Long Range Transport Models for  Photochemical
  Oxidants and Their Precursor.  U.S. EPA;  Research Triangle Park,
  NC, 1983 April 12-14.

Kim Y, Nickola R. A heat-resistant aluminized steel for high-
  temperature applications. SAE Paper #800316, Society of Auto-
  motive Engineers, Warrendale, PA 1980.

Kinosian J. Ozone precursor relationships from EKMA diagrams.
  Environmental Science and Technology 1982; 16  (12).

Kipp KL, et al. Ability of gasoline additives to clean engines and
  reduce exhaust emissions. SAE Paper #700456, Society of Automo-
  tive Engineers, Warrendale, PA 1970.

Klein JP, et al. Hemoglobin affinity for oxygen  during short-term
  exhaustive exercise.  Journal of Applied Physiology 1980; 48.

Kleinbaum DG, Morgenstern H, Keyser LL. Epidemiologic Research.
  1982.                                                °=

Kopp R, Vaughan W. Agricultural benefit analysis:  alternative
  ozone and photochemical oxidant standards.  Resources for the
  Future 1983 June 30; and discussion with authors, 1984.

-------
                                -15-


Kopp SJ, Glonek T, Erlanger M, Perry EF, Perry HM Jr, Barany M.
  Cadmium and lead effects on myocardial function and metabolism.
  J Environ Pathol Toxicol 1980; 4: 205-27.

Kromhout D, Coulande CL.  Trace metals and CHD risk indicators
  in 152 elderly men (the Zutphen Study).  Eur Heart J 1984;
  5 (Abstr suppl 1) :  101

Kuroishi N. Recent technology of sintered metals for automotive
  use. SAE Paper #811416, Society of Automotive Engineers,
  Warrendale, PA 1981.

Lamola A-A, Joselow M, Yamane T. Zinc protoporphyrin (ZPP):  a
  simple, sensitive, fluorometric screening test for lead poison-
  ing.  Clin Chem (Winston Salem, NC) 1975a;  21: 93-7.

Laraola A-A, Piomelli S. Poh-Fitzpartick MB, Yamane T, Barber LC.
  Erythropoietic protoporphyria and lead intoxication: the
  molecular basis for difference in cutaneous photosensitivity:
  II. different binding of erythrocyte proteporphyrin to hemo-
  globin.  J Clin Invest 1975b; 56: 1528-5.

Landrigan PJ, Baker EL, Feldman RG, Cox DH, Eden KV, Orenstein WA,
  Mather JA, Yankel AJ, von Lindern IH.  Increased lead absorption
  with anemia and slower nerve conduction in  children near a lead
  smelter. Journal of Pediatrics 1976; 89: 904-10.

Larson J, of Weyerhauser.  personal communication, 1984.

Lauwerys R, Buchet JP, Roels H, Hubermont G.  Placenta I transfer
  of lead, mercury, cadmium, and carbon monoxide in women.
  I.  Comparison of the frequency distributions of the biological
  indices in maternal and umbilical cord blood. Environmental
  Research 1978; 15: 278-89.

Lave L, Seskin E. Air Pollution and Human Health. Baltimore:
  Johns Hopkins University Press 1977.

Lax D. Analysis of fuel economy and vehicle use data for the 1977
  truck inventory and use survey. SAE Paper #810388, Society of
  Automotive Engineers, Warrendale, PA 1981.

Layland D, Cole H. A review of recent applications of the SAI
  urban airshed model.  U.S. EPA; Research Triangle Park,
  NC 1983 December.

Lebowitz MD, et al. The effect of air pollution and weather  on
  lung function in exercising children and adolescents.  American
  Review of Respiratory Diseases 1974; 109.

-------
                                -16-


Lebowitz MD, O'Rourke MK, Dodge R. Holberg CJ,  Gorman G, Hoshaw
  RW, Pinnas JL, Barbe RA, Sweller MR.   The adverse health
  effects of biological aerosols,  other aerosols,  and indoor
  microclimate on asthmatics and nonasthmatics.  Environ Int
  1982; 8: 375-80.

Lebowitz MD, Holberg CJ,  Dodge RR. Repiratory effects on population
  from low level exposures to ozone,  presented at:  34th annual
  meeting of the Air Pollution Control  Association, 1983 June,
  Atlanta, GA. Pittsburgh, PA:  Air Pollution Control Association,
  paper no 83 - 12.5.

Lebowitz, MD.  The effects of environmental tobacco smoke exposure
  and gas stoves on daily peak flow rates in asthmatic families.
  Eur T Respir Dis 1984;  64 (133): 90-7.

Leighton J, Shehadi A, Wolcott R.  Aggregate benefits of air pollu-
  tion control, prepared  for U.S.  EPA,  Office of Policy Analysis,
  by Public Interest Economics Foundation; Washington, DC 1983
  June.

Leikkanen HE, Beckman EW.  Effect  of leaded and unleaded gaso-
  lines on exhaust emissions as influenced by combustion chamber
  deposits.  SAE Paper #710843, Society of Automotive Engineers,
  Warrendale, PA 1971.

Lenane DL. MMT - a further evaluation.  SAE Paper #770656, Society
  of Automotive Engineers, Warrendale,  PA 1977.

Levy RI. The decline in coronary heart  disease  mortality: Status
  and perspective on the  role of cholesterol. American Journal
  of Cardiology 1984 August 27; 54 (4).

Lewis BW, Collin RJ, Wilson HS. Seasonal incidence of lead poisoning
 in children in St. Louis. Southern Medical Journal 1955; 48: 298-301

Lewis RA, Newhall HK, Peyla RJ, Voss DA, Welstand JS. A New concept
  in engine deposit control additives for unleaded gasoline.  SAE
  Paper #830938, Society  of Automotive  Engineers,  Warrendale, PA
  1983.

Lilienfeld AM, Lilienfeld DE.  Foundations of Epidemiology.
  New York, 1980.

Lincoln RH. written statement before the Senate Committee on
  Environmental and Public Works in Opposition to Senate Bill 2609,
  1984.

Lin-Fu JS. Vulnerability  of children to lead exposure and toxicity:
  parts one and two. New England Journal of Medicine 1973; 289:
  1229-33, 1289-93.

-------
                                   -17-


   Linn WS, et al. Respiratory function and symptoms in urban office
     workers in relation to oxidant air pollution exposure.  Amer
     Rev Resp Disease 1976; 114: 477-83.

   Linn W, et al. Human respiratory effects of heavy exercise in
     oxidant-polluted ambient air. American Review of Respiratory
     Disease 1981; 123 (4) .

   Lippmann M, et al. Effects of ozone on the pulmonary function of
     children. Lee S, et al. eds, The Biomedical Effects of Ozone and
     Related Photochemical Oxidants. Princeton Scientific Publishers,
     Inc; Princeton, NJ:  Advances in Modern Environmental Toxicology
     1983; V: 423-46.

   Litman DA, Corriea MA. L-tryptophan:  a common denominator of
     biochemial and neurological events of acute hepatic porphyrias?
     Science 1983; 222: 1031-3.

   Loehman E, et al. Distributional Analysis of regional benefits and
     cost of air quality control.  Journal of Environmental Economics
     and Management 1979 Steptember; 6(3).

   Love GJ, et al. Acute respiratory illness in families exposed
     to nitrogen dioxide ambient air pollution in Chattanooga, TN.
     Arch Environ Health 37: 75-80.

   MacLean DC.  Stickstoffoxide ALS phytotoxis CHE Luftuerurein 16
     UNGEN.  Staub-Reinhalt Luft 1975; 35(5).

   Mage D, et al.  A sensitivity analysis of the effects of CO time
     patterns on computer  blood carboyhemogologin levels,  presented
     at: Air Pollution Control Association Meetings 1983, Atlanta, Ga.

   Mahaffey KR, Annest JL, Roberts J, Murphy MS. National estimation
     of blood lead levels: United States (1976-1980).  New England
     Journal of Medicine;  307: 573-9.

   Mahaffey KR, Rosen JF,  Chesney RW, Peeler JT, Smith CM, Deluca HF.
     Association between age, blood lead concentrations and serum
     1,25-dihydroxycholecalciferol levels in children. Amercian Jour-
     nal of Clinical Nutrition 1982; 35: 1327-31.

   Makino K, Mizoguchi I.  Symptoms caused by photochemical smog.
     Japan Journal of Public Health 1975; 22(8).

/  /Manton WI. Sources of lead in blood:  identification by stable
'   '  isotopes. Archives of Environmental Health 1977; 32:149-59.

   McBride WG, Black BP, English BJ. Blood lead levels and behavior  of
     400 preschool children. Med Journal Aust  1982; 2: 26-9.

-------
                                -18-


McCauley PT, Bull RJ.  Lead-induced delays in synaptogenesis in the
  rat cerebral cortex. Fed Proc Fed Am Soc Exp Bio 1978; 37: 740.

McCormick H. TRW Piston Ring Division, personal communication.

McDonnel TF. The effects of engine oil additives on vehicle fuel
  economy, emissions,  emission control components and engine wear.
  SAE Paper #780962, Society of Automotive Engineers, Warrendale,
  PA 1978.

McDonnell WF, Hortsmann DH, Hazucha DH, Hazucha MJ, Seal E Jr, Haak
  ED, Salaam S, House  DE.   Pulmonary effects on ozone exposure
  during exercise:  dose-response characteristics.  J Appl Physiol
  Respir Environ Exercise Physiol 1983; 54: 1345-52.

McGartland A, Ostro B. Benefits Analysis of New Source Performance
  Standards:  The control of HC and NOX. U.S. EPA, Office of Policy
  Analysis, Draft reported 1984 October.

McGartland A, Ostro B.  Benefit Analysis of New Source Performance
  Standards:  The Control of VOC and NOX.  U.S. EPA, Office of
  Policy analysis,  1985.

McGee D, Gordon T.   The results of the Framingham Study applied to
  four other U.S. - based epidemeologic studies of coronary heart
  disease.  The Farmingham Study.  Section 31.  DHEW Pub No. (NIH)
  76-1083.  National Institutes of Health, Washington, D.C. U.S.
  Government Printing  Office 1976.

McLaughlin S, et al. Measuring effects of air pollution stress on
  forest productivity.  Tapp Journal 1984; 67: 74.

Melia RJW, et al. Association between gas cooking and respiratory
  disease in children.  Br Med J 1977; 2: 149-52.

Melia RJW, et al. Differences in N02 levels in kitchens with gas or
  electric cookers.  Atm Environ 1978; 12: 149-52.

Melia RJW, et al. The  relation between respiratory illness in
  primary schoolchildren and the use of gas for cooking.  I -
  Results from a national survey.  Int J Epid 1979; 8: 333.

Mellins RB, Jenkins CD. Epidemiological and psychological study of
  lead poisoning in children. Journal American Medical Association
  1955; 158: 15-20.

Mentzer WC. Differentration of iron deficiency from thalessemia
  trait. Lancet 1973;  1: 882.

Menzel DB.  The role of free radicals in the toxicity of air pollu-
  tants (nitrogen oxides and ozone).  Bryor WA ed.  Free Radicals
  in Biology, vol III. New York: Academic 1976: 181-202.

-------
                                -19-


Meredith PA, Moore MR, Campbell BC, Thompson GG, Goldberg A. Delta-
  animolaevulinic acid metabolism in normal and lead-exposed humans,
  Toxicology 1978; 9: 1-9.

Milar CR, Schroeder SR, Mushak P, Dolcourt J, Grant LD. Contribu-
  tions of the caregiving environment to increased lead burden of
  children.  American Journal Mental Deficiency 1980; 84: 339-44.

Milar CR, Schroeder SR, Mushak P, Boone L. Failure to find hyper-
  activity in preschool children with moderately elevated lead bur-
  den.  Journal Pediatric Psychology 1981; 6: 85-95.

Moffitt JV. Compatability of military standard engines with leaded
  and low-leaded fuels. AFLRL-19, Southwest Research Institute,
  San Antonio, Texas (NTIS IAD-756 511)  1972.

Monthly Energy Review.  U.S. Department of Energy, various issues,
  1978 - 1982.

Moore MR, Meredith PA. The association of delta aminolevulinic acid
  with the neurological and behavioral effects of lead exposure.
  Hemphill DD ed. Trace Substances in Environmental Health - X.
  Columbia, MO: University of Missouri - Columbia 1976.

Moran JB. The environmental implications of manganese as an alter-
  native antiknock, SAE Paper #750926, Society of Automotive
  Engineers, Warrendale, PA 1975.

Morbidity and Mortality Weekly Reports,  various issues, 1976-1981,
  U.S. Public Health Service.

Morbidity and Mortality Weekly Reports:   Reports of the Lead Point
  Poisoning Prevention Screening Program 1973-1981.

Morgan JM. Hyperkalemia and acidosis in lead neuropathy.  South
   Med J 1976; 69(7):  881-3.

Moreau T, Orssaud G, Juguet B, Busquet G. Plombemie et press ion
  arterielle: premiers resultats d'une enquete transversable de 431
  sujets de sexe masculin.  [Blood lead levels and arterial pres-
  sure: initial results of a cross sectional study of 431 male
  subjects.]  [Letter]  Rev Epidemol Sante Fpulique 1982; 30: 395-7.

Morris WE, Rogers JD Jr, Poskitt RW. 1971 cars and the  'new1 gaso-
  lines.  SAE Paper #710624, Society of Automotive Engineers,
  Warrendale, PA 1971.

Motor Vehicle Manufacturing Association.  Motor Vehicle Facts and
  Figures "83".  Detroit 1983.

Motor Vehicle Manufacturing Association.  Incentives for Proper
  Usage of Unleaded Fuel.  Detroit, 1984.

-------
                                -20-


Multiple Risk Factor Intervention Trial:  Risk factor changes and
  mortality results.  Multiple Risk Factor Intervention Trial
  Research Group.  JAMA 1982; 248: 1465-77.

National Academy of Sciences. Lead: airborne lead in perspective.
  Washington, DC:  National Academy of Sciences 1972. (Biologic
  effects of atmospheric pollutants.)

National Academy of Sciences. Air quality and automobile emission
  control, Vol 4. prepared for the Committee on Public Works, U.S.
  Senate, U.S. Government Printing Office 1974.

National Academy of Sciences. Nitrogen oxides:  medical and bio-
  logic effects of environmental pollutants.  National Research
  Council, Washington, DC 1976.

National Academy of Sciences. Nitrogen oxides. National Academy of
  Sciences, Washington, DC 1977: 197-214.

National Center for Health Statistics.  National Hospital Discharge
  Survey, various issues.

National Center for Health Statistics.  Plan and Operation of the
  Second National Health and Nutrition Examination Survey 1976-
  1980.  National Center for Health Statistics 1981 (Vital and
  Health Statistics Series 1, No. 15)

National Petroleum Council. Refinery Flexibility, an interim report
  of the National Petroleum Council, 1979 December.

National Research Council.  Air quality and automobile emisison
  control, Vol 4: The costs and benefits of automobiles emission
  control.  A report by the coordinating Committee on Air Quality
  Studies of the National Academy of Sciences and the National
  Academy of Engineering.  Prepared for the Committee on Public
  works, United States Senate, 93rd Congress, 2nd Session.
  Washington, DC: U.S. Government Printing Office 1974.

National Research Council. Nitrogen oxides: medical and biologi-
  cal effects of environmental pollutants.  National Academy of
  Sciences, Washington, DC 1977: 1-333.

National Research Council. Nitrates, an environmental assessment.
  National Academy of Sciences, Washington, DC 1978: 723.

National Research Council.  Acid deposition:  atmospheric processes
  in eastern North America.  NRC, National Academy Press, Washing-
  ton DC 1983.

Needleman HL, Gunnoe C, Leviton A, Reed R, Peresie H, Maher C,
  Barret P. Deficits in psychological and classroom performance of
  children with elevated dentine lead levels.  New England Journal
  of Medicine 1979; 300: 689-95.

-------
                                -21-
Needleman HLf Rabinowitz M, Leviton A, Linn S, Schoenbaums S.  The
  relationship between prenatal exposure to lead and congenital
  anomalies. Journal of the American Medical Association 1984;  251:
  2956-9.

New approaches to screening for iron deficiency. Editorial.
  Journal of Pediatrics; 90: 678.

Nisbet ICT. Sulfates and acidity in precipitation:   their relation-
  ship to emissions and regional transport of sulfur oxides.  Air
  quality and stationary source emission control, a report by the
  Commission on Natural Resources of the National Research Council.
  prepared for the Committee on Public Works, U.S.  Senate, 94th
  Congress, 1st Session.  Washington, DC: U.S. Government Printing
  Office 1975.

Odenbro A, Greenberg N, Vroegh K, Bederka J, Kihlstrom J-E.   Func-
  tional disturbances in lead-exposed children. Ambio 1983;  12:
  40-4.

Orehek J, et al. Effect  of short-term, low-level nitrogen dioxide
  exposure on brochial sensitivity of asthmatic patients.  J  Clin
  Invest 1976; 57:37

Orrin DS, Miner WR, Kipp KL. Unleaded gasoline - lubricant require-
  ments and fuel additive performance. SAE Paper #720689, Society
  of Automotive Engineers, Warrendale, PA 1972.

Otto DA, Benignus VA, Muller KE, Barton CN. Effects of age and  body
  lead burden on CNS function in young children.  I: slow cortical
  potentials.  Electroencephalog Clin Neurophysiol 1981; 52:  229-39.

Otto, et al. Effects of low to moderate lead exposure on slow
  cortical potentials in young children:  two year follow up study.
  Neurobehav Toxicol Teratol 1982; 4:733-7.

Otto D, Rovinson G, Baumann S, Schroeder S, Kleinbaum D, Barton C,
  Mushak P, Boone L. Five-year follow-up study of children with
  low-to-moderate lead absorption:  electrophysiological evaluation.
  presented at second international conference on prospective lead
  studies 1984 April; Cincinnati, OH. available for inspection  at:
  U.S. Environmental Protection Agency, Environmental Criteria  and
  Assessment Office, Research Triangle Park, NC.

Page WP.  Estimation of Economic losses to the agricultural sector
  from airborne residuals in the Ohio River Basin.  Journal of  the
  Air Pollution Control Association, 1982; 32:  151-4.

Paglia DE, Valentine WN, Dahlgren JG.  Effects of low level lead
  exposure on pyrimidine 5' nucleotidase and other erythrocyte
  enzymes:  possible role of pyrimidine 5' nucleotidase  in the
  pathogenesis of lead induced anemia.  Journal of Clinical Inves-
  tigation 1975; 56:1164-9.

-------
                                -22-


Pahnke AJ,  Conte JF.  Effects of combustion-chamber deposits and
  driving conditions on vehicle exhaust emissions.  SAE Paper
  #690017,  Society of Automotive Engineers,  Warrendale PA 1969.

Pahnke AJ,  Bettoney WE. Role of lead antiknocks in  modern gasoline.
  SAE Paper #710842, Society of Automotive Engineers Warrendale, PA
  1971.

Patterson WR. Materials, design and corrosion effects on exhaust-
  system life. SAE Paper #780921, Society of Automotive Engineers,
  Warrendale, PA 1978.

Perlstein MA, Attala R. Neurologic sequelae  of plumbism in children.
  Clinic Pediatric (Philadelphia) 1966; 5: 292-8.

Perry HM Jr, Erlanger MW. Pressor effects of chronically feeding
  cadmium and lead together. Hemphill DD ed. Trace  Substances
  in Environmental Health-XII;   [proceedings of University of
  Missouri's 12th annual conference on trace substances in environ-
  mental health]; 1978 June, Columbia, MO. Columbia MO:  University
  of Missouri-Columbia:  268-75.

Pitts JN Jr, et al.  Comment on effect of nitrogen  oxide emissions
  on ozone levels in metropolitan regions effects of NOX emission
  rates on smog formation in the California  south coast air basin,
  and effects of hydrocarbon and NOX on photochemical smog forma-
  tion under simulated transport conditions.  Environ Sci Technol
  1983; 54-7.

Piomelli S, Davidow B, Guinee VF, Yound P, Gay G. The FEP (free
  erythrocyte porphyrins) test:  a screening micromethod for lead
  poisoning.  Pediatrics 1973;  51: 254-9.

Piomelli S, Seaman C, Zullow D, Curran A, Davidow B. Metabolic
  evidence of lead toxicity in  "normal" urban children.  Clinic Res
  1977; 25: 459A.

Piomelli S, Seaman C, Zullow D, Curran A, Davidow B.  Threshold for
  lead damage to heme synthesis in urban children.   Proceeding of
  the National Academy of Sciences in 1982 May; 79: 3335-9.

Piomelli S, Rosen JF, Chisolm JJ, Graef JW.   Management of child-
  hood lead poisoning.  Journal of Pediatrics 1984; 4:105.

Pirkle JL,  Annest JL.  Blood lead levels 1976-1980  - Reply (letter).
  N Eng J Med 1984; 310(17): 1125-6.

Pirkle JL,  Schwartz J, Landis JR, Harlan WR.  The  relationship
  between blood lead levels and blood pressure and  its cardiovas-
  cular risk implications.  American Journal of Epidemology, 1985-
  121(2):  246-58

-------
                                -23-


Platts Oilgram Price Report, Price Average Supplement.  December
  1983 Monthly Averages.  1984 January 26.

Pless LG, Bennett PA. Effects of some engine, operating, and oil
  variables on engine rusting in short-trip service. Lubrication
  Engineering 1969 October; 384.

Pless LG. Effects of some engine, fuel, and oil additive factors on
  engine rusting in short-trip service. SAE Paper #700457, Society
  of Automotive Engineers, Warrendale, PA 1970.

Pless LG. Interactions among oil additive and engine operating
  parameters affecting engine deposits and wear. SAE Paper #720686,
  Society of Automotive Engineers, Warrendale, PA 1972.

Pless LG. A study of lengthened engine oil-change intervals. SAE
  Paper #740139, Society of Automotive Engineers, Warrendale, PA
  1974.

Pless LG.  Deposits, wear, and catalyst performance with low ash
  and ashless engine oils. SAE Paper #750900, Society of Automotive
  Engineers, Warrendale, PA, 1975.

Pocock SJ, Ashby D.  Environmental lead and children's intelli-
  gence: a review of recent epidemiological studies.  Statician (in
  press).

The Pooling Project Research Group.  Relationship of blood pres-
  sure, serum cholesterol, smoking habit, relative weight and ECG
  abnormalities to incidence of major coronary events:  Final
  report of the pooling project.  J Chron Dis 1978; 31:  201-306.

Port CD. A comparative study of experimental and spontaneous emphy-
  sema.  J Toxicol Environ Health 1977; 2:589-604.

Portney P, Mullahy J. Ambient ozone and human health: an epidemio-
  logical analysis, prepared for U.S. EPA, Office of Air Quality
  Planning and Standards, 1983 September.

Portney P, et al. A multinomial model of the air pollution respira-
  tory health relationship.  Resources for the Furture 1984.

Portney P, Mullahy J. Urban air quality and acute respiratory
  illness. Journal of Urban Economics forthcoming 1985.

Posin C, et al. Nitrogen dioxide and human blood biochemistry.
  Arch Environ Health 1978 Nov/Dec; 318-24.

Provenzano G.  The social costs of excessive lead-exposure during
  childhood.  HL Needleman ED. Low Level Lead Exposure;  The
  Clinical Implications of Current Research.  New York: Raven
  Press 1980.

-------
                                -24-


Provvedini DM, Tsoukas CD, Deftos LJ,  Manolagas SC.  1,25-dihydroxy-
  vitamin 03 receptors in human lukocytes.   Science  (Washington,
  DC) 1983; 221: 1181-2.

Public Use Data Tape Documentation, Hematology and Biochemistry.
  Second National Health and Nutrition Examination Survey 1976-
  1980.  Catalog number 5411, U.S. Public Health Service, National
  Center for Health Statistics.

Rabinowitz WB, Needleman H. Petrol lead sales and umbilical cord
  blood lead levels in Boston, Massachusetts. Lancet 1983.

Rabinowitz WB, Wetherill GW, Kopple JD. Kinetic analysis of lead
  metabolism in human health. Journal  of Clinical Investigation
  1976; 58:260-70.

Ramirez-Cervantes B, Embree JW, Hine CH, Nelson KM,  Varner MO,
  Putnam RD. Health assessment of employees with different body
  burdens of lead. Journal Occupational Medicine 1978;  20: 610-7.

Randall A, et al. Bidding games for valuation of aesthetic
  environmental improvements.  J Environ Econ Manag  1974; 1:
  132-49.

Rasmussen H, Waisman DM. Modulation of cell function in the calcium
  messenger system.  Rev Physiol Biochemical Pharmacol  1983; 95:
  111-48.

Raven P, et al. Effect of carbon monoxide and peroxyacetylitrate on
  man's maximal aerobic capacity.  Journal of applied physiology
  1974; 36.

Raymond L. Today's fuels and lubricants and how they got that way.
  SAE Paper #801341, Society of Automotive Engineers, Warrendale,
  PA 1980.

Richet G, Albahary C, Morel-Maroqer L, Guillaume P,  Galle P. Les
  alterations renales dans 23 cas de saturnisme professionnel.
  [Renal changes  in 23 cases of occupational lead poisoning] Bull
  Memorial Society Medical Hop Paris 1966;  117: 441-66.

Roels H, Buchet JP, Lauwerys R, Hubermont G, Braux P, Claeys-
  Thoreau  F, LaFontaine A, Van Overschelde J. Impact of air pollu-
  tion by  lead on the heme biosynthetic pathway in schoolage
  children.  Arch Environ Health 1976; 31: 310-6.

Rogers JJ, Kabel RH. The sequence IID  engine oil test.  SAE Paper
  #780931, Society of Automotive Engineers, Warrendale, PA 1978.

Rosen JF, Chesney RW, Hamstra A, De Luca HF, Mahaffey KR. Reduction
  in 1,25-dihydroxy vitamin D in children with increased lead
  absorption.  New England Journal of  Medicine 1980a; 302: 1128-31.

-------
                                 -25-
 Rosen JF, Chesney RW, Hamstra A, Deluca HF, Mahaffey KR.  Reduction
   in 1,25-dihydroxyvitamin D in children with increased lead
   absorption.  Brown SS, Davis DS ed.  Organ Directed Toxicity
   Chemical Indices and Mechanisms.  Pergamon Press 1980b.

 Rosen JF, Chesney RW. Circulating calcitriol concentrations in
   health and disease. Journal Pediatric (St. Louis)  1983;  103: 1-7.

 Rowley DW. Exhaust system considerations for heavy duty trucks.
   SAE Paper #770893, Society of Automotive Engineers, Warrendale,
   PA 1977.

 Royal Comission on Environmental Pollution. Ninth report - lead
   in the environment.  Cmnd 8852, Her Majesty's Stationery Office,
   London, England 1983.

 Rowe R, Chestnut L. Ozone and asthmatics in Los Angeles: a benefits
   analysis.  Report for Office of policy Analysis by Energy and
   Resource Consultants, 1984.

 Rummo JH, Routh DK, Rummo NJ, Brown JF. Behavioral and neurological
   effects of symptomatic and asymptomatic lead-exposure in children.
   Arch Env He 1979; 34(2):  120-4

 Russell JA, Tosh JD, Johnson AA. Performance of army engines with
   leaded and unleaded gasoline. AFLRL-21, Southwest Research
   Institute, San Antonio, Texas (NTIS IAD-766 760) 1973.

 SAI.  Simulations of the Regional Air Quality Impacts of
   Industrial Emission Controls.  Final report to U.S. EPA,
   Economic Analysis Division 1984.

 Sanders M. U.S. Postal Service, personal communication.

 SAS Users Guide. Statistical Analysis System, inc.  Gary,
   NC 1982.

 Sassa S, Garnick JL, Garnick S, Kappas A, Levere RD. Studies in
   lead poisoning. I: Microanalysis of erythrocyte protoporphyrin
   levels by spectrofluorometry in the detection of chronic lead
   intoxication  in the subclinical range. Biochem Medical 1973;
   8: 135-48.

 Schwartz J, Leggett J, Ostro B, Pitcher H, Levin R.  Costs and
   Benefits of Reducing Lead in Gasoline.  U.S. EPA, Office of
   Policy Analysis 1984a.

/Schwartz J, Janney A, Pitcher H. The relationship between gasoline
   lead and blood lead.  U.S. EPA, Office of Policy Analysis 1984b.

 Schwartz J. Blood lead and blood pressure.  Memo to EPA Central
   Docket EN-84-05, 1984c September 7.

-------
                                -26-


Schwartz J. Change in aromatics emissions due to lead phasedown.
  Memo to EPA Central Docket EN-84-05,  1984d September 17.

Schwartz J. Availability and severity of catalytic reforming
  equipment.  Memo to EPA Central Docket EN-84-05, 1984e.
  September 11.

Schwochert HW. Performance of a catalytic converter that operates
  with nonleaded fuel. SAE Paper #690503, Society of Automotive
  Engineers, Warrendale, PA 1969.

Seigneur C, Saxena P, Roth P. Preliminary results of acid rain
  modeling, submitted at a Specialty Conference on Atomspheric
  Deposition sponsored by the Air Pollution Control Association;
  Detroit, Michigan 1982 November 7-10.

Seigneur C, et al. Modeling studies of  sulfate and nitrate chemis-
 try:  the effect changes in sulfur dioxide, nitrogen oxide, and
 reactive hydrocarbon levels.  final report 1984.

Shah BV,  Surregr; Standard Errors of Regression Coefficients
  from Sample Survey Data.  Research Traingle Park, NC,
  Research Triangle Institute 1982.

Shepard CD. Instrumentation and computer techniques aid  automotive
  exhaust system design for low noise levels; materials  to reduce
  system corrosion. SAE Paper #690006,  Society of Automotive
  Engineers, Warrendale, PA 1969.

Sherwin RP, et al. Sequestration of exogenous peroxidase in the
  lungs of animals exposed to continuous 0.5 ppm nitrogen oxide.
  Fed Proc 1977; 36:1091.

Shier DR, Hall A. Analysis of housing and data collected in a lead
  based paint survey in Pittsburgh.  National Bureau of  Standards
  1977.

Shlossman M, Brown M, Shapiro E, Dziak  R. Calcitonin effects on
  isolated bone cells. Calcif Tissue Int 1982; 34: 190-6.

Shy CM, et al. The Chattanooga school children study: effects of
  community exposure of nitrogen dioxide.  I Methods, description
  of polllutant exposure and results of  ventilatory function test-
  ing.  J Air Pollut Control Assoc 1970; 20 (8):539-45.

Shy CM, Love GJ. Recent evidence on the  human health effects of
  nitrogen dioxide.  Proceedings of the Symposium on Nitrogen
  Oxides; Honolulu, Hawaii, 1979 April  4-5.

Sigsby J, et al.  Automotive emissions  of ethylene dibromide.
  SAE Paper #820786.  Society of Engineers, Warrendale,  PA 1982.

-------
                                -27-


Silbergeld EK, Adler HS. Subcellular mechanisms of lead neuro-
  toxicity.  Brain Research 1978; 148: 451-67.

Silbergeld EKf Hruska RE, Miller LP, Eng N. Effects of lead ir\ vivo
  and in vitro on GABAergic neurochemistry.  J Neurochem 1980;
  34:1712-8.

Silver W, Rodriguez-Torres R. Electrocardiographic studies in
  children with lead poisoning.  Pediatrics 1968; 41: 1124-7.

Smith M, Delves T, Lansdown R, Clayton B, Graham P.  The effects of
  lead exposure on urban children: the Institute of Child Health/
  Southampton study.  United Kingdom Department of the Environment
  1983.

Smith RS. The feasibility of an injury tax approach to occupational
  safety.  Law and Contemporary poblems 1974; 38: 730-44.

Smith RS. The Occupational Safety and Health Act.  Washington, DC:
  American Enterprise Institute for Public Policy Research, 1976.

Society of Automotive Engineers.  Fuel Economy Measurement - Road
  Test Procedure.  SAE Standard J 1082, Warrendale, PA 1980
  September.

Soloman and Associates.  Letter to the National Petroleum
  Refiners Association.  Available in the U.S. EPA Central
  Docket EN-84-05, 1984.

Sorem SS. Effects of fuel factors on emissions. SAE Paper #710364.
  Society of Automotive Engineers, Warrendale, PA 1971.

Speizer FE, et al. Respiratory disease rates and pulmonary function
  in children associated with NC>2 exposure.  Am Rev Resp Dis 1980;
  121:3-10.

Spengler JD, et al. Sulfur dioxide and nitrogen dioxide levels
  inside and outside homes and the implications on health effects
  research.  Environ Sci Techn 1979; 13:1271-6.

Spengler JD. Nitrogen dioxide inside and outside 137 homes and
  implications for ambient air quality standards and health effects
  research.  Environ Sci Technol 1983; 17:164-8.

SRI International. An estimate of the nonhealth benefits of meeting
  the Secondary National Ambient Air Quality Standards,  a final
  report to the National Commission on Air Quality 1981.

Stambaugh RL, Kopko RJ, Franklin TM. Effects of unleaded fuel and
  exhaust gas recirculation on sludge and varnish formation. SAE
  Paper #720944, Society of Automotive Engineers, Warrendale, PA
  1972.

-------
                                -28-


Sugiura K. Present status of Japanese motor oils and extension of
  oils drain intervals. SAE Paper #780953, Society of Automotive
  Engineers, Warrendale, PA 1978.

Supplemental Analysis of Refining Costs.   Submitted to U.S. EPA
  Central Docket EN-84-05, 1984 September 19.

Tera, et al. Identification of gasoline lead in childrens1  blood
  using isotopic analysis.  Arch Env Health 1985 January.

Thaler R, Rosen S. The value of saving a  life:  evidence from the
  labor market.  Taleckji NE ed.  Household Production and  Consump-
  tion.  New York:  Columbia University Press  1976.

Thomas HV, et al. Lipoperoxidation of lung lipids in rats exposed
  to nitrogen dioxide.  Science 1968; 159:532-4.

Thompson CR, et al. Effects of continous  exposure of navel  oranges
  to NO2. Atm Environ 1970; 4:349-55.

Tosh JD, Johnson AA, Frame EA. Performance of  army engines  with
  leaded and unleaded gasoline. AFLRL-54, Southwest Research
  institute, San Antonio, Texas (NTIS #AD-A005  577) 1975.

Tosh JD. Performance of army engines with unleaded gasoline—
  field study. AFLRL-82, Southwest Research Institute, San  Antonio,
  Texas (NTIS #AD-A032 075) 1976.

Trijonis JC.  Empirical Relationships Between  Atmospheric Nitrogen
  Dioxide and Its Precursors.  U.S. EPA,  Office of Research and
  Development.  Environmental Sciences Research Laboratory,
  Research Triangle Park, NC, 1978 February.

Trijonis JC, et al.  The Relationship of  Ambient NC<2 to Hydrocarbon
  and NOX Emissions.  Draft report from Technology Service  Corpor-
  ation to U.S. EPA under Contract NOX 68-02-2299.  U.S. EPA,
  Office of Research and Development.  Triangle Park, NC 1979.

Turner, Mason and Associates.  Reply on behalf  of the Lead  Indus-
  tries Association, July 1984.  Available in  the EPA Central
  Docket EN-84-05.

Union Carbide, personal communication, 1984.

Unzelman GH, Michalski GW. Octane improvement  update — refinery
  processing, antiknocks, and oxygenates. NPRA Paper #AM-84-43,
  National Petroleum Refiners Association, Washington, DC 1984.

Urban CM, Springer KJ, McFadden JJ. Emissions  control of gasoline
  engines for heavy-duty vehicles. SAE Paper #750903, Society of
  Automotive Engineers, Warrendale, PA 1975.

-------
                                -29-


Urban CM, Kaupert AW. Evaluation of emission control technology
  approaches for heavy-duty gasoline engines. SAE Paper #780646,
  Society of Automotive Engineers, Warrendale, PA  1978.

U.S. Department of Agriculture. Agricultural Statistics, 1980.
  Washington, DC:  U.S. Government Printing Office 1982.

U.S. Department of Energy, Energy Information Administration.
  Annual Report to Congress. 1983.

U.S. Department of Energy, Energy Information Administration.
  Annual Report to Congress. 1984a.

U.S.  Department of Energy, Energy Information Administration,
  Office of Oil and Gas Refinery Evaluation Modeling System (REMS),
  model documentation.  1984b July.

U.S. Department of Energy. Refinery Evaluation Modeling System
  (REMS), 1983 Analysis of Verification.  1985 January.

U.S. Department of Health, Education, and Welfare.  Limitations of
  activity and mobility due to chronic conditions.  Public Health
  Service, Vital and Health Statistics Series 10, 1973a; (96).

U.S. Department of Health, Education, and Welfare. Prevalance of
  selected chronic respiratory conditions:  United States - 1970.
  Public Health Service, Vital and Health Statistics Series 10,
  1973b; (84).

U.S. Department of Health and Human Services, Public Health Ser-
  vice.  Blood carbon monoxide levels in persons 3-74 years of  age:
  United States, 1976-80. Advance Data 1982 March 17; (76).

U.S. Environmental Protection Agency. Ozone SIP Data Base and
  Summary Report.  Office of Air Quality Planning and Standards,
  Research Triangle Park, NC.

U.S. Environmental Protection Agency.  Air Quality Criteria for
  Ozone and Other Photochemical Oxidants.  Office of Research  and
  Development, 1978 April.

U.S. Environmental Protection Agency, Carcinogen Assesmment Group.
  The Carcinogen Assessment Group's Final Report on Population  Risk
  to Ambient Benzene Exposure.  1979a January 10.

U.S. Environmental Protection Agency, Office of Mobile Source  Air
  Pollution Control.  Draft Regulatory Analysis:  Heavy-Duty Diesel
  Particulate Regulations.  U.S. EPA; Ann Arbor, Michigan 1979b.

U.S. Environmental Protection Agency.  Regulatory Impact Analysis
  of the National Ambient Air Quality Standards for Carbon Monoxide,
  Office of Air Quality Planning and Standards, 1980.

-------
                               -30-


U.S. Environmental Protection Agency.  The Costs of Controlling
  Emissions of 1981 Model Year Automobiles.  Office of Mobile
  Source Air Pollution Control, 1981 June.

U.S. Environmental Protection Agency. 1982 NCLAN Annual Report.
  Environmental Research Lab, Corvallis, Oregon 1982a.

U.S. Environmental Protection Agency. Air Quality Criteria for
  Oxides of Nitrogen (Criteria Document).  Research Triangle Park,
  NC  1982b.

U.S. Environmental Protection Agency.  Review of the National
  Ambient Air Quality Standards for Nitrogen Oxides:  Assessment of
  Scientific and Technological Information. OAQPS, NC, 1982c August.

U.S. Environmental Protection Agency. NAAQS Environmental Impact
  Statement for N02.  Research Triangle Park, NC 1982d.

U.S. Environmental Protection Agency. Cost and Economic Assessment
  of Regulatory Alternatives for NO2 NAAQS (DRAFT).  Research Tri-
  angle Park, NC 1982e.

U.S. Environmental Protection Agency. National Air Pollutant
  Emissions Estimates, 1940 - 1980.  Monitoring and Data Analysis
  Division 1982f January.

U.S. Environmental Protection Agency.  Regulation of Fuel and Fuel
   Additives. 47 Federal Register 49382, 1982g October 28.

U.S. Environmental Protection Agency.  Motor Vehicle Tampering
  Survey 1982.  National Enforcement Investigation Center, 1983a
  April.

U.S. Environmental Protection Agency, OAR, QMS.  Anti-Tampering and
  Anti-Misfueling Programs to Reduce In-Use Emissions from Motor
  Vehicles.  1983b December 31.

U.S. Environmental Protection Agency.  The API Study and Its
  Possible Human Health Implications,  memo from Al Lorang (Chief,
  Technical Support Staff) to Charles Gray, Jr. (Director, Emission
  Control Technology Division), 1983c May 16.

U.S. Environmental Protection Agency, Environmental Criteria and
  Assessment Office. Air Quality Criteria for Lead.  1983d.

U.S. Environmental Protection Agency, Office of Mobile Source Air
  Pollution Control, Standards Development and Support Branch.
  Issues Analysis - Final Heavy-Duty Engine HC and CO Standards.
  U.S. EPA; Ann Arbor, Michigan 1983e.

U.S. Environmental Protection Agency.  Preliminary Draft Revised
  Air Quality Criteria for Ozone and Other Photochemical Oxidants.
  Office of Researach and Development 1983f.

-------
                                -31-


U.S. Environmental Protection Agency.  The Feasibility, Cost,  and
  Cost-Effectiveness of Onboard Vapor Control, prepared by Glenn
  Passavant, 1984a March.

U.S. Environmental Protection Agency.  VOC/Ozone Relationships from
  EKMA. memo from Warren Freas (Air Management Technology Branch)
  to Alan McGartland (Benefits Branch), 1984b January 27.

U.S. Environmental Protection Agency.  Response to Public Comments
  on EPA's Listing of Benzene Under Section 112 and Relevant Proce-
  dures for the Regulation of Hazardous Air Pollutants. OAQPS,
  1984c.

U.S. Environmental Protection Agency, Office of Air, Noise and
  Radiation.  Regulation of Fuels and Fuel Additives:  Lead Phase
  Down.  Federal Register 1984d August 2; 31032.

U.S. Environmental Protection Agency.  Motor Vehicle Tampering
  Survey 1983.  National Enforcement Investigation Center, 1984e.

U.S. Environmental Protection Agency, Office of Policy Analysis.
  Preliminary Regulatory Impact Analysis of Proposed Rules Limiting
  the Lead Content of Gasoline.  1984f July 23.

U.S. Environmental Protection Agency.  Regulatory Impact Analysis
  Guidelines.  1984g.

U.S. Environmental Protection Agency. Draft Revised Air Quality
  Criteria for Ozone and Other Photochemical Oxidants. Office  of
  Research and Development 1984h.

U.S. Environmental Protection Agency.  Regulation of Fuels and Fuel
  Additives; Banking of Lead Rights.  50 Federal Register 718, 1985
  January 4.

U.S. Office of Management and Budget. Interim Regulatory Impact
  Analysis Guidance. 1981, p. 4.

Victery W, Vander AJ, Shulak JM, Schoeps P, Juluis S. Lead, hyper-
  tension, and the renin-angiotensin system in rats. Journal Labor-
  tory Clinic Medical 1982; 99: 354-62.

Violette, Chestnut. Valuing Reductions in Risks:  A Review of the
  Empirical Estimates.  1983.

Viscusi WK. Labor market valuations of life and limb: empirical
  evidence and policy implications.  Public Policy 1978; 26: 359-86

Von Nieding G, et al. Minimum concentration of N02 causing acute
  effects on the respiratory gas exchange and airway resistance in
  patients with chronic bronchitis.  Int Arch Arbeitsmed 1971;
  27:338-48.  U.S. Environmental Protection Agency, Researach
  Triangle Park, NC.

-------
                                -32-


Von Nieding G, et al.  Studies of the acute effect of NC-2 on lung
  function:  influence on diffusion, perfusion and ventilation in
  the lungs.  Int Arch Arbeitsmed 1973; 31:61-72.

Von Nieding G, et al. Acute effects of ozone on lung function of
  men.  VDI-Ber 1977; 270:123-9.

Wade WA III, et al. A study of indoor air quality.  J Air Pollut
  Control Assoc 1975; 25-93-939.

Wagner TO. Motor gasoline and automotive air pollution.  SAE Paper
  #710485, Society of Automotie Engineers, Warrendale, PA 1971.

Wagner TO, Russum LW. Optimum octane number for unleaded gasoline.
  SAE Paper #720552, Society of Automotive Engineers, Warrendale,
  PA 1973.

Ware JH, et al.  Passive smoking, gas cooking and respiratory
  health of children living in six cities.  America Review of
  Respiratory Diseases 1984; 129.

Weaver C. Particulate control technology and particulate standards
  for heavy-duty diesel engines. SAE Paper #840174, Diesel Particu-
  late Traps: 140, Society of Automotive Engineers, Warrendale, PA
  1984.

Weaver CS.  The Effects of Low-Lead and Unleaded Fuels on Gasoline
  Engines.  Prepared for:  U.S. Environmental Protection Agency,
  Office of Policy Analysis.  Energy and Resource Consultants, Inc,
  1984b September 28.

Webb RC, Winquist RJ, Victery W, Vander AJ. In vivo and in vitro
  effects on lead on vascular reactivity in rats.  AM J Physiol
  1981; 24:  H211-6.

Wedeen RP, Maesaka JK, Weiner B, Lipat GA, Lyons MM, Vitale LF,
  Joselow MM. Occupational lead nephropathy.  American Journal of
  Medicine 1975; 59: 630-41.

Weekly Petroleum Status Report. 1984, various issues.

Weinstein MC, Stason WB.  Allocation of Resources to Manage Hyper-
  tension.  N Eng J Med 1977; 296(13):  732-9.

White JM, Harvey DR. Defective synthesis of A and B globin chains
  in  lead poisoning.  Nature (London) 1972; 236: 71-3.

Whittemore A, Korn EL. Asthma and air pollution in the Los Angeles
  area. American Journal of Public Health 1980; 70.

Wibberly DG, Khera AK, Edwards JH, Rushton DI. Lead levels in human
  placentae from normal and malformed births.  Journal of Medical
  Genetics 1977; 14: 339.

-------
                                -33-


Williams BJ, Griffith WH III, Albrecht CM, Pirch JH, Hejtmancik MR
  Jr. Effects of chronic lead treatment on some cardiovascular
  responses to norepinephrine in the rat.  Toxicology Appl Pharma-
  col 1977; 40: 407-13.

Williams BJ, Goldman D, Hejtmancik MR, Ziegler MG.  Noradrenergic
  effects of lead in neonatal rat.  Pharmacology 1978; 20(3):  186.

Williams BJ, Hejtmancik M.  Time and level of perinatal lead-
  exposure  for development of norepinephrine cardiotoxicity.   Res
  Comm CP 1979; 24(2):  367-76.

Willits N, Ott W.  Modeling the dynamic response of an automobile
  for air pollution exposure studies.  Environraetrics, Summaries of
  Conference Presentations 1981; 81:  104-5.

Winneke G,  Brockhaus A, Baltissen R. Nerve behavioral and systemic
  effects of longterm blood lead elevation in rats. Arch Toxicol
  1977; 37: 247-63.

Winneke G.  Impaired intelligence in children from environmental
  lead (letter).  Mun Med Woe 1979; 121(26): 865.

Winneke G,  Lilienthal H, Werner W. Task dependent nerve behavioral
  effects of lead in rats. Arch Toxicol  1982a; Supp 5: 84-93.

Winneke G,  Hrdina K-G, Brockhaus A. Neuropsychological studies in
  children  with elevated tooth-lead concentrations, part 1: pilot
  study. Int Arch Occupational and Environmental Health 1982b;
  51:  169-83.

Winneke G,  Kramer U, Brockhaus A, Ewers U, Kujanek G, Lechner  H,
  Janke W.  Neuropsychological studies in children with elevated
  tooth lead concentrations.  Part II:  extended study.  Int Arch
  Occup Environ Health 1983; 51:  231-52.

Winneke G,  Beginn U, Ewert T, Havestadt C, Kramer U, Krause C,
  Thron HL, Wagner HM.  Studie zur erfassung subklinischer blei-
  wirkungen auf das nervensystem bei kindern mit bekannter prana-
  taler exposition in Nordenham. [Study on the determination of
  subclinical lead effects on the nervous system of Nordenham
  children  with known pre-natal exposure.]  BGA-Berichte 1984.

Wintringham JS, et al. Car maintenance expense in owner service
  with leaded and nonleaded gasolines.  SAE Paper #720499, Society
  of Automotive Engineers, Warrendale, PA 1972.

Wong GL. Actions of parathyroid hormone and 1,25-dihydroxcholecal-
  ciferol on citrate decarboxylation in osteoblast-like bone cells
  differ in calcium requirement and in sensitivity to trifluopera-
  zine.  Calcif Tissue Int 1983; 35: 426-31.

-------
                                -34-


World Health Organization, United Nations Environmental Program;
  Lead: Environmental Health Criteria 3, Geneva,  Switzerland 1977.

Worthen RP, Tunnecliffe TN. Temperature controlled engine valves.
  SAE Paper #820501, Society of Automotive Engineers,  Warrendale,
  PA 1982.

WR Grace, Davison Chemical Division.  Davison Catalagram #69, 1984.

Yip R, Norris TN, Anderson AS. Iron status of children with ele-
  vated blood lead concentrations. Journal Pediatr (St. Louis)
  1981; 98: 922-5.

Young CT, Grimes DA. Erosion mechanisms of automotive  spark plug
  electrodes. SAE Paper #780330, Society of Automotive Engineers,
  Warrendale, PA 1978.

Yule W, Landsdown R, Millar IB, Urbanowicz MA.  The relationship
  between blood lead concentrations, intelligence and  attainment in
  a school population: A pilot study. Dev Medical Child Neurology
  1981; 23:567-76.

Yule w, Lansdown R. Lead and children's development:  recent
  findings, presented at international conference: Management
  and Control of Heavy Metals in the Environment, 1983 September;
  Heidelberg, West Germany.

Yule W, Urbanowicz M-A, Lansdown R, Millar IB.  Teacher's ratings
  of children's behaviour in relation to blood lead levels.
  British Journal of Developmental Psychology 1984.

Zagraniski R, Leaderer B, Stolwuk J. Ambient sulfates, photochemical
  oxidants and acute adverse health effects: an epidemiologic study.
  Environmental Research 1979; 19.

Ziskind RA, et al.  Carbon Monoxide Intrusion in Sustained-Use
  Vehicles, #SAI-068-80-535. U.S. EPA; Research Triangle Park,
  NC, 1979 November.

-------
                            APPENDIX A




                        REFINERY PROCESSES






     In refining, crude oil is first separated by molecular size



into fractions, each of which can be blended directly into final



petroleum products or processed further.  In the downstream pro-



cessing operations, the molecular size and structure of petroleum



fractions are altered to conform to desired characteristics of



refined products.  Table li-l in the text classifies the various



refinery processes according to their principal functions.  The



actual processing configuration will depend on the characteristics



of the crude oil processed and on the desired final product mix.



These major processing steps are described briefly below.



     Fluid Catalytic Cracking uses high temperature in the pres-



ence of a catalyst to convert or "crack" heavier fractions into



lighter products, primarily gasoline and distillates.   Feed is



brought to process conditions (1000°F and 20 pounds per square



inch pressure  [psi]) and then mixed with a powdered catalyst in



a reaction vessel.  In the reactor, the cracking process is



completed and the hydrocarbon products pass to a fractionating




section for separation.



     Coke, a coal-like by-product, is formed on the catalyst as a



result of the cracking reaction.  Coked catalyst is transferred




from the reactor to a regenerator vessel where air is injected




to burn the coke to CO and CO2.  The regenerator flue gases are



passed through cyclones and, sometimes, electrostatic precipita-



tors, to remove entrained catalyst.  They are then vented to the

-------
                               A-2
atmosphere or sent to a CO boiler where carbon monoxide is burned




to produce CC>2.  The regenerated catalyst is returned to the



reactor.



     Hydrotreating (also known as hydrodesulfurization) is a




catalytic process designed to remove sulfur, nitrogen, and



heavy metals from petroleum fractions.  Feed is heated to process



temperatures (650° to 705°F), mixed with hydrogen, and fed to



a reactor containing a fixed bed of catalyst.  The primary reac-



tions convert sulfur compounds in the feed to hydrogen sulfide



(H2S) and the nitrogen compounds to ammonia.  The B^S and



ammonia are separated from the desulfurized product; the t^S



is sent to sulfur recovery facilities.



     Catalytic reforming is used to upgrade low-octane naphtha



to produce high-octane gasoline blending stocks.   The flow pat-



tern is similar to that of hydrotreating except that several



reactor vessels are used.   The required temperature is about



1000"F and the required pressure is about 200 pounds per square



inch.  Reforming catalysts are readily poisoned by sulfur,



nitrogen, or heavy metals, and therefore the feed is normally



hydrotreated before being charged to the reforming unit.



     In hydrocracking the cracking reaction takes place in the



presence of hydrogen.  The process produces high quality desul-



furized gasoline and distillates from a wide variety of feed-



stocks.  The process employs one or more fixed bed reactors and



is similar in flow to the hydrotreating process.   Process




conditions are 800°F and 2000 psi.  Like hydrotreating, hydro-

-------
                               A-3






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



recovery.




     Coking  is another  type of cracking which does not employ a



catalyst or  hydrogen.   The process is utilized to convert heavy



fuel oils  into light products and a solid residue (coke).  Feed



is brought to process conditions (900°P and 50 psi) and fed to



the coking vessel.  Cracked products are routed to a fractionation



section.   Coke accumulates in the vessel and is drilled out about



once a day.  In one version of the coking process, fluid coking,



a portion  of the coke is used for process fuel and the balance



is removed as small particles.



     Acid gas treating  and sulfur recovery units are used to



recover hydrogen sulfide (H2S) from refinery gas streams and



convert it to elemental sulfur.  Sour gas containing H2S is



produced in  several refinery units, particularly cracking and



hydrotreating.  In the  acid gas treating units, f^S is removed



from the fuel gas by absorbing it in an alkaline solution.  This



solution,  in turn, is heated and steam-stripped to remove the



H2S to form sulfur and water.  Sulfur recovery is high but



never 100%.  The remaining sulfur is incinerated and discharged



to the atmosphere or removed by a tail gas treating unit.



     The purpose of the tail gas treating unit is to convert any



remaining sulfur compounds from the sulfur recovery unit to




elemental sulfur.  There are several processes available, the



most common of which are the Beavon and SCOT processes.  In both



processes, sulfur compounds in the sulfur unit tail gas are

-------
                               A-4






converted to H2S.  The Beavon process converts H2S to sulfur



through a series of absorption and oxidation steps.   The SCOT



process concentrates the f^S and returns it to the sulfur



recovery facilities.  In both processes, the treated tail gas



is virtually free of sulfur compounds when released  to the



atmosphere.

-------
                           APPENDIX B



                         THE FLEET MODEL





     We developed a fleet model to predict the number of cars and



gasoline powered trucks (for six weight classes) on the road; the



model also forecasts the number of miles driven per year for each



vehicle type for the years 1985-1992.  In this document we have



used our fleet model for several different purposes.  Estimates



of total miles driven per year by vehicle type, age and fuel



consumed are used to estimate the maintenance savings attributable



to the use of lower lead or unleaded fuel.  Estimates of lifetime



vehicle-miles traveled by new misfuelers in each year provide



the basis for estimating the value of the reduced emissions due



to reduced misfueling.



B.I  Vehicles on the Road



     The extrapolation procedure for total vehicle miles per year



by age and type of vehicle relies on two basic sets of data:



1)  Vehicles on-the-road for 1968-1983 as published annually by



    the Motor Vehicle Manufacturers Association(MVMA) in Motor



    Vehicle Facts and Figures.  Data are available for cars



    (Table B-l) and trucks (Table B-2) ages 1 to 15 years, as



    well as 16 and older.   No breakdown of trucks by weight and



    engine type (diesel versus gasoline) is available, however.



2)  Data Resources, Inc. (DRI) projections of the total fleet



    size for cars, as well as annual sales for cars and trucks.



Using these data sets, the fleet model projected how many cars



and trucks (by age) would be on the road for each of the years,



1985 through 1992.  For trucks, the projections reflect only

-------
                                B-2
gasoline-powered vehicles and the fleet is divided into six



weight classes.



     The model required two sets of inputs:  an inventory of



vehicles on the road by age and a set of survival rates for



determining how many vehicles in each age category would still be



on the road the next year.   The initial inventory for vehicles



aged "New1 through 15 years of age is taken from the 1983 column



in Tables B-l and B-2.  The category of vehicles aged 16 and over



is broken into 15 additional age categories.   The procedures we



used are discussed in sections B.I.a for cars and B.l.b. for



trucks.



     The survival rates used in the forecast  are based on an



analysis of historical survival rates.   These historical survival



rates are derived by dividing the number of vehicles in a model



year cohort in one year by the number of vehicles in the same



cohort the previous year.  For cars, the historical survival



rates are given in Table B-3, while Table B-4 presents the rates



for trucks.  As both tables show, there is significant variation



in the rates over time.  A statistical analysis of the data



showed that both the overall scrappage rate (defined as total



vehicle retirements divided by fleet size) and the fraction of



the total fleet that each model year comprised were important



explanatory variables.  So, for the forecast, we adjusted the



survival rates to reflect changes in the scrappage rates and the



relative size of the cohort in each year.  As explained below,



the scrappage rates used for cars are derived from the DRI

-------
                               B-3
Table B-l  Cars on the Road 1968-1983
Year
Age of
Car
New
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 +
Total
1968


6182
8122
8836
8939
7667
7058
6183
4657
4615
3347
1709
1990
1612
1496
743
623
1517
75358
1969


6467
8927
8054
8798
8855
7532
6829
5804
4087
3726
2452
1188
1421
1139
1063
525
1578
80449
1970


6288
9299
8816
7878
8538
8506
7116
6268
5058
3267
2776
1692
799
996
794
753
1583
80449
1971


5927
8888
9280
8802
7772
8313
8171
6651
5624
4274
2525
2035
1183
563
730
580
1804
83138
1972


7169
8915
8851
9122
8596
7499
7930
7583
5920
4713
3343
1824
1413
805
389
526
1813
86439
1973


7988
10158
8715
8612
8881
8291
7120
7333
6715
4963
3698
2470
1268
967
548
274
1780
89805
1974


6433
11269
10147
8622
8493
8615
7931
6624
6531
5710
3976
2824
1813
901
682
391
1621
92608
1975


4684
9763
11332
10098
8549
8341
8339
7556
6113
5796
4825
3234
2229
1407
689
523
1742
95241
Year
1976
1977   1978   1979   1980    1981    1982
1983
Age of
Car
New
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 +


6472
7683
9746
11130
9872
8249
7966
7774
6856
5361
4888
3923
2578
1740
1083
526
1943


7177
9557
7477
9594
10854
9563
7866
7449
6963
5859
4416
3887
3023
1969
1315
818
2093


7426
10382
9483
7291
9431
10559
9140
7326
6784
6087
4917
3589
3093
2369
1545
1021
2496


7288
10699
10219
9203
6990
9004
9965
8431
6573
5909
5034
3999
2862
2460
1874
1223
2930


5868
10402
10483
9931
8900
6682
8499
9151
7544
5653
4939
4049
3172
2280
1969
1516
3514


5140
8818
10245
10290
9758
8735
6463
8050
8458
6791
4929
4238
3369
2635
1910
1654
4346


4399
8280
8825
10075
10155
9661
8471
6190
7498
7629
5989
4243
3581
2822
2208
1609
5220


5044
7429
8273
8749
10014
10038
9434
8195
5867
6885
6798
5239
3632
3052
2395
1869
6037
 Total   97818  99904 102957 104677 104564 105839 106867 108961

-------
                               B-4
Table B-2  Trucks on the Road 1968-1983
Year
Age of
Truck
NEW
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 +
TOTAL
Year
Age of
Truck
NEW
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 +
1968


1058
1452
1512
1380
1208
1035
889
692
749
682
461
508
514
537
370
425
2182
15685
1976


1893
2148
2732
2799
2346
1697
1635
1731
1345
1220
1191
1024
828
642
503
351
2455
1969


1262
1581
1447
1495
1357
1177
1004
860
661
706
632
421
459
459
480
330
2242
16586
1977


2177
2746
2109
2689
2752
2291
1639
1573
1645
1267
1129
1096
922
736
566
442
2422
1970


1263
1881
1536
1428
1483
1339
1154
975
826
621
658
583
383
417
414
432
2278
17686
1978


2533
3240
2743
2076
2656
2681
2227
1567
1509
1554
1189
1043
998
832
663
506
2531
1971


1193
1736
1872
1496
1398
1441
1298
1112
927
774
585
610
532
347
376
369
2383
18465
1979


2402
3541
3231
2679
2006
2589
2587
2140
1501
1435
1459
1111
958
906
749
595
2687
1972


1637
1784
1744
1858
1468
1372
1409
1260
1066
877
721
525
544
471
303
328
2380
19773
1980


1362
3765
3663
3332
2750
2046
2641
2609
2163
1540
1429
1452
1082
919
856
688
2948
1973


1883
2385
1753
1709
1825
1446
1336
1363
1208
1005
817
663
474
487
419
268
2356
21412
1981


1244
2225
3808
3633
3318
2722
2005
2567
2516
2078
1438
1337
1348
999
842
779
3192
1974


1834
2829
2396
1742
1698
1804
1418
1305
1309
1145
941
752
603
427
434
370
2289
23312
1982


1291
2099
2183
3817
3621
3297
2689
1950
2469
2399
1942
1364
1258
1256
930
779
3628
1975


1326
2739
2848
2384
1730
1668
1779
1395
1273
1256
1085
884
697
554
388
391
2393
24813
1983


1564
2214
2106
2160
3779
3566
3230
2615
1887
2368
2247
1848
1288
1184
1177
870
4026
TOTAL 26560  28222  30565  32583  35268  36069  36987  38143

-------
                               B-5
Table B-3  Year to Year Survival Rates for Cars(historical)
Year

Age of
Car
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16+
Year

Age of
Car
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16+
1968
-1969


1.242
.992
.996
.991
.982
.968
.939
.878
.807
.733
.695
.714
.707
.711
.707
2.533
1976
-1977


1.260
.973
.984
.975
.969
.954
.935
.896
.855
.824
.795
.771
.764
.756
.755
3.979
1969
-1970


1.284
.988
.978
.970
.961
.945
.918
.871
.799
.745
.690
.673
.701
.697
.708
3.015
1977
-1978


1.238
.992
.975
.983
.973
.956
.931
.911
.874
.839
.813
.796
.784
.785
.776
3.051
1970
-1971


1.411
.998
.998
.987
.974
.961
.935
.897
.845
.773
.733
.699
.705
.733
.730
2.396
1978
-1979


1.261
.984
.970
.959
.955
.944
.922
.897
.871
.827
.813
.797
.795
.791
.792
2.870
1971
-1972


1.161
.996
.983
.977
.965
.954
.928
.890
.838
.782
.722
.694
.680
.691
.721
3.126
1979
-1980


1.299
.980
.972
.967
.956
.944
.918
.895
.860
.836
.804
.793
.797
.800
.809
2.873
1972
-1973


1.238
.978
.973
.974
.965
.949
.925
.886
.838
.785
.739
.695
.684
.681
.704
3.384
1980
-1981


1.309
.985
.982
.983
.981
.967
.947
.924
.900
.872
.858
.832
.831
.838
.840
2.867
1973
-1974


1.315
.999
.989
.986
.970
.957
.930
.891
.850
.801
.764
.734
.711
.705
.714
5.916
1981
-1982


1.293
1.001
.983
.987
.990
.970
.958
.931
.902
.882
.861
.845
.838
.838
.842
3.156
1974
-1975


1.470
1.006
.995
.992
.982
.968
.953
.923
.887
.845
.813
.789
.776
.765
.767
4.455
1982
-1983


1.241
.999
.991
.994
.988
.977
.967
.948
.918
.891
.875
.856
.852
.849
.846
3.752
1975
-1976


1.188
.998
.982
.978
.965
.955
.932
.907
.877
.843
.813
.797
.781
.770
.763
3.715





















-------
                                B-6
Table B-4  Truck Year to Year Survival Rates (historical;
Year

Age of
Truck
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16+
Year

Age of
Truck
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
15-16+
1968
-1969


1.151
.997
.989
.983
.974
.970
.967
.955
.943
.927
.913
.904
.893
.894
.892
5.275
1976
-1977


1.123
.982
.984
.983
.977
.966
.962
.950
.942
.925
.920
.900
.889
.882
.879
6.900
1969
-1970


1.273
.972
.987
.992
.987
.980
.971
.960
.939
.932
.922
.910
.908
.902
.900
6.903
1977
-1978


1.149
.999
.984
.988
.974
.972
.956
.959
.945
.938
.924
.911
.902
.901
.894
5.726
1970
-1971


1.278
.995
.974
.979
.972
.969
.964
.951
.937
.942
.927
.913
.906
.902
.891
5.516
1978
-1979


1.132
.997
.977
.966
.975
.965
.961
.958
.951
.939
.934
.919
.908
.900
.897
5.310
1971
-1972


1.133
1.005
.993
.981
.981
.978
.971
.959
.946
.932
.897
.892
.885
.873
.872
6.450
1979
-1980


1.414
1.034
1.031
1.027
1.020
1.020
1.009
1.011
1.026
.996
.995
.974
.959
.945
.919
4.955
1972
-1973


1.191
.983
.980
.982
.985
.974
.967
.959
.943
.932
.920
.903
.895
.890
.884
7.183
1980
-1981


1.187
1.011
.992
.996
.990
.980
.972
.964
.961
.934
.936
.928
.923
.916
.910
4.640
1973
-1974


1.171
1.005
.994
.994
.988
.981
.977
.960
.948
.936
.920
.910
.901
.891
.883
8.541
1981
-1982


1.233
.981
1.002
.997
.994
.988
.973
.962
.953
.935
.949
.941
.932
.931
.925
4.657
1974
-1975


1.333
1.007
.995
.993
.982
.986
.984
.975
.960
.948
.939
.927
.919
.909
.901
6.468
1982
-1983


1.153
1.003
.989
.990
.985
.980
.972
.968
.959
.937
.952
.944
.941
.937
.935
5.168
1975
-1976


1.123
.997
.983
.984
.981
.980
.973
.964
.958
.948
.944
.937
.921
.908
.905
6.279





















-------
                                B-7


forecast of sales and fleet size  for cars.  Since DRI does not

forecast fleet size for trucks, we modified the adjustment pro-

cedure for truck year to year survival rates to use car scrappage

rates.  A detailed description of the adjustment procedure is

given in the next two sections.


B. 1.a  Projection Model for Cars on the Road

     We begin with a description of the projection procedure for

cars.  Let C^j denote cars of age i in year j.  Thus C^o,1985

represents the number of ten year old cars in 1985.  Then,

letting S^j be the probability of a car of age i in year j still

being on the road in year j + 1, we have C^ + i ,j + i= C^jX S-[j.

Further, we have

         RJ = SUM Cj_jx(l - Sij )
               i
where RJ is total retirements in the fleet in year j.  However,

we also have

         R-J = Fj + 1 - Fj + N-J,

where Fj is the size of the fleet in year j and Nj is new car

sales in that year.  Thus, we have two equations which determine

retirements and these must be consistent if the model is to

operate properly.  We used the DRI forecast as the source of the

overall loss rates and adjusted year to year survival rates until

they gave the same total losses as the DRI forecast.

     We adjusted age specific survival rates by using the

regression coefficient of the scrappage rate in the model

(B.I)    Sij = a0 + bSRj + c(Cij/SUM Cij) - e

where SR-; is the overall scrappage rate for year j and

-------
                                B-8






Ci;j/SUM Vjj is the fraction of the fleet that vintage i is of the



fleet in year j.  The survival rates were varied by changing the



scrappage rate in equation B.I until the total losses from the



survival rates were within 0.01% of those in the DRI forecast.



     In equation B.I the scrappage rate value necessary to achieve



equality in total losses with the DRI forecast was typically



larger than the DRI scrappage rate.  Thus, the survival rate



model overpredicts the size of the fleet in the next year when



compared to the DRI forecast.  However, the advantage of adjusting



using the model is that the survival rates maintain a consistent



pattern by vehicle age as model forecasts of retirements from



each cohort in the fleet are changed in order to obtain equality



with the DRI forecast.



     For the car fleet, the adjustment equations were estimated



using sixteen years of data from the MVMA for cars on the road in



July of each year.  The resulting estimates are given in Table



B-5.  For forecasting purposes, the results of the regression



analysis were not used for the first two age groups.  Because new



cars on the road on July 1st of the year fluctuated sharply as a



percentage of total car sales for the model year, we assumed they



were equal to 75% of total sales for that model year.  One year



old cars were assummed to be 99.9 percent of the appropriate new



car model year sales and two year old cars were assumed to be



99.5 percent of the original sales level.  One and two year old



cars show little variation in survival rates and the correction



mechanism occassionally gave survival rates greater than one —



therefore these rates were imposed.

-------
                           B-9
Table B-5.
Regression Coeficients for the Car Year to Year
Survival Rates
Survival
Rate
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
Constant
1.426
1.009
.9854
1.019
.9967
.9857
1.013
1.013
.8523
.5212
.5580
.6533
.6067
.6810
.7057
Scrappage
Rate
-.000239
-.003030
-.003828
-.004151
-.004715
-.004911
-.006537
-.006537
-.01218
-.001298
-.004541
-.008412
-.000991
-.005657
-.004773
Cohort
Weight
-1.880
.0721
.2983
-.05617
.1383
.1298
-.3060
-.7581
1.619
5.467
5.969
5.550
6.832
7.202
7.637

-------
                               B-10






     The other two pieces of information necessary to start the




forecast procedure were a 1983 inventory of vehicles and an



initial set of survival rates.  The initial inventory data were



extended from the 15 years available in the data to 30 years.



The extension was necessary because consumers have been keeping



their cars longer, and consequently, older vehicle have been



steadily growing as a fraction of the fleet.   As Tables B-l and



B-2 indicate, the cohort of vehicles aged 16 years old and



greater has been growing in size during the last 15 years, in



both absolute and relative terms.



     We began the inventory extension by noting, as Tables 3 and



4 show, that the survival rates for vehicles aged 11 through 15



were similar in size and showed a consistent pattern of change



from year to year.  Therefore, we assumed that the survival rate



for older vehicles was equal to the rate for 14 to 15 year-old



cars and that it changed in the same way as this rate did in



response to a change in the overall scrappage rate.   Using this



assumption, we constructed our estimates of 15 through 30 year



old cars in 1983 iteratively.   We began in 1968 with 15 year old



cars and estimated the number  of 16 year old cars in 1969 using



the survival rate for 14 to 15 year old cars for 1968 to 1969.



Thus, in 1969 we had an estimate for 15 and 16 year old cars.   We



then repeated the same process for two cohorts to get estimates



of 15, 16 and 17 year-old cars in 1970.  A similar process was



carried on through 1983 to give us estimates of cars 15 through



30 years of age.  In 1983, the total number of cars resulting



from our estimate of the individual cohorts was 5 percent less

-------
                               B-ll






than the actual total for cars over 15 years of age.  We corrected



for this error by adjusting each age group by



(B.2)    (1 + r)i - 15,




where i is the age of the cohort and r was chosen to yield the



desired equality.  This adjusts older vehicles more than younger



vehicles to reflect their longer extrapolation period.  The value



for r was .0181 for cars.  The largest correction, for the oldest



cohort, was a 31 percent increase.




     The other required set of initial information was the



1983-1984 survival rates.   These were the 82 to 83 set with



rates for age 15 and older cohorts set to 0.84.  This was the



rate for 14 year old cars in 1983. This was was used because the



assumption had been used in the inventory extension process and



had worked quite well .



     With these two inputs determined, the model could forecast



the total number cars on the road for each year from 1985 through



1992; the results are in Table B-6.  As the table shows the



number of older vehicles continues to grow.  By 1992, cohorts



aged 16 and older exceed 12 million cars and are nearly 10



percent of the fleet.  While this is to some extent an artifact



of the extrapolation procedure, the overall size of the pool of



cars over 15 years of age is consistent with existing behavior



within the fleet and the DRI forecast of cars on the road.






B.l.b  Modification of the Model for Trucks



     For trucks, it was necessary to modify the car forecasting




procedure because DRI does not provide a forecast of the size of

-------
                                 B-12
 Table B-6  Predicted Number of Cars on the Road  1984-1992
 Year 1984   1985   1986   1987   1988
1989
1990   1991   1992
Age
Car
NEW
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
of

7950
9191
7392
8117
8526
9620
9545
8741
7345
4988
5993
5715
4317
3070
2525
1973
1539
1326
979
861
650
422
279
172
86
67
36
18
26
21
20


7950
10589
9145
7256
7915
8196
9154
8852
7842
6257
4342
5042
4716
3650
2543
2082
1626
1268
1092
807
709
536
348
230
142
71
55
30
15
21
17


7875
10589
10536
9016
7109
7651
7844
8554
8007
6787
5456
3675
4208
3993
3046
2111
1727
1349
1052
906
669
588
445
289
191
118
59
46
25
12
17


8100
10490
10536
10408
8852
6888
7341
7355
7765
6978
5923
4631
3083
3565
3344
2535
1756
1437
1122
875
754
557
489
370
240
159
98
49
38
21
10


8400
10789
10438
10396
10207
8565
6600
6870
6663
6741
6087
5019
3874
2611
2980
2779
2106
1459
1194
932
727
626
463
406
307
199
132
81
41
32
17


8625
11189
10735
10288
10182
9862
8194
6164
6211
5760
5878
5150
4186
3279
2178
2472
2304
1746
1210
990
773
603
519
384
337
255
165
109
67
34
27


8850
11489
11133
10558
10053
9812
9409
7625
5550
5328
5018
4959
4271
3541
2725
1801
2043
1904
1443
1000
818
639
498
429
317
278
211
136
90
55
28


9000
11788
11432
10936
10303
9673
9346
8737
6850
4740
4640
4226
4100
3612
2937
2250
1486
1685
1570
1190
825
675
527
411
354
261
229
174
112
74
45


9075
11988
11729
11219
10661
9901
9202
8664
7835
5830
4126
3903
3485
3466
2990
2421
1853
1224
1388
1293
980
679
556
434
338
292
215
189
143
92
61
Tot 111510 112498 113950  115769  117741 119876  122011 124188 126232

-------
                                B-13


the truck fleet.  Rather than attempt to develop an independent

estimate of the total size of the fleet, we assumed that the year

to year survival rates were a function of the scrappage rates

observed in the car fleet.  Since these can he derived from the

DRI forecast and from historical data, we can estimate the size

of the fleet by adding trucks surviving from the previous year to

new trucks delivered.  The precise details are given below.

     We regressed truck survival rates on car scrappage rates

and the fraction of the fleet that the truck cohort of age i was

of all trucks.  Thus,

(B.3)       SijT = a + bSRjfCars + c(Tij/SUM TIj) + e
                                          i
Because the truck fleet data for 1980 reflect the reclassifi-

cation of all passenger vans from cars to trucks, we excluded

the survival rates from 1979 to 1980 from the data base used for

the regressions.  The regression results are presented in Table

B-7.  The regression coefficients for SRj are used to adjust

the survival rates in forecast years for each truck cohort.

Aggregate fleet size is the sum of the individual cohorts plus

new truck sales for the year.  As in the car model, new trucks

on the road were assumed to be 75 percent of model year sales

and one year-old trucks were assumed to be 99.9 percent of orig-

inal sales.   Trucks on the road were extended to 30 years of

age using the same process as for cars.  The extrapolation

resulted in a 17 percent overestimate of the fleet between 16

and 30 years of age.  The adjustment coefficient (r)  in equation

B.2 was -.019, which resulted in a  25 percent reduction in the

oldest cohort.

-------
                                B-14
Table B-7  Regression Coefficients for Truck Year to Year
           Survival Rates
Survival
Rate
New-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
9-10
10-11
11-12
12-13
13-14
14-15
Constant
1.227
1.018
1.000
.9843
.9504
.9824
.9799
.9589
.9509
.9503
.8771
.8680
.8721
.8759
.8784
Scrappage
Rate
.007742
-.004245
-.000509
-.000364
.001117
-.000491
-.000722
.000908
-.002385
-.001290
-.001353
-.002685
-.001531
-.002938
-.003051
Cohort
Weight
-1.079
.1060
-.08927
.05054
.2416
-.02511
-.06452
.1289
.2915
-.07566
1.326
1.813
1.483
1.756
1.781

-------
                                B-15





     The  initial set of survival rates for trucks for 1983-84 was



taken  to  be  the same as the 82-83 rates.  However, the rates for



cohorts age  16 and greater were not set equal to the 14-15 rate



for 1983  for two reasons.  First, in the inventory extrapolation,



it was necessary to revise the older cohorts downward significantly



to make them equal to the number of trucks 16 years of age and



older  in  1983.  Secondly, there was no overall forecast to con-



strain the year to year survival rates, so it was essential to



reflect the  information from the extrapolation process in the



survival  rates. The rates were set to decline from 0.93 to 0.82



as cohorts ranged from 16 to 30 years of age.  As seen in Table



B-8, even with this reduction, older trucks become about one



sixth  of  the fleet by 1992.  Further, the total number of trucks



increases substantially relative to cars during the forecast



period.



     Once total trucks on the road were determined, several



additional adjustments to the data were made.  First, total



trucks on the road had to be disaggregated by weight class.



Trucks were  divided into six weight classes, derived from the



usual eight  classes by combining classes three through five into



one class.'  We did this because classes three(10,000 to 14,000



pounds),  four (14,000 to 16,000 pounds) and five (16,000 to



19,500 pounds) are very small.  Currently, they constitute less



than one  tenth of 1 percent of all trucks sold.



     The  adjustment procedure required the weight composition of



new trucks on the road in every year from 1954 to 1992.  For the



years from 1968 through 1983, we used published data on truck

-------
                                B-16
Table
Year
B-8
1984
Predi.ted Trucks on the Road 1984-1992
198F
1986
1987
1988
1989
1990
1991
1992
Age of
Truck
New
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

3139
3139
2189
2084
2130
3700
3464
3122
2500
1764
2249
2114
1736
1202
1103
1090
801
731
591
525
415
314
232
173
120
117
95
59
61
60

334 )
40 8 i
310.1
2165
2058
2084
3591
3345
2975
2333
1673
2109
1981
1613
1115
1017
1000
731
663
533
471
371
277
202
149
103
99
79
48
50

3275
4299
4034
3069
2137
2014
2024
3469
3192
2779
2214
1571
1978
1845
1500
1030
935
914
664
600
480
421
328
243
175
127
87
82
65
39

3468
4243
4250
3991
3029
2092
1956
1955
3312
2982
2638
2080
1474
1843
1716
1386
947
855
831
601
539
429
372
286
210
149
107
72
68
53

3670
4499
4194
4205
3939
2964
2031
1890
1867
3094
2830
2477
1952
1373
1714
1586
1274
866
111
751
540
482
379
325
248
179
126
90
60
55

3830
4757
4447
4150
4150
3856
2879
1962
1804
1744
2936
2658
2325
1818
1277
1584
1458
1165
788
703
676
483
426
332
281
212
151
105
74
48

3995
4963
4702
4400
4096
4062
3744
2780
1873
1685
1655
2757
2494
2165
1690
1180
1456
1333
1059
712
632
604
427
373
286
240
179
126
87
60

4148
5171
4905
4652
4343
4009
3944
3616
2654
1749
1599
1554
2587
2323
2012
1561
1084
1330
1211
957
640
565
534
373
322
245
203
149
104
71

4295
5368
5111
4853
4592
4250
3892
3809
3451
2479
1660
1501
1458
2409
2158
1859
1435
991
1209
1095
861
572
499
467
323
275
206
169
123
85
Tot 41020  43360  45590  47938  50440  53078  55812  58614   61453

-------
                                B-17






sales by weight class.  These data were converted to percentages



and applied to new trucks on the road to yield estimates of the



fleet by weight.  For the years before 1968, we assumed a constant



composition of the fleet at 1968 levels.  For the years after



1983, we used DRl's decomposition of new truck sales into



light trucks (those under 14000 pounds) and medium and heavy



trucks (those over 14,000 pounds).   We broke the light truck



category into two classes light duty trucks classes 1 and 2.   We



used the 1981-1983 average ratio of class 1 to class 2 trucks and



applied it to the DRI light trucks forecast to derive these two



categories of trucks.  We assumed there were no trucks between 10



and 14,000 pounds (1983 sales were 145). Similarly,  we applied



the average ratios for the other four weight categories to get a



breakdown of the DRI forecast of medium and heavy trucks.  The



resulting percentage composition of the fleet is presented in



Table B-9.



     The other adjustment that we made in the data was to estimate



the fraction of each class of truck that was diesel; we removed



these from the fleet.  For the years 1968 through 1983, we esti-



mated the fraction of truck sales that were diesel from actual



data.  Because it is difficult to find a coherent set of data on



all aspects of truck sales, we used total factory sales of diesels



for the numerator and total sales of trucks for the denominator;




this may understate the diesel fraction somewhat.  For the years



before 1968 we used the data in column two of Table B-10 as the




fraction of each of the six truck classes that were diesel.  These



correspond to the actual ratios in 1965.

-------
                                B-18
Table B-9  Fraction of Truck  Fleet  in  Each Weight Class
Year
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
1974
1973
1972
1971
1970
1969
1968
1967
1966
1965
1964
1963
1962
1961
1960
1959
1958
1957
1956
1955
1954
LDTl
.4874
.4874
.4874
.4865
.4862
.4848
.4833
.4815
.4814
.4824
.4879
.4899
.4850
.4901
.4543
.4415
.3929
.3409
.3746
.4331
.4684
.5635
.5825
.5781
.5828
.5615
.5830
.5992
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
.5812
LDT2
.4448
.4448
.4448
.4440
.4437
.4424
.4410
.4395
.4393
.4403
.4453
.4471
.4453
.4276
.4309
.4368
.4865
.5467
.5172
.4602
.4048
.2682
.2556
.2389
.2369
.2373
.2106
.2035
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
.2171
10,000
-19,500
.0007
.0007
.0007
.0007
.0008
.0008
.0008
.0008
.0009
.0008
.0007
.0007
.0005
.0011
.0014
.0027
.0065
.0209
.0128
.0172
.0142
.0097
.0239
.0336
.0442
.0452
.0514
.0536
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
.0500
19,500
-26,000
.0182
.0182
.0181
.0186
.0188
.0195
.0203
.0212
.0212
.0207
.0179
.0169
.0172
.0197
.0365
.0402
.0451
.0398
.0469
.0502
.0675
.0823
.0682
.0744
.0644
.0736
.0766
.0745
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
.0749
26,000
-33,000
.0199
.0199
.0198
.0204
.0205
.0213
.0222
.0232
.0232
.0227
.0196
.0185
.0219
.0278
.0261
.0262
.0153
.0105
.0082
.0073
.0098
.0126
.0142
.0173
.0177
.0227
.0173
.0221
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
.0207
33,000
plus
.0291
.0290
.0290
.0298
.0300
.0312
.0324
.0338
.0340
.0331
.0286
.0270
.0301
.0337
.0509
.0526
.0536
.0413
.0404
.0320
.0354
.0636
.0557
.0577
.0539
.0597
.0610
.0472
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560
.0560

-------
                                B-19






Table B-10.  Fraction of Trucks that are Diesel by Weight Class
Weight
Class
0-6,000
6,000-10,000
10,000-19,500
19,500-26,000
26,000-33,000
33,000+
Before
1968
0.0
0.001
0.05
0.06
0.45
0.80
After
1983
0.02
.10 to .135
.00
.24
.54
.975
For the years after 1983, we used the values given in column 3



of Table B-10 as the fraction of the fleet that will be diesel.



These forecasts are the result of an inspection of historical



data on diesel sales ratios and an overall forecast that the



price of gasoline and diesel fuel relative to other goods is not



going to rise significantly.   Class two trucks are the only



class that gives any indication of a growth in diesel penetration



at the current time.  Therefore, we increased diesel penetration



rates by 0.05 per year for this class.





B.2  Miles Driven per Year by Vehicle Class



     Data for cars and light duty trucks on miles driven per year



are taken directly from the Transportation Energy Data Book



(Department of Energy, 1982).  For heavier trucks, only average



miles per year for all trucks in the class are available in this



reference.  To develop a representation of how miles per year



declined with the age of these vehicles, we used data from

-------
                                B-20






Jambekar and Johnson (1978), and normalized this to the average




found in the DOE report.



     An important division of gasoline fueled vehicles for our



purposes was the extent to which vehicles designed for unleaded



fuel were misfueled with leaded gasoline.  This was important to



compute both the maintenance benefits and the benefits due to the



reduction of emissions from engines that would have been misfueled



in the absence of the rule.  To calculate this, we estimated the



extent of misfueling by vehicle age from EPA's 1983 survey of



tampering and misfueling.   This data was quite variable when



broken down by age of vehicle, so we smoothed it by fitting



a regression.  The resulting regression is



        MSFR = .0326 + .02314 T



where MSFR is the misfueling rate by vehicle age and T is the age



of the vehicle (T = 0 for  new vehicles).   The truck sample size



in the survey was too small to support a separate estimate for



them.  Therefore, we used  the same regression for trucks and



cars although the available data suggests that trucks misfuel at



a higher rate than cars.






B.3 Total Miles per Year for Maintenance Benefits



     For maintenance benefits we had to compute the total mileage



driven by both misfuelers  and legal leaded-gasoline users in a



given year.   Miles driven  by misfuelers were estimated by com-



puting the fraction of misfuelers in each model year using the



regression given above.  Cars and light duty trucks over 10 years



of age were assumed to experience no further increase in misfueling.

-------
                                    B-21






For legal leaded users, we used estimates of the legal leaded



fleet and the corresponding estimates of miles diriven per year.



In Table B-ll,  we present the estimated miles per year traveled



by cars and by light duty trucks classes 1 and 2 by leaded,



misfueled, and unleaded status.  Because we computed no maintenance



benefits for heavier trucks or other gasoline powered vehicles,



no estimates of usage are provided in this table.  The results



reflect the decline in the size of the fleet of legal leaded-gas-



oline users; the results also show how miles traveled by misfuelers



are projected to grow over time.






B.4  Miles per Year for Conventional Pollutant Emissions



     Emissions due to misfueling constitute the other major source



of benefits calculated on the basis of the fleet model results.



For conventional benefits estimates, we had to calculate the



increase in misfueling for each model year in a given calendar



year.   For each year, the incremental increase in misfueling



for a given model year is given by the slope coefficient.   Since



we restricted the increase in overall misfueling so that misfueling



was constant for vehicles 10 years of age and older, there is no




increase in misfueling for cars more than 10 years of age and,



thus, no conventional pollutant benefits are attributed to these



vehicles.  We then computed the expected number of miles the



additional misfuelers will travel each year for the next 20




years.  These estimates are corrected to reflect expected loss



rates and reductions in miles traveled as the vehicle ages.  The



resulting estimates are present valued to the year of initial

-------
                                B-22
Table B-ll
Mileage Estimates by Misfueling Status and Type of
Vehicle (millions of miles per year)	
Year
  1984
1985
1986
1987
1988
1989
LEGAL LEADED
CARS
LDT1
LDT2
MISFUELED
CARS
LDT1
LDT2
UNLEADED
CARS
LDT1
LOTS
Year
LEGAL LEADED
CARS
LDT1
LDT2
MISFUELED
CARS
LDT1
LDT2
UNLEADED
CARS 1
LDT1
LDT2
312.454
50.772
77.076
119.747
13.264
6.612
861.141
110.594
70.709
1990
76.682
22.638
55.804
181.770
27.194
17.965
162.103
199.305
136.531
250.001
44.513
71.901
132.948
15.494
8.413
921.376
128.196
84.593
1991
59.441
19.645
54.239
189.069
29.746
19.874
1200.156
213.001
146.108
199.925 159.257
39.015 34.124
67.040 62.740
144.940 155.805
17.742 20.020
10.312 12.299
974.934 1025.857
142.936 157.037
96.027 106.681
1992
46.191
17.092
53.122
195.837
32.328
21.703
1233.052
226.255
154.995
125.563 98.420
29.796 26.000
59.971 57.729
165.381 173.989
22.351 24.724
14.185 16.062
1074.616 1120.383
171.227 185.365
116.523 126.642





-------
                                B-23






misfueling using a 10 percent discount rate.  Total lifetime



miles driven by new misfuelers in each year are given in Table



B-12.  The total miles are then multiplied by estimates of the



increase in emissions due to poisoning of the catalyst.  These



increases vary by model year and are given in Table VI-1.  The




resulting increases in total emissions are given in Table VI-3.

-------
                                    B-24
Table B-12  Discounted Mileage Traveled in Following Twenty Years
            by Vehicles that Initially Misfuel in Given Year
            (millions of vehicle miles)
Year
1985
1986
1987
1988
1989
1990
1991
1992
Cars
132,104
135,764
139,178
143,644
147,955
152,671
157,174
161 ,523
LDT1
17,917
19,309
20,812
22,398
23,899
25,710
27,478
29,134
LDT2
12,032
13,509
14,973
16,392
16,948
17,986
19,046
20,076

-------
                            APPENDIX C

                        REGRESSION RESULTS


     This appendix consists of four sectiions.  Part C.I contains

of a description of the variables used in the regressions reported

in later sections.  Part C.2 contains logistic regressions for

white and black preteens, obtained using PROC LOGIST, a statistical

software package available within the Statistical Analysis System

(SAS).  The third section, Part C.3, presents linear regression

results for whites and blacks obtained by application of SURREGR,

a statistical routine available as an adjunct of SAS which corrects

regression results for the complex sample survey design used in

the NHANES data collection.  Part 4 of appendix C contains regres-

sion results where the solder content of cans has been added to

the variables used in the gas lead blood lead regressions reported

in Chapter Three .


C.I Variable Descriptions

     In addition to the regressions shown in Chapter Three, we

have used the regressions presented in this appendix in our

forecasts of child health effects.  We have used the following

variables in these regressions:

Variable Name                     Description

Gaslead                    Lead Used in gasoline, in hundreds of
                           metric tons a day, lagged one month.

Poor                       1 if Income l(see below); 0 otherwise

Age 1                      1 is age >^ 6 months and < 2 years;
                           0 otherwise

Age 2                      1 if age ^> 2 years and < 4 years;
                           0 otherwise

-------
                                C-2
Age 3


Age 4


Age 5


Age 6


Age 7


Teen


Income 1


Income 2


Male

Teen Male


Adult Male


Small City


Rural



Drinker


Heavy Drinker
Northeast, Mideast,
South
1 if age _> 4 years and < 6 years;
0 otherwise

1 if age >_ 6 years and < 8 years;
0 otherwise

1 if age >_ 8 years and < 10 years;
0 otherwise

1 if age _> 10 years and < 12 years;
0 otherwise

1 if age ^> 12 years and < 14 years;
0 otherwise

1 if age _> 14 years and < 18 years;
0 otherwise

1 if family income < $6,000; 0
otherwise

1 if family income > $6,000 and <
$15,000; 0 otherwise

1 if gender is male; 0 otherwise

1 if gender is male and age >^14 and
< 18 years; 0 otherwise

1 if gender is male and age >^18 years;
0 otherwise

1 if residence is in city with
population _> 1,000,000; 0 otherwise

1 if residence is in a rural area as
defined by the Bureau of the Census;
0 otherwise

1 if alcohol consumption is _>. 1 drink/
week and < 1 drink/day; 0 otherwise

1 if alcohol consumption is _>. 1
drink/day; 0 otherwise      ~~

1 if in this Census Region; 0 otherwise

-------
                                C-3
Education                  0 if the person never completed grade
                           school; 1  if grade school was the highest
                           level completed; 2 if high school was
                           the highest level completed; and 3 if
                           college was completed

Kid                        1 if age < 6; 0 otherwise

-------
                                       C-4

C.2  Logistic Regression Results

Black Preteens = 8-13 years old, 112 observations

Dependent variable;  1 if blood lead is over 20 ug/dl; 0 otherwise

  Model Chi-square =6.42 with 4 D.F-
Variable
Intercept
Gaslead
Poor
Age 5
Age 6
Beta
-6.0148
0.9786
0.2356
0.6158
0.2397
Std. Error
2.4044
0.4943
0.5289
0.6304
0.6208
Chi-Square
6.26
3.92
0.20
0.95
0.15
P-Value
0.0124
0.0477
0.6560
0.3286
0.6994
Fraction of concordant pairs of predicted probabilities and responses = 0.656


White Preteens = 8-13 years old, 660 observations

Dependent variable:  1 if blood lead is over 20 ug/dl; 0 otherwise

  Model Chi-square = 21.35 with 4 D.F.
Variable
Intercept
Gaslead
Poor
Age 5
Age 6
Beta
-8.9395
1.0674
0.8355
1.4199
1.2041
Std. Error
1.6782
0.3374
0.4883
0.5810
0.5904
Chi-Square
28.38
10.01
2.93
5.97
4.16
P-Value
0.0000
0.0016
0.0871
0.0145
0.0414
Fraction of concordant pairs of predicted probabilities and responses = 0.710
All logistic regression results were run using PRDC LOGISTIC within the Statistical
Analysis System (SAS).  This procedure uses individual data where the dependent
variable is 1 if the individual is above the threshold, and 0 otherwise.

-------
                                       C-5
APPENDIX C.3.  Linear Regression Results'
Whites :
Dependent variable:
Variable
Intercept
Gaslead
Income 1
Income 2
Age 1
Age 2
Age 3
Age 4
Teen
Male
Teen Male
Adule Male
Small City
Rural
Drinker
Heavy DrinKer
Northeast
Midwest
South
Education Level

individual
Beta
5.4436
2.1835
0.7675
0.3381
3.2352
4.0452
3.2020
2.1818
-0.7386
0.5763
1.7556
3.9812
-0.8490
-1.3215
0.8582
2.0871
-1.0908
-1.2243
-1.0598
-0.9440

blood lead levels
Std. Error
1.1842
0.0345
0.0553
0.0288
0.2015
0.1713
0.1267
0.2118
0.0519
0.1040
0.2150
0.1203
0.1080
0.1188
0.0296
0.0889
0.1302
0.1631
0.2493
0.0182


F-Statistic
—
138.19
10.65
3.97
51.95
95.51
80.91
22.48
10.52
3.19
14.34
131.72
6.67
14.70
24.84
48.97
9.14
9.19
4.51
48.90


P-Value
—
0.0000
0.0026
0.0548
0.0000
0.0000
0.0000
0.0000
0.0028
0.0834
0.0006
0.0000
0.0146
0.0006
0.0000
0.0000
0.0049
0.0048
0.0416
0.0000
* Procedure used is SURREGR in SAS.

-------
                                 06
Whites:
Dependent variable:
Variable
Intercept
Gas lead
Income 1
Income 2
Kid
Teen Male
Rural
Small City
Teen
Male
Adult Male
Age 4
Age 5
Age 6
Age 7
Drinker
Heavy Drinker
Northeast
Midwest
South
Education Level
individual
Beta
5.4593
2.1821
0.7542
0.3386
3.2344
2.0860
-1.3350
-0.8443
-1.5987
1.1333
3.4231
1.8952
0.5581
0.4784
0.3958
0.8672
2.0789
-1.0823
-1.2414
-1.0619
-0.9461
blood lead levels
Std. Error
1.1766
0.0344
0.0559
0.0284
0.0926
0.2093
0.1221
0.1098
0.0910
0.0348
0.0504
0.2205
0.1126
0.1629
0.0727
0.0303
0.0894
0.1312
0.1663
0.2504
0.1808

F-Statistic
—
138.53
10.17
4.04
112.97
20.79
14.59
6.49
28.08
36.90
232.33
16.29
2.77
1.41
2.15
24.92
48.35
8.92
9.27
4.50
49.51

P-Value
—
—
0.0000
0.0032
0.0531
0.0000
0.0001
0.0006
0.0159
0.0000
0.0000
0.0003
0.1060
0.2445
0.1520
0.0000
0.0000
0.0054
0.0046
0.0417
0.0000

-------
                                       C-7
Blacks:
Dependent variable:
Variable
Intercept
Gaslead
Income 1
Income 2
Age 1
Age 2
Age 3
Age 4
Teen
Male
Adule Male
Drinker
Heavy Drinker
Education Level
individual
Beta
4.8847
1.9342
1.1457
1.0941
6.1030
8.8867
6.6989
4.8920
0.6352
1.8280
4.2469
1.0359
1.4088
-0.8329
blood lead levels
Std. Error
2.4116
0.1432
2.2593
0.2902
1.3729
0.5052
0.4592
0.7706
0.1869
0.3413
0.6157
0.4713
1.2531
0.0874

F-Statistic
—
26.12
5.06
4.13
27.13
156.32
97.73
31.06
2.16
9.79
29.29
2.28
1.58
7.93

P-Value
—
0.0000
0.0328
0.0522
0.0000
0.0000
0.0000
0.0000
0.1533
0.0042
0.0000
0.1429
0.2190
0.0090

-------
                                       C-8
Blacks:
Dependent variable:
Variable
Intercept
Gas lead
Income 1
Income 2
Kid
Teen
Male
Adule Male
Age 4
Age 5
Age 6
Age 7
Drinker
Heavy Drinker
Northeast
Midwest
South
Education Level
individual
Beta
4.795
2.041
1.016
1.063
7.204
-0.806
1.860
4.061
4.869
2.494
2.215
0.417
1.063
1.386
-1.460
0.145
-0.1173
-0.826
blood lead levels
Std. Error
2.48
0.12
0.26
0.33
0.29
0.35
0.24
0.48
0.81
1.10
0.44
0.59
0.44
1.17
0.84
1.05
0.501
0.086

F-Statistic
—
33.84
3.90
3.44
180.91
1.84
14.39
34.08
29.22
5.67
11.07
0.30
3.03
1.64
2.53
0.02
0.03
7.91

P-Value
—
0.0000
0.0587
0.0748
0.0000
0.1857
0.0008
0.0000
0.0000
0.0246
0.0025
0.5910
0.0933
0.2117
0.1230
0.8885
0.8695
0.0091

-------
                                        C-9
APPENDIX C.4.   Regression Results:  Whites with Solder Used  in Cans
Effect                    Coefficient
Intercept
Gasoline Lead
Low  Income
Moderate Income
Child
Solder
Teenager
Male
Male Teenager
Adult Male
Small City
Rural
Drinker
Heavy Drinker
Northeast
South
Midwest
Education  Level
•6U.S. GOVERNMENT PRINTING OFFICE:  1965  526  259  30280
Standard Error
P-Value
4.16
2.16
0.78
0.36
3.47
0.74
-0.32
0.68
1.38
3.89
-0.82
-1.32
0.83
2.09
-1.02
-1.15
0.88
-0.93
—
0.04
0.06
0.03
0.11
1.01
0.05
0.08
0.13
0.11
0.11
0.12
0.03
0.09
0.13
0.17
0.29
0.02
—
0.0000
0.0024
0.0383
0.0000
0.4651
0.1667
0.0251
0.0005
0.0000
0.0212
0.0007
0.0000
0.0000
0.0091
0.0084
0.1093
0.0000

-------